Runs - Precision Medicine 2017¶

1_ec_simple¶

Run ID: 1_ec_simple
Participant: TREC_UB
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 2be3834463d1e97d60d423e977f4d17d
Run description: This is the first run using Apache Lucene and BM25.

2_ec_complex¶

Run ID: 2_ec_complex
Participant: TREC_UB
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 2ea9965fa0d89a1e4ead4c9eb6432c25
Run description: This is the second more complex run using Apache Lucene and BM25.

aCSIROmedAll¶

Run ID: aCSIROmedAll
Participant: CSIROmed
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: abstracts
MD5: 9111539b371a3210da833265925037e0
Run description: Two sets of abstracts indexed separately merged using Round Robin Federated Search Algorithm. Query processing includes: Gene Expansion, Wikipedia/Metamap Filter, Word Embedding Query Expansion, Age expansion, Gene boost matching, Gender Expansion.

aCSIROmedMCB¶

Run ID: aCSIROmedMCB
Participant: CSIROmed
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: abstracts
MD5: 800a14c753174351734ec3c23d316666
Run description: Federated search on indexes (abstract + medline) with query processing which include: Gene Expansion, Wikipedia/Metamap Filter, Word Embedding Query Expansion (Wikipedia+Medline), Age expansion, Gene boost matching, Gender Expansion using a Generic Document Scoring algorithm with weights without citation boosting.

aCSIROmedMGB¶

Run ID: aCSIROmedMGB
Participant: CSIROmed
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: abstracts
MD5: bb3db3cb5a72f75dbcd2fd7d86db399e
Run description: Federated search on indexes (abstract + medline) with query processing which include: Gene Expansion, Wikipedia/Metamap Filter, Word Embedding Query Expansion (Wikipedia+Medline), Age expansion, Gender Expansion using a Generic Document Scoring algorithm with weights and with citation boosting without gene boosting.

aCSIROmedNEG¶

Run ID: aCSIROmedNEG
Participant: CSIROmed
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: abstracts
MD5: 6c55503a4af51f7b118f469201a55686
Run description: Federated search on indexes (abstract + medline) with query processing which include: Gene Expansion, Wikipedia/Metamap Filter, Word Embedding Query Expansion (Wikipedia+Medline), Age expansion, Gene boost matching, Gender Expansion using a Generic Document Scoring algorithm with weights with negation.

aCSIROmedPCB¶

Run ID: aCSIROmedPCB
Participant: CSIROmed
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: abstracts
MD5: 9d8d2be7b6ad5a9df7c4b56e93da4bba
Run description: Federated search on indexes (abstract + medline) with query processing which include: Gene Expansion, Wikipedia/Metamap Filter, Word Embedding Query Expansion (Wikipedia+Medline), Age expansion, Gene boost matching, Gender Expansion using a Generic Document Scoring algorithm with weights and with citation boosting (higher).

Broad¶

Run ID: Broad
Participant: NaCTeM
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: eca7a1d4b5cca8e84c2f18ea9eba8542
Run description: This is our broad query, which is a combination of textual and semantics-based and aiming at maximize recall.

Broadc¶

Run ID: Broadc
Participant: NaCTeM
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 2937bf3833fffbe59821327da9b3ea92
Run description: This is our broad query, which is a combination of textual and semantics-based and aiming at maximize recall by looking at multiple fields of the clinical trial.

cbnuCT1¶

Run ID: cbnuCT1
Participant: cbnu
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 59aa963e9128b792bcb34b6569405d98
Run description: query expansion (extracting gene fields information of wikipedia articles)

cbnuCT2¶

Run ID: cbnuCT2
Participant: cbnu
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 1a96ba506be53bd4d87a70a450045bd2
Run description: query expansion (genes are extracted all fields information of wikipedia articles)

cbnuCT3¶

Run ID: cbnuCT3
Participant: cbnu
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 92891b9f112079f664b2b284937f8f73
Run description: pseudo relevance feedback using diseases document clusters

cbnuSA1¶

Run ID: cbnuSA1
Participant: cbnu
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: d5263529d75c603f5a6503d51d29605e
Run description: query expansion (extracting gene fields information of wikipedia articles)

cbnuSA2¶

Run ID: cbnuSA2
Participant: cbnu
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: e678913e27ad2d4b5d010528ceb5d380
Run description: query expansion (genes are extracted all fields information of wikipedia articles)

cbnuSA3¶

Run ID: cbnuSA3
Participant: cbnu
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 9126ef17857fe9dc679dd4620eb19a5a
Run description: pseudo relevance feedback using diseases document clusters

cCSIROmedAll¶

Run ID: cCSIROmedAll
Participant: CSIROmed
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: trials
MD5: 1e3606a8934aad8aba7bae8018e1254d
Run description: Query processing includes Cohort Matching, Gene Expansion, Wikipedia/Metamap Filter, Word Embedding Query Expansion, Age expansion, Gene boost matching, Gender Expansion.

cCSIROmedHGB¶

Run ID: cCSIROmedHGB
Participant: CSIROmed
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: trials
MD5: 4927cd0c0e4190d1a06eb2fd3a1a7828
Run description: Query processing includes Gene Expansion, Wikipedia/Metamap Filter, Word Embedding Query Expansion (Wikipedia+Medline), Age expansion, Gene boost matching (higher compared to other runs), Gender Expansion using a Generic Document Scoring algorithm with weights.

cCSIROmedMCB¶

Run ID: cCSIROmedMCB
Participant: CSIROmed
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: trials
MD5: 4ceef961a6b700c7d198e76ed6277546
Run description: Query processing includes Gene Expansion, Wikipedia/Metamap Filter, Word Embedding Query Expansion (Wikipedia+Medline), Age expansion, Gene boost matching, Gender Expansion using a Generic Document Scoring algorithm with weights without citation boost.

cCSIROmedMCM¶

Run ID: cCSIROmedMCM
Participant: CSIROmed
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: trials
MD5: ebb4520ddbb9c509ebad76a98fe186a3
Run description: Query processing includes Gene Expansion, Wikipedia/Metamap Filter, Word Embedding Query Expansion (Wikipedia+Medline), Age expansion, Gene boost matching, Gender Expansion using a Generic Document Scoring algorithm with weights without Cohort Matching.

cCSIROmedNEG¶

Run ID: cCSIROmedNEG
Participant: CSIROmed
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: trials
MD5: 876efef88e83644164d457e656a84a44
Run description: Query processing includes Cohort Matching, Gene Expansion, Wikipedia/Metamap Filter, Word Embedding Query Expansion (Wikipedia+Medline), Age expansion, Gene boost matching, Gender Expansion using a Generic Document Scoring algorithm with weights, and negation detection.

cRun1Bsl¶

Results | Participants | Input | Summary | Appendix

Run ID: cRun1Bsl
Participant: iris
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: b5acd85d7fe5bc93b007fda0f0248255
Run description: baseline, unigram terms from topics

cRun2BslOth¶

Results | Participants | Input | Summary | Appendix

Run ID: cRun2BslOth
Participant: iris
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: dbaf2304d741f0f30adf33dba80a01ec
Run description: baseline, unigram terms from topics, including "other" area

cRun3MRF¶

Results | Participants | Input | Summary | Appendix

Run ID: cRun3MRF
Participant: iris
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 7691a563449e06842a08b87638badc3a
Run description: MRF model, with extra syn names auto selected.

cRun4MRFrf¶

Results | Participants | Input | Summary | Appendix

Run ID: cRun4MRFrf
Participant: iris
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 6ec595bec422c66032021a0428160a43
Run description: MRF model, with extra syn names auto selected. With Relevance feedback model.

cRun5MRFBow¶

Results | Participants | Input | Summary | Appendix

Run ID: cRun5MRFBow
Participant: iris
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: f08e5f026227c792c66c6cee6feb3f13
Run description: MRF model, with extra syn names auto selected. With word2vec cos similarity.

DA_IICTrun1¶

Results | Participants | Input | Summary | Appendix

Run ID: DA_IICTrun1
Participant: DA_IICT
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: a7d37b333708169ac3593fc4ae0b29ae
Run description: DA_IICTrun1 is automatic retrieval by In_expC2 model with query expansion.

DA_IICTrun3¶

Results | Participants | Input | Summary | Appendix

Run ID: DA_IICTrun3
Participant: DA_IICT
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: a4e8b9a077cbb1888870756824d6bb35
Run description: DA_IICTrun3 is an automatic retrieval using BM25 model with query expansion.

DA_IICTrun4¶

Results | Participants | Input | Summary | Appendix

Run ID: DA_IICTrun4
Participant: DA_IICT
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: 18f7d22971bab759d8982493a038bb18
Run description: DA_IICTrun4 is an automatic retrieval using TF_IDF model with query expansion.

DA_IICTrun5¶

Results | Participants | Input | Summary | Appendix

Run ID: DA_IICTrun5
Participant: DA_IICT
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: 4a0bc421a158c79cdf75d8c01afc8f33
Run description: DA_IICTrun5 is an automatic retrieval using Hiemstra_LM model with query expansion.

dumbmethod¶

Results | Participants | Input | Summary | Appendix

Run ID: dumbmethod
Participant: SDSFU
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 25fc2d232cacf7a0da66c496ca687363
Run description: This run only use a single python filetaking consideration of disease,gene,gender,age and others. However,the evaluation criteria is quite simple,so I call this method dumbmethod.

DUTIR003¶

Run ID: DUTIR003
Participant: DUTIRL
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 9b58dc8fcfc74a9087f4e786db0941f5
Run description: This runs is using Lucene technology which a fixed frame. And this runs used the three model including input, process and output.

DUTIR004¶

Results | Participants | Input | Summary | Appendix

Run ID: DUTIR004
Participant: DUTIRL
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: fc036c6c46c127c452871b59601029cd
Run description: This runs is using Lucene technology which a fixed frame. And this runs used the three model including input, process and output.

ECNU_C_1¶

Results | Participants | Input | Summary | Appendix

Run ID: ECNU_C_1
Participant: ECNUica
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 560397ced578b6acbc8405d7f8a4f45d
Run description: This run is a combination of the results retrieved by BM25，BB2，PL2 and DFR_BM25 models. We use PV-DBOW model provided by Gensim to learn document embeddings of "Inclusion" part in documents. We also apply PV-DBOW to represent queries. After each retriveal model, we compute the similarities between inclusions and queries and re-rank the results of the retrieval model. We also rule out ineligible clinical trials.

ECNU_C_2¶

Results | Participants | Input | Summary | Appendix

Run ID: ECNU_C_2
Participant: ECNUica
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 9fecc73c85f3302380eb8b24af09627f
Run description: This run is retrieved by BM25 model. We translate queries into Chinese and use Chinese terms to do query expansion. The expansion terms are finally translated into English. We use terrier-4.0 to run the retrieval tasks.

ECNU_C_3¶

Results | Participants | Input | Summary | Appendix

Run ID: ECNU_C_3
Participant: ECNUica
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: f7aba9f998c7d933fa0f306ec1aafc7d
Run description: This run is a combination of the retrieval results of BM25, BB2 and PL2 models. The query is the content of the label "disease" in topics and query expansion is not applied in this run. We use terrier-4.0 to run the retrieval tasks. We also rule out ineligible trials by the criteria like age and gender.

ECNU_C_4¶

Results | Participants | Input | Summary | Appendix

Run ID: ECNU_C_4
Participant: ECNUica
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: bd9a91bab115d56ff8be4d6579e5e8e3
Run description: This run is retrieved by DFR_BM25 model. We use Word Vectors and Wiki data to do query expansion. We use terrier-4.0 to run the retrieval tasks.

ECNU_C_5¶

Results | Participants | Input | Summary | Appendix

Run ID: ECNU_C_5
Participant: ECNUica
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 13b0237d8e82472780e7117248e97af0
Run description: This run is a combination of two results of different query expansion strategies. The one is using MeSH to do query expansion and the other one is based on Google. We also rule out ineligible trials by the criteria like age and gender.

ECNU_M_1¶

Run ID: ECNU_M_1
Participant: ECNUica
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 774995377b2e2da29cf837f8f1efe3bf
Run description: This run is a combination of ECNU_M_2, ECNU_M_3, ECNU_M_4 and ECNU_M_5.

ECNU_M_2¶

Run ID: ECNU_M_2
Participant: ECNUica
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 6ba7c2b20f668966d3fe4716b3d823c0
Run description: This run is retrieved by BM25 model. We translate queries into Chinese and use Chinese terms to do query expansion. The expansion terms are finally translated into English. We use terrier-4.0 to run the retrieval tasks.

ECNU_M_3¶

Run ID: ECNU_M_3
Participant: ECNUica
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: e38d71985c44bd30f983c5be5e8dfde1
Run description: This run is retrieved by DFR_BM25 model. We use Word Vectors and Wiki data to do query expansion. We use terrier-4.0 to run the retrieval tasks.

ECNU_M_4¶

Run ID: ECNU_M_4
Participant: ECNUica
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: ed8d3613ee3f454000cb6677ffcd21f3
Run description: This run is a combination of several retrieval results whose queries are different labels of topics. The queries are expanded by Google search engine. We use terrier-4.0 to run the retrieval tasks.

ECNU_M_5¶

Run ID: ECNU_M_5
Participant: ECNUica
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: bf3f086c9009773840b95f942784606a
Run description: This run is a combination of the retrieval results of BM25, BB2 and PL2 models. The query is the content of the label "disease" in topics and query expansion is not applied in this run. We use terrier-4.0 to run the retrieval tasks.

eth_a_gws¶

Run ID: eth_a_gws
Participant: ETH
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 24d08d4e79df075ebcd9488a10cf6eff
Run description: In this run an initial list of documents was retrieved using LUCENE. These documents were solely scored using metrics gained from mutation code embeddings and adversarial training. Finally, the scores were combined in a weighted sum based on judged documents. Used Resources: http://cancer.sanger.ac.uk/census/ https://www.ncbi.nlm.nih.gov/gene http://www.genenames.org/

eth_a_luc¶

Run ID: eth_a_luc
Participant: ETH
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: e012d912e93143f52299c615ebd459d5
Run description: In this run an initial list of documents was retrieved using LUCENE. The scores for this run were solely retrieved from LUCENE. Used Resources: http://cancer.sanger.ac.uk/census/ https://www.ncbi.nlm.nih.gov/gene http://www.genenames.org/

eth_a_nn¶

Run ID: eth_a_nn
Participant: ETH
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: cd0c262a481b4e2c4f75b5cc67d475f2
Run description: In this run an initial list of documents was retrieved using LUCENE. These documents were additionally scored using metrics gained from mutation code embeddings and adversarial training. Finally, the scores were combined using a neural network classifier trained on judged documents. Used Resources: http://cancer.sanger.ac.uk/census/ https://www.ncbi.nlm.nih.gov/gene http://www.genenames.org/

eth_a_ws¶

Run ID: eth_a_ws
Participant: ETH
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: c2f312e8fa28e602eedcde46b74a0b56
Run description: In this run an initial list of documents was retrieved using LUCENE. These documents were additionally scored using metrics gained from mutation code embeddings and adversarial training. Finally, the scores were combined in a weighted sum based on judged documents. Used Resources: http://cancer.sanger.ac.uk/census/ https://www.ncbi.nlm.nih.gov/gene http://www.genenames.org/

eth_a_ws_q¶

Run ID: eth_a_ws_q
Participant: ETH
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 0c42a4177600b4efe118457841b849e2
Run description: In this run an initial list of documents was retrieved using LUCENE. These documents were additionally scored using metrics gained from mutation code embeddings and adversarial training. Finally, the scores were combined in a weighted sum to fit the qrels from judged documents. Used Resources: http://cancer.sanger.ac.uk/census/ https://www.ncbi.nlm.nih.gov/gene http://www.genenames.org/

eth_t_gwsq¶

Run ID: eth_t_gwsq
Participant: ETH
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: af420e667a0c8c1c569f6265d3b46826
Run description: In this run an initial list of documents was retrieved using LUCENE. These documents were solely scored using metrics gained from mutation code embeddings and adversarial training. Finally, the scores were combined in a weighted sum (with bias and quadratic terms) based on judged documents. Used Resources: http://cancer.sanger.ac.uk/census/ https://www.ncbi.nlm.nih.gov/gene http://www.genenames.org/

eth_t_luc¶

Run ID: eth_t_luc
Participant: ETH
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: e3cbea83134089fad1df5eb8c1630a6a
Run description: In this run an initial list of documents was retrieved using LUCENE. The scores for this run were solely retrieved from LUCENE. Used Resources: http://cancer.sanger.ac.uk/census/ https://www.ncbi.nlm.nih.gov/gene http://www.genenames.org/

eth_t_nn¶

Run ID: eth_t_nn
Participant: ETH
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: f7b83638308542d1364026f9f0e60c6c
Run description: In this run an initial list of documents was retrieved using LUCENE. These documents were additionally scored using metrics gained from mutation code embeddings and adversarial training. Finally, the scores were combined using a neural network classifier trained on judged documents. Used Resources: http://cancer.sanger.ac.uk/census/ https://www.ncbi.nlm.nih.gov/gene http://www.genenames.org/

eth_t_ws¶

Run ID: eth_t_ws
Participant: ETH
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 756040c801f7f6cfedbc3a0827a10bbb
Run description: In this run an initial list of documents was retrieved using LUCENE. These documents were additionally scored using metrics gained from mutation code embeddings and adversarial training. Finally, the scores were combined in a weighted sum based on judged documents. Used Resources: http://cancer.sanger.ac.uk/census/ https://www.ncbi.nlm.nih.gov/gene http://www.genenames.org/

eth_t_wsb_q¶

Run ID: eth_t_wsb_q
Participant: ETH
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: f837596505577e8ca5aaa0f47199f552
Run description: In this run an initial list of documents was retrieved using LUCENE. These documents were additionally scored using metrics gained from mutation code embeddings and adversarial training. Finally, the scores were combined in a weighted sum (with bias) to fit the qrels from judged documents. Used Resources: http://cancer.sanger.ac.uk/census/ https://www.ncbi.nlm.nih.gov/gene http://www.genenames.org/

Focused¶

Run ID: Focused
Participant: NaCTeM
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: e0b55ca0faf0dfc412dd09d34b92139e
Run description: This is our focused query, which uses a combination of textual- and semantics-based retrieval, and aiming at maximizing precision

Focusedc¶

Run ID: Focusedc
Participant: NaCTeM
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 792341d0dbd1e17791c27e68113d1af0
Run description: This is our focused query, which uses a combination of textual and semantics-based retrieval, and aiming at maximizing precision by only looking at the most critical fields of the clinical trials

GP14Medline¶

Run ID: GP14Medline
Participant: HokieGo
Track: Precision Medicine
Year: 2017
Submission: 7/29/2017
Type: automatic
Task: abstracts
MD5: e0a13747623c6289b69f1d8e9c7fb478
Run description: Building Medline re-rank model by Genetic Programming training on 2014 trec-cds data.

GP14Trail¶

Run ID: GP14Trail
Participant: HokieGo
Track: Precision Medicine
Year: 2017
Submission: 7/29/2017
Type: automatic
Task: trials
MD5: 110eed42a3a40a847760211d76dae760
Run description: Building Clinical Trail re-rank model by Genetic Programming training on 2014 trec-cds data.

GP16Medline¶

Run ID: GP16Medline
Participant: HokieGo
Track: Precision Medicine
Year: 2017
Submission: 7/29/2017
Type: automatic
Task: abstracts
MD5: 31129b0644c59fe80cf20102c5bf2747
Run description: Building Medline re-rank model by Genetic Programming training on 2016 trec-cds data.

GP16Trail¶

Run ID: GP16Trail
Participant: HokieGo
Track: Precision Medicine
Year: 2017
Submission: 7/29/2017
Type: automatic
Task: trials
MD5: 464bc1efb2d29a33a752af8bfe23a96a
Run description: Building Clinical Trail re-rank model by Genetic Programming training on 2016 trec-cds data.

Gwave¶

Run ID: Gwave
Participant: GravityWave
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: manual
Task: abstracts
MD5: c68fb7030466f16a17423b1f9b093676
Run description: In this run, we expanded the query. Indexed the corpus using TFIDF and used BM25 to retrieve the documents.

Gwave_ct¶

Results | Participants | Input | Summary | Appendix

Run ID: Gwave_ct
Participant: GravityWave
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: manual
Task: trials
MD5: 96d14a126467e99b9cbf2f68b6bc6cb2
Run description: Used Solr to index and retrieve this run.

ielabRun1¶

Run ID: ielabRun1
Participant: ielab-CSIRO-QUT
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 9c8f0c1000236d85b48f6ef34b2dae90
Run description: Base query + treatment field filter (Filtering based on if the document has a treatments field) Uses QuickUMLS
Code: https://bitbucket.org/ielab/prec-med-search/wiki/Home

ielabRun21¶

Run ID: ielabRun21
Participant: ielab-CSIRO-QUT
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 4ade1eccbfa33e97a884802b7d3ca26d
Run description: Base query + treatment field filter + treatment cui Uses QuickUMLS Elasticsearch
Code: https://bitbucket.org/ielab/prec-med-search/wiki/Home

ielabRun22¶

Run ID: ielabRun22
Participant: ielab-CSIRO-QUT
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 9dc28c08a09ffde9eb9d247ee8d13f3f
Run description: Base query + treatment field filter + treatment name Uses QuickUMLS Elasticsearch
Code: https://bitbucket.org/ielab/prec-med-search/wiki/Home

ielabRun23¶

Run ID: ielabRun23
Participant: ielab-CSIRO-QUT
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 2e5d33c80a496867aefedd360140bf38
Run description: Base query + treatment field filter + treatment cui + treatment name Uses QuickUMLS Elasticsearch
Code: https://bitbucket.org/ielab/prec-med-search/wiki/Home

ielabRun3¶

Run ID: ielabRun3
Participant: ielab-CSIRO-QUT
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 522b6be4baa952c627c692eb52411dc1
Run description: Disease-Treatment Association boost based on SemMedDB mentions for the treatment associated with the disease. Use of SemMedDB Uses QuickUMLS Elasticsearch
Code: https://bitbucket.org/ielab/prec-med-search/wiki/Home

KISTI01¶

Run ID: KISTI01
Participant: KISTI
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 49c0ea605506263e59bca5fdba70d351
Run description: dirichlet-smoothing language model

KISTI01CT¶

Results | Participants | Input | Summary | Appendix

Run ID: KISTI01CT
Participant: KISTI
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 6c82cc9cb24aeb64b2056b7ab0c09949
Run description: dirichlet-smoothing language model

KISTI02¶

Run ID: KISTI02
Participant: KISTI
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 99d8264339f12c166f4aa6786421779a
Run description: dirichlet-smoothing language model + relevance feedback model

KISTI02CT¶

Results | Participants | Input | Summary | Appendix

Run ID: KISTI02CT
Participant: KISTI
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: af490d5dc3d91f5b218dc6506740ffbc
Run description: dirichlet-smoothing language model + relevance feedback model

KISTI03¶

Run ID: KISTI03
Participant: KISTI
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: ecc883a65674dfbfbade8ed819a00efc
Run description: dirichlet-smoothing language model + random-walk feedback model

KISTI03CT¶

Results | Participants | Input | Summary | Appendix

Run ID: KISTI03CT
Participant: KISTI
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 7babf7c93f745406a31631bcbfeb4389
Run description: dirichlet-smoothing language model + random-walk feedback model

KISTI04¶

Run ID: KISTI04
Participant: KISTI
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 239a9301fb4007f3fe639141d2fa19c9
Run description: dirichlet-smoothing language model + external expansion model

KISTI04CT¶

Results | Participants | Input | Summary | Appendix

Run ID: KISTI04CT
Participant: KISTI
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: aaeb26fb02371db0e993ab69f9de3e12
Run description: dirichlet-smoothing language model + external expansion model

KISTI05¶

Run ID: KISTI05
Participant: KISTI
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: a652566f38afbd17c798a653ea3d941e
Run description: dirichlet-smoothing language model + random-walk feedback model + external expansion model

KISTI05CT¶

Results | Participants | Input | Summary | Appendix

Run ID: KISTI05CT
Participant: KISTI
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 2871493a70c3df20899aef9734c837cc
Run description: dirichlet-smoothing language model + random-walk feedback model + external expansion model

kkseabs1¶

Run ID: kkseabs1
Participant: kaist-kse
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: abstracts
MD5: d9f07a4c11cbcf0170924d8e0670af05
Run description: We filtered the PubMed documents through the detection of certain keywords in the content so that we only work with those that are related to cancer treatment. The extra abstracts were assumed to be all related to the task at hand, hence, no filtering process was done for this set. A search index was built based on each document's title, abstract, MeSH keywords (primary and secondary topics), keywords, and chemicals. We used the given disease's original term, synonyms, and parent concepts along with the specified genetic information and its synonym to retrieve the top 1,500 presumably relevant documents. From this initial set of documents, we performed an ad-hoc search to determine which ones have mentioned the co-morbidities listed in the others field of the patient's case. The initial set of documents were then re-ranked based on a relevance score that placed varying weights on the initial retrieval score, appearance of terms like 'treat', 'therapy', 'drug', etc. in the content, mention of the other diseases specified, and gender match. This run did not use the age of the patients.

kkseabs3¶

Run ID: kkseabs3
Participant: kaist-kse
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 09ce4d75d3e3668e76929d2c2a912e6c
Run description: For abstracts that did not come with associated MeSH terms, we derived the main topics by subjecting their titles to Metamap. We also detected concepts using the same tool in the abstract. The identified concepts were filtered using a list of pre-determined semantic types that are related to the task. We built an index of the documents based on the title and MeSH main topics. The retrieval algorithm used is Okapi-BM25 as implemented in Graphlab. The query was expanded by using the parent, child, and synonym relationships in UMLS.

kkseabs4¶

Run ID: kkseabs4
Participant: kaist-kse
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 524dc914ff7085d26500f6e25f60f2ff
Run description: We filtered out the PubMed documents that are not related to cancer treatment. For abstracts that did not come with associated MeSH terms, we derived the main topics by subjecting their titles to Metamap. We also detected concepts using the same tool in the abstract. The identified concepts were filtered using a list of pre-determined semantic types that are related to the task. We built an index of the documents based on the title and MeSH main topics. The retrieval algorithm used is Okapi-BM25 as implemented in Graphlab. The initial query was expanded by using the parent, child, and synonym relationships of the disease in UMLS (MRREL table). In the first retrieval, we derived the top 1,500 results. From this, we did two additional ad-hoc searches for the other conditions and the patient's demographic profile. We performed re-ranking on the initial results to take the later criteria into account.

kksetri1¶

Results | Participants | Input | Summary | Appendix

Run ID: kksetri1
Participant: kaist-kse
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: trials
MD5: 8915d70ec2bd8ae627de04985a39e922
Run description: We built the index based on the trials' title, brief summary, detailed description, condition, and MeSH terms. The initial query is comprised of the patient's disease, its synonyms, and parent concepts. The synonyms and parent concepts were derived from relationships in UMLS (MRREL table). Using this initial query, the top 1,500 presumably relevant documents are retrieved using the BM25 algorithm as implemented in GraphLab. We performed multiple adhoc searches on the initial set of relevant documents using the genetic information and its synonyms along with the comorbidities specified in the others field and their respective synonyms. The initial retrieved documents were re-ranked based on a linear combination of their scores in the initial search and the adhoc searches that followed. We set the coefficients/weights for each retrieval score (initial, gene, others, demographic) to reflect the relevance judgment mechanism for the precision medicine track.

kksetri2¶

Results | Participants | Input | Summary | Appendix

Run ID: kksetri2
Participant: kaist-kse
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: trials
MD5: f3706dac9770cb0a49894425fc646ebb
Run description: We filtered non-cancer related documents from the given set of clinical trials based on the presence of certain keywords/terms in the title and condition. This process was able to identify more than 40,000 cancer trials out of 200,000 treatments. The method is the same as the kksetri1 run but with a pre-filtered document pool and slightly different coefficients used in the re-ranking function.

kksetri4¶

Results | Participants | Input | Summary | Appendix

Run ID: kksetri4
Participant: kaist-kse
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: trials
MD5: 3f9714897ede6ae5575fedcd91446e46
Run description: We filtered out non-cancer related trials from the given document pool based on the presence of certain keywords/terms in the title and conditions specified in the trial. Metamap was used to extract biomedical concepts and their corresponding unique identifiers (CUIs) from the title, condition, and criteria. We defined a set of semantic types that are most likely related to cancer treatment and used it to filter out the biomedical concepts derived in the previous step. The criteria were parsed into their respective inclusion and exclusion parts. The Metamap tool can also detect negation, hence all negated concepts were automatically put in the exclusion part. The query preprocessing was done by first identifying the corresponding concept identifiers (CUIs) of the patient's disease and extracting its parent and child concepts in UMLS (MRREL table). We retrieved all trials where the CUIs of the disease, its synonyms, parents, and child concepts appeared in the concepts detected from the title. From this initial set of relevant documents, we performed an ad-hoc search using the genetic information and its synonyms on the title, brief summary, detailed description, and inclusion criteria. Another ad-hoc search was performed using the CUIs of the diseases specified in the others field and their respective synonyms. We automatically remove documents where the condition in the others field is part of the exclusion criteria. We also check for the matching of the age and gender in the given case and a retrieved trial. We use a rubric scoring mechanism to account for the matching of conditions, genetic information, inclusion criteria, and demographic profile. The scoring mechanism places more weight on the matching of disease (cancer type) and genetic information. Based on this, the initial documents are re-ranked and the top 1000 documents are included in the submission.

LDGprfStrict¶

Run ID: LDGprfStrict
Participant: POZNAN_SEMMED
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: cde1b7368707881cf60ccc57c0a7c3f8
Run description: File generated with data aggregation method over multiple Terrier runs (based on LGD scoring function) with various queries for a single topic. Exclusion criteria(eg. gender, age) are incorporated with use of dedicated method. Words which express gene or disease are required to appear within the documents. Query expansion applied for the gene and disease words. Pseudo relevance feedback is applied.

LGDnoprfGOpt¶

Run ID: LGDnoprfGOpt
Participant: POZNAN_SEMMED
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 6f3e810f82fbad43e0290d27617e23bc
Run description: File generated with data aggregation method over multiple Terrier runs (based on LGD scoring function) with various queries for a single topic. Exclusion criteria(eg. gender, age) are incorporated with use of dedicated method. Words which express disease are required to appear within the documents. Words which express gene are optional, but taken into account for scoring. Query expansion applied for the gene and disease words. No pseudo relevance feedback, which we belive should be an improvement over the appliance of the method.

LGDprfGOpt¶

Run ID: LGDprfGOpt
Participant: POZNAN_SEMMED
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 9316cc5e3fc5ade7bbd2d52762aa7d45
Run description: File generated with data aggregation method over multiple Terrier runs (based on LGD scoring function) with various queries for a single topic. Exclusion criteria(eg. gender, age) are incorporated with use of dedicated method. Words which express disease are required to appear within the documents. Words which express gene are optional, but taken into account for scoring. Query expansion applied for the gene and disease words. Pseudo relevance feedback is applied.

LGDraw¶

Run ID: LGDraw
Participant: POZNAN_SEMMED
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: ff7db14552151fdd9d742986dd1e0d0f
Run description: Simple LGD baseline file generated with Terrier. Exclusion criteria(eg. gender, age) incorporated with use of dedicated method.

LGDStrict¶

Run ID: LGDStrict
Participant: POZNAN_SEMMED
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: b0bcf1c8382c35f4cb2b523f77d5abf1
Run description: File generated with data aggregation method over multiple Terrier runs (based on LGD scoring function) with various queries for a single topic. Exclusion criteria(e.g. gender, age) are incorporated with use of dedicated method. Words which express gene or disease are required to appear within the documents. Query expansion applied for the gene and disease words.

mayonlpct1¶

Run ID: mayonlpct1
Participant: MayoNLPTeam
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 4b201cc6f6268885c2918fa37e35bed2
Run description: Exact keryword searching.

mayonlpct2¶

Run ID: mayonlpct2
Participant: MayoNLPTeam
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 242485a044676a094e1fcbbd724aa049
Run description: Use top (ten) related genes for each drug from semantic medline

mayonlpct3¶

Run ID: mayonlpct3
Participant: MayoNLPTeam
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 0de70dee129ea67a9aeb682204f90722
Run description: Search pubmed articles, get top 10 ranked (according to appearance) pseudo relevant terms for disease, gene, mesh terms

mayonlpct4¶

Run ID: mayonlpct4
Participant: MayoNLPTeam
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: f4f9616cd2a598feea80cb58a6ae6d18
Run description: Use the results from MeSH on Demand to search

mayonlpct5¶

Run ID: mayonlpct5
Participant: MayoNLPTeam
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 93b9018dbda09d79393949a7bb01928c
Run description: a combination of runs 2-4.

mayonlppm1¶

Run ID: mayonlppm1
Participant: MayoNLPTeam
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: abstracts
MD5: 5faae25209fd6e9c34cbef578fe28607
Run description: MRF model for topic terms. Abstract must contain disease & gene & (if has variant, must contain variant &(variant|variants)) from topics.

mayonlppm2¶

Run ID: mayonlppm2
Participant: MayoNLPTeam
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: abstracts
MD5: 48a4766887776cbd72c1d1eacf58b4b7
Run description: MRF model for topic terms. First, extract gene and disease entities and construct structured data. Second, search in the structured data.

mayonlppm3¶

Run ID: mayonlppm3
Participant: MayoNLPTeam
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: abstracts
MD5: 92f9982ab2e1bb600378b5ca7de8c14e
Run description: MRF model for topic terms. First, extract gene and disease entities and construct structured data. Second, search in the unstructured data. Third, rerank using structured data.

mayonlppm4¶

Run ID: mayonlppm4
Participant: MayoNLPTeam
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: abstracts
MD5: b124a85b3d2c7cef626921b294d12462
Run description: Pseudo relevance feedback. search in unstructured abstract, get mesh heading, gene mentions, and disease mentions for top ranked documents, add to query -> search in mesh heading, gene mentions, and disease mentions

mayonlppm5¶

Run ID: mayonlppm5
Participant: MayoNLPTeam
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 332f1dbc8603a1f25c188c9d8f6cb2ee
Run description: An ensemble model of using pseudo relevance feedback and re-ranking.

MedIER_ct1¶

Run ID: MedIER_ct1
Participant: MedIER
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: e4918cb28f3506ea14b2c81e2a7ff30b
Run description: normal query + filtered by age and gender MeshID for disease and gene comes from https://www.ncbi.nlm.nih.gov/gene

MedIER_ct2¶

Run ID: MedIER_ct2
Participant: MedIER
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: a12aa05d8b45ba0f39f05b1c46b9c706
Run description: pseudo relevance feedback query + filtered by age and gender MeshID for disease and gene comes from https://www.ncbi.nlm.nih.gov/gene

MedIER_ct3¶

Run ID: MedIER_ct3
Participant: MedIER
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 03ff06e1c909889141db4dc262374f43
Run description: normal query expanded to 3000 results + filtered by age and gender, stop at top 500 MeshID for disease and gene comes from https://www.ncbi.nlm.nih.gov/gene

MedIER_ct4¶

Run ID: MedIER_ct4
Participant: MedIER
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 6092e6ba7448748c964cc9f29fb97fd4
Run description: pseudo relevance feedback query expanded to 3000 results + filtered by age and gender, stop at 500 MeshID for disease and gene comes from https://www.ncbi.nlm.nih.gov/gene

MedIER_sa1¶

Run ID: MedIER_sa1
Participant: MedIER
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 014d369849612a3d0887c942578f28c1
Run description: normal query + MeshID; normal query + MeshID； MeshID for disease and gene comes from https://www.ncbi.nlm.nih.gov/gene

MedIER_sa2¶

Run ID: MedIER_sa2
Participant: MedIER
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: abstracts
MD5: 741a0cf0b3df13efd2e9898ecd2d63e0
Run description: MeshList of Disease, Gene. Use dataset from https://www.ncbi.nlm.nih.gov/gene to get the meshList id of disease and gene.
Code: https://github.com/tongyin04/TREC2017

MedIER_sa3¶

Run ID: MedIER_sa3
Participant: MedIER
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: b58f4aa1d8ec18f2eca0ffaa1baaae8f
Run description: pseudo relevance feedback query+ MeshID query MeshID for disease and gene comes from https://www.ncbi.nlm.nih.gov/gene

MedIER_sa4¶

Run ID: MedIER_sa4
Participant: MedIER
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: c314d46b034a84e9e9e29eb2af8e3252
Run description: normal query as baseline; MeshID for disease and gene comes from https://www.ncbi.nlm.nih.gov/gene

medline1¶

Run ID: medline1
Participant: FDUDMIIP
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 0e8fe97926787d56c5f82a399401e00d
Run description: Baseline of medline

medline2¶

Run ID: medline2
Participant: FDUDMIIP
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: f2ba281251d3ebb1e08e7abe221db965
Run description: medline expansion

medline3¶

Run ID: medline3
Participant: FDUDMIIP
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 7f3050268ce804954f91cfd71431f79f
Run description: medline

medline4¶

Run ID: medline4
Participant: FDUDMIIP
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 1cc3a742ca652f2d138b1efcff35efd2
Run description: medline

medline5¶

Run ID: medline5
Participant: FDUDMIIP
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 5646c4cc574221f9fca7d64660d3b248
Run description: medline

mRun1Bsl¶

Run ID: mRun1Bsl
Participant: iris
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: d568a7e2753bf1f3fd407b787285e9de
Run description: MRF model, with extra syn names auto selected. With word2vec cos similarity.

mRun2BslOth¶

Run ID: mRun2BslOth
Participant: iris
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 4b65503dd8470b2c52f6aff2d4e188ce
Run description: MRF model, with extra syn names auto selected. With word2vec cos similarity.

mRun3MRF¶

Run ID: mRun3MRF
Participant: iris
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 1fc752577f53b61d13f7bf03c3dfa4a5
Run description: MRF model, with extra syn names auto selected. MRF

mRun4MRFrf¶

Run ID: mRun4MRFrf
Participant: iris
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: bb40a1ee770dc71a6e1c015d91f6058b
Run description: MRF model, with extra syn names auto selected. MRF with RM3

mRun5MRFBow¶

Run ID: mRun5MRFBow
Participant: iris
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: b48bbf98ff436b3418d2cc9a9cb8b3a1
Run description: MRF model, with extra syn names auto selected. MRF with Word2vec cos similarity.

mugctbase¶

Run ID: mugctbase
Participant: imi_mug
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: trials
MD5: 18ad0b7166dd22b7d95a76eab65c5977
Run description: Baseline run checking age, sex + inclusion criteria (disease and gene). No exclusion criteria used.
Code: https://github.com/bst-mug/trec2017

mugctboost¶

Run ID: mugctboost
Participant: imi_mug
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: d89a6055110e96a23b3efce4192fad0e
Run description: Using positive boosters.
Code: https://github.com/bst-mug/trec2017/

mugctdisease¶

Run ID: mugctdisease
Participant: imi_mug
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 99b6aff960b7dcbf958f0b371e5ed57e
Run description: Expanding disesases using Lexigram API.
Code: https://github.com/bst-mug/trec2017/

mugctgene¶

Run ID: mugctgene
Participant: imi_mug
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 83e277a974abcfad274e34bfe5ad2f7d
Run description: Using gene synonym expanded in loco with NIH gene list.
Code: https://github.com/bst-mug/trec2017/

mugctmust¶

Run ID: mugctmust
Participant: imi_mug
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: e71d96deba29bf9969dc1f4e3489d841
Run description: New baseline forcing match of gene/disease and priorizing comorbidities in exclusion criteria.
Code: https://github.com/bst-mug/trec2017/

mugpubbase¶

Run ID: mugpubbase
Participant: imi_mug
Track: Precision Medicine
Year: 2017
Submission: 7/28/2017
Type: automatic
Task: abstracts
MD5: 7f6a0247b5aac6e87f3d0cd984cde430
Run description: Baseline query with basic synonyms and heuristic boosting.
Code: https://github.com/bst-mug/trec2017

mugpubboost¶

Run ID: mugpubboost
Participant: imi_mug
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 4369673f0bf4bff8ee1d0fb2fba92460
Run description: Using positive and negative boosters and regex for chemoterapics suffixes.
Code: https://github.com/bst-mug/trec2017/

mugpubdiseas¶

Run ID: mugpubdiseas
Participant: imi_mug
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: f02ff9d99d07491a5f2cfb51cd1b1850
Run description: Expanding disesases using Lexigram API.
Code: https://github.com/bst-mug/trec2017/

mugpubgene¶

Run ID: mugpubgene
Participant: imi_mug
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: e5d62ab5dc783d48c16551afaa3a8251
Run description: Using gene synonym expanded in loco with NIH gene list.
Code: https://github.com/bst-mug/trec2017/

mugpubshould¶

Run ID: mugpubshould
Participant: imi_mug
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 292602b2b68886ae1565a31a43dad4df
Run description: Relaxing criteria to "should" queries.
Code: https://github.com/bst-mug/trec2017/

NOVAsa1¶

Run ID: NOVAsa1
Participant: NOVASearch
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 77e345fb980703d52a57fd8723354e5f
Run description: Search (BM25L similarity) in the Medline/ASCO/AACR title and abstract text. Query is the disease, gene variant, patient demographics and existing conditions text, expanded by PRF using terms from the top 25 results. Query also expanded with synonyms, alternative and preferred terms from MeSH.

NOVAsa2¶

Run ID: NOVAsa2
Participant: NOVASearch
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: d32f725d3ed926bf82aa75ba611e9f3e
Run description: Search (multiple similarities) in the Medline/ASCO/AACR title and abstract text. Query is the disease, gene variant, patient demographics and existing conditions text, expanded by PRF using terms from the top 25 results. Query also expanded with synonyms, alternative and preferred terms from MeSH. Final rank is the fusion of runs using BM25L,BM25+,TF-IDF and Dirichlet Language Model similarities, using RRF.

NOVAsa3¶

Run ID: NOVAsa3
Participant: NOVASearch
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: b5f239e1af0a159051825e77a35e951c
Run description: Search (multiple similarities) in the Medline/ASCO/AACR title and abstract text. Query is the disease, gene variant, patient demographics and existing conditions text, expanded by PRF using terms from the top 25 results. Query also expanded with synonyms, alternative and preferred terms from MeSH and SNOMed. Final rank is the fusion of runs using BM25L,BM25+,TF-IDF and Dirichlet Language Model similarities using RRF.

NOVAtr1¶

Results | Participants | Input | Summary | Appendix

Run ID: NOVAtr1
Participant: NOVASearch
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 8d82129c2628943fc0de9bdbca4229ce
Run description: Search (BM25 similarity) in the trial title, summary and description text. Query is the disease and gene variant text.

NOVAtr2¶

Results | Participants | Input | Summary | Appendix

Run ID: NOVAtr2
Participant: NOVASearch
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 660f3f7a0788d049b81cd01c05632b41
Run description: Search (BM25 similarity) in the trial title, summary and description text. Query is the disease and gene variant text. Results filtering by the patient age and gender.

NOVAtr3¶

Results | Participants | Input | Summary | Appendix

Run ID: NOVAtr3
Participant: NOVASearch
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 4f11b23240565d52065836a3cc1adfb8
Run description: Search (BM25 similarity) in the trial title, summary and description text. Query is the disease and gene variant text, expanded by PRF using terms from the top 3 results. Results filtering by the patient age and gender.

NOVAtr4¶

Results | Participants | Input | Summary | Appendix

Run ID: NOVAtr4
Participant: NOVASearch
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: bafb9bf89572cbb5a324caa7426bbeb0
Run description: Search (BM25 similarity) in the trial title, summary and description text. Query is the disease and gene variant text. Filtered results by matching the patient age and gender to trial's criteria, and where the patient's existing conditions exclusion criteria matched the trails exclusion criteria.

NOVAtr5¶

Results | Participants | Input | Summary | Appendix

Run ID: NOVAtr5
Participant: NOVASearch
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 91601760e8cdf1ba927e4d0ac5323443
Run description: Search (BM25 similarity) in the trial title, summary and description text. Query is the disease and gene variant text, expanded by PRF using terms from the top 3 results. Filtered results by matching the patient age and gender to trial's criteria, and where the patient's existing conditions exclusion criteria matched the trails exclusion criteria.

Ontological¶

Run ID: Ontological
Participant: NaCTeM
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: manual
Task: abstracts
MD5: efd1a340cc619dbfb6d147fe6d77d411
Run description: This is our ontology-based query, where the UMLS ontology is used to expand the list of concepts to be retrieved. The selection of concepts from the ontology was performed manually.

Ontologicalc¶

Run ID: Ontologicalc
Participant: NaCTeM
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: manual
Task: trials
MD5: 8ec4ee2c5e5629ef67382b0730dfe2f7
Run description: This is our ontology-based query, where the UMLS ontology is used to expand the list of concepts to be retrieved. The selection of concepts from the ontology was performed manually.

pms_run1¶

Run ID: pms_run1
Participant: prna-mit-suny
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 38dc72698de0634014a9afc413f83bde
Run description: This run uses various ontologies such as MESH, GO etc. along with BM25 retrieval scheme to extract the oncology-relevant articles; extra abstracts were not included in this run to see their impact besides using other internal parameters and normalization factors.

pms_run1_tri¶

Run ID: pms_run1_tri
Participant: prna-mit-suny
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 65b13ddc511e749389f07e86ab5f99e6
Run description: This run uses various ontologies such as MESH, GO etc. along with BM25 retrieval scheme to extract the oncology-relevant trials; we use various internal parameters and normalization factors to see their impact on the results.

pms_run2_abs¶

Run ID: pms_run2_abs
Participant: prna-mit-suny
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 8ef4323de51d1f059ffcc612a3496a6a
Run description: This run uses hybrid clinical NLP-driven approaches to information retrieval by leveraging relevant ontologies/knowledge sources with appropriate statistical and heuristic techniques

pms_run2_tri¶

Run ID: pms_run2_tri
Participant: prna-mit-suny
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: f076fd375de9d5805b4d60a04d46a57c
Run description: This run uses hybrid clinical NLP-driven approaches to information retrieval by leveraging relevant ontologies/knowledge sources with appropriate statistical and heuristic techniques

pms_run3_abs¶

Run ID: pms_run3_abs
Participant: prna-mit-suny
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 6da26c0bce4dff4b7868990d93cda7d8
Run description: This runs uses an algorithm to combine the results of run2 and run5.

pms_run3_tri¶

Run ID: pms_run3_tri
Participant: prna-mit-suny
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: e184e562951f2f3c98c1fd65712b3ff7
Run description: This runs uses an algorithm to combine the results of run2 and run5.

pms_run4_abs¶

Run ID: pms_run4_abs
Participant: prna-mit-suny
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 34445d919a7acfd1d95f794c6fe36f98
Run description: This run uses hybrid clinical NLP-driven approaches to information retrieval by leveraging relevant ontologies/knowledge sources with appropriate statistical and heuristic techniques; we explore more strict matching here.

pms_run4_tri¶

Run ID: pms_run4_tri
Participant: prna-mit-suny
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: dfd21ae117b354e828bec717c8aeef21
Run description: This run uses hybrid clinical NLP-driven approaches to information retrieval by leveraging relevant ontologies/knowledge sources with appropriate statistical and heuristic techniques; we explore more strict matching here.

pms_run5_abs¶

Run ID: pms_run5_abs
Participant: prna-mit-suny
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 355d608c646d327deb3124bc3deeccaa
Run description: This run uses various ontologies such as MESH, GO etc. along with BM25 retrieval scheme to extract the oncology-relevant articles; extra abstracts were included in this run to see their impact besides using other internal parameters and normalization factors.

pms_run5_tri¶

Run ID: pms_run5_tri
Participant: prna-mit-suny
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 023bb4ce55187a16ad5dcfbd58df79cc
Run description: This run uses various ontologies such as MESH, GO etc. along with BM25 retrieval scheme to extract the oncology-relevant trials; extra abstracts were included in this run to see their impact besides using other internal parameters and normalization factors.

POZabsBB2GRn¶

Run ID: POZabsBB2GRn
Participant: POZNAN_SEMMED
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 7b9f2c7aa391c53b21e47fc5fc7a46d3
Run description: File generated with data aggregation method over multiple Terrier runs (based on BB2 scoring function) with various queries for a single topic. Exclusion criteria are omitted. Words which express gene or disease are required to appear within the documents. Query expansion applied for the gene and disease words.

POZabsBB2sn¶

Run ID: POZabsBB2sn
Participant: POZNAN_SEMMED
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: c52b8624f52384b53249298e9cc53ebf
Run description: LGD based retrieval with Terrier over the Medline abstracts corpus. No additional processing. Pseudo relevance feedback applied.

POZabsLGDGRn¶

Run ID: POZabsLGDGRn
Participant: POZNAN_SEMMED
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 3a3fa2a9a96addd0025e59a705d2efec
Run description: File generated with data aggregation method over multiple Terrier runs (based on LGD scoring function) with various queries for a single topic. Exclusion criteria are omitted. Words which express gene or disease are required to appear within the documents. Query expansion applied for the gene and disease words.

SDSFU_BASE¶

Results | Participants | Input | Summary | Appendix

Run ID: SDSFU_BASE
Participant: SDSFU
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: 81a705eb3bcdf2f432c9c4fe6c463456
Run description: This run use the information of topics, including disease, gene, demographic and other, to retrieve and score for clinical trials.

SDSFU_Ensem¶

Run ID: SDSFU_Ensem
Participant: SDSFU
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: 128b710bf10f02c3bfbb7705d6adffaa
Run description: BM25 with query expansion based on UMLS and pseudo feedback Using an ensemble version of several learn-to-rank models to rerank

SDSFU_Jnal¶

Run ID: SDSFU_Jnal
Participant: SDSFU
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: efbbac5be1f71d69db596eec406aea43
Run description: BM25 plus query expansion with UMLS and pseudo feedback Use journal frequency to revise the result.

SDSFU_Lambda¶

Run ID: SDSFU_Lambda
Participant: SDSFU
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: a17b31ec9c3111c4b70129ae082a5e5c
Run description: baseline:BM25 query expansion:UMLS qseudo feedback added rerank:LambdaMart

SDSFU_PF¶

Results | Participants | Input | Summary | Appendix

Run ID: SDSFU_PF
Participant: SDSFU
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: e865a56344d900da59a4bfb998617e00
Run description: This run use the information of topics to do query expansion via pseudo feedback, and then the query was used to retrieve and score for clinical trials.

SDSFU_PF_SA¶

Run ID: SDSFU_PF_SA
Participant: SDSFU
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: 2aad15c4e0a8794718f647477963b576
Run description: This run just use pseudo feedback to do query expansion, and then these terms and information of topics were use to retrieve and score for scientific abstracts. This run is the baseline of scientific abstracts.

SDSFU_PFUMLS¶

Results | Participants | Input | Summary | Appendix

Run ID: SDSFU_PFUMLS
Participant: SDSFU
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: 6b6dda2f5b39d3f311b17b431e1d803f
Run description: This run use the information of topics to do query expansion via UMLS and pseudo feedback. Then the these terms and information of topics were used to retrieve and score clinical trials.

SDSFU_PU_SA¶

Run ID: SDSFU_PU_SA
Participant: SDSFU
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: ad64393829602f07b1cbb637232cf3ff
Run description: This run use the information of topics to do query expansion via pseudo feedback and UMLS.

SDSFU_UMLS¶

Results | Participants | Input | Summary | Appendix

Run ID: SDSFU_UMLS
Participant: SDSFU
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: c160108d82af7aab4d2d192df2109b20
Run description: This run use disease and gene of topics to do query expansion via UMLS, and then the terms from UMLS and information of topics were used to retrieve and score for clinical trials.

Semantic¶

Run ID: Semantic
Participant: NaCTeM
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: ef54530990c5dbdfd93e379bc7594790
Run description: This is our semantic query, where semantic information extracted from data is used to retrieve documents

Semanticc¶

Run ID: Semanticc
Participant: NaCTeM
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 349388add99801e3d87720fd562abf2b
Run description: This is our semantic query, where semantic information extracted from data is used to retrieve documents

SIBTct1¶

Run ID: SIBTct1
Participant: BiTeM
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 40994c9f1549dde345c7f8796e212491
Run description: Demographic : filtering Genes : IR in the whole CT (except for exclusion criteria) Disease : use of NCI (no hierarchy), filtering in tag Other : MeSH matching and down-weighting reranking

SIBTct2¶

Run ID: SIBTct2
Participant: BiTeM
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 078f1df9fd620cf4433386c8d334cb40
Run description: Demographic : filtering Genes : IR in the whole CT (except for exclusion criteria) Disease : use of NCI (with children), filtering in tag Other : MeSH matching and down-weighting reranking

SIBTct3¶

Run ID: SIBTct3
Participant: BiTeM
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: ab5ce92598c69e345255ceb309688363
Run description: Demographic : filtering Genes : IR in the whole CT (except for exclusion criteria) Disease : use of NCI (with children), filtering in , and tags Other : MeSH matching and down-weighting reranking

SIBTct4¶

Run ID: SIBTct4
Participant: BiTeM
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 3c158aea3bd711b2ca6de6a4d63b01df
Run description: Demographic : filtering Genes : IR in the whole CT (except for exclusion criteria) Disease : use of NCI (with children and parents), filtering in , and tags Other : MeSH matching and down-weighting reranking

SIBTct5¶

Run ID: SIBTct5
Participant: BiTeM
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: c2fe40f4a7ee6ee2552df1eca61193bc
Run description: Demographic : filtering Genes : IR in the whole CT (except for exclusion criteria) Disease : use of NCI (with children and parents), filtering in all the CT Other : MeSH matching and down-weighting reranking

SIBTMlit1¶

Run ID: SIBTMlit1
Participant: BiTeM
Track: Precision Medicine
Year: 2017
Submission: 7/27/2017
Type: automatic
Task: abstracts
MD5: 06fd519198be1a4fba065b6525075317
Run description: Run 1 is our baseline run. Three queries, enabling constraint relaxing (disease+gene+variant; disease+gene; gene+variant) are sent to our search engine. Results are re-ranked based on drug density. Drug density is based on a list of drugs and products provided by DrugBank.

SIBTMlit2¶

Run ID: SIBTMlit2
Participant: BiTeM
Track: Precision Medicine
Year: 2017
Submission: 7/27/2017
Type: automatic
Task: abstracts
MD5: 4b8df3eb35d1b99b127e9904a8d4f6a5
Run description: Run 2 is based on run 1. Results are further re-ranked based on a list of positive keywords (e.g. treat, therap) and negative keywords (e.g. marker, sequencing), thus aiming to rank higher publications about treatment, prognosis and prevention.

SIBTMlit3¶

Run ID: SIBTMlit3
Participant: BiTeM
Track: Precision Medicine
Year: 2017
Submission: 7/27/2017
Type: automatic
Task: abstracts
MD5: a43a472c77084b9a100e5e0e7070f165
Run description: Run 3 is expanding the search to publications about more specific/general diseases. The simplified neoplasms hierarchy provided by NCI thesaurus is used (https://evs.nci.nih.gov/ftp1/NCI_Thesaurus/Neoplasm/Neoplasm_Core_Hierarchy.html). The run is then merged with run 2 through linear combination.

SIBTMlit4¶

Run ID: SIBTMlit4
Participant: BiTeM
Track: Precision Medicine
Year: 2017
Submission: 7/27/2017
Type: automatic
Task: abstracts
MD5: b9ce055116e6770deb0c4070639c3369
Run description: Run 4 is based on run 3. Results are further re-ranked based on the type of publications. Clinical trials, cohort studies and papers from ASCO and AACR are promoted.

SIBTMlit5¶

Run ID: SIBTMlit5
Participant: BiTeM
Track: Precision Medicine
Year: 2017
Submission: 7/27/2017
Type: automatic
Task: abstracts
MD5: 927b3884586b104a78803bd599b596c4
Run description: Run 5 is based on the clinical trial track. For each clinical trial returned for the clinical trial run submitted by our group, the cited MEDLINE publications are retrieved and assigned the same score than the clinical trial. Results are then merged with the run 4 through linear combination.

teckro1¶

Run ID: teckro1
Participant: teckro
Track: Precision Medicine
Year: 2017
Submission: 7/21/2017
Type: automatic
Task: trials
MD5: 7b90de111836303539dcfe04c15185c2
Run description: first run

teckro2¶

Run ID: teckro2
Participant: teckro
Track: Precision Medicine
Year: 2017
Submission: 7/21/2017
Type: automatic
Task: trials
MD5: 7ba32893e1a9ba7c6582faa302a8dd2e
Run description: second run

teckro3¶

Run ID: teckro3
Participant: teckro
Track: Precision Medicine
Year: 2017
Submission: 7/21/2017
Type: automatic
Task: trials
MD5: 9cca46928dff6c754730b14096f35e29
Run description: third run

teckro4¶

Run ID: teckro4
Participant: teckro
Track: Precision Medicine
Year: 2017
Submission: 7/21/2017
Type: automatic
Task: trials
MD5: 422a7de576c39fe196979652ee933ba1
Run description: fourth run

teckro5¶

Run ID: teckro5
Participant: teckro
Track: Precision Medicine
Year: 2017
Submission: 7/21/2017
Type: automatic
Task: trials
MD5: 56c4f21c524b540b653adb2933f2369f
Run description: fifth run

Textual¶

Run ID: Textual
Participant: NaCTeM
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 42994c7f5bf46b7227a731291a389576
Run description: This is our text-based retrieval method, including all the elements in a disjunction query: expected to have high recall, but low precision.

Textualc¶

Run ID: Textualc
Participant: NaCTeM
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: f297eb0f5d81c77581eb3ada03a21c45
Run description: This is our text-based retrieval method, including all the elements in a disjunction query.

trial1¶

Results | Participants | Input | Summary | Appendix

Run ID: trial1
Participant: FDUDMIIP
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: adb41062b2307c9e4c2dcd5bf0d059b3
Run description: baseline of trial

trial2¶

Results | Participants | Input | Summary | Appendix

Run ID: trial2
Participant: FDUDMIIP
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: cf0ba3ecb5a11e4374694641c43f4d61
Run description: trial

trial3¶

Results | Participants | Input | Summary | Appendix

Run ID: trial3
Participant: FDUDMIIP
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 16b6d3d5e09d81b3a0bd61a949220d37
Run description: trial3

trial4¶

Results | Participants | Input | Summary | Appendix

Run ID: trial4
Participant: FDUDMIIP
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 394121b75971d5ac8fa518c023a4a12c
Run description: trial4

UCASBASE¶

Run ID: UCASBASE
Participant: UCAS
Track: Precision Medicine
Year: 2017
Submission: 7/27/2017
Type: automatic
Task: trials
MD5: 079fc1f595f13ee3092acdc4a3a1179c
Run description: This is a baseline run. The UCAS team applies the DPH parameter free model with Bo1 query expansion model to produce this baseline.

UCASBASEa¶

Run ID: UCASBASEa
Participant: UCAS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: f96212f8d52f0b851e8cd6cb8ad09239
Run description: This baseline run uses DPH + Bo1 query expansion.

UCASSEM1¶

Run ID: UCASSEM1
Participant: UCAS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 1d3d746c1c6667d944172238bdb4c31f
Run description: This run uses DPH with Bo1 query expansion. Semantic relevance score based on D2D similarity using word embeddings is also applied.

UCASSEM1a¶

Run ID: UCASSEM1a
Participant: UCAS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 57eb09eb01bca51756ce26412888c75f
Run description: This run uses DPH + Bo1 query expansion. Semantic relevance score using word embeddings is also applied.

UCASSEM2¶

Run ID: UCASSEM2
Participant: UCAS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 074408235e444ce4b397de0088c9e6e4
Run description: This run uses DPH with Bo1 query expansion. Semantic relevance score based on D2D similarity using paragraph vectors is also applied.

UCASSEM2a¶

Run ID: UCASSEM2a
Participant: UCAS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: b5120490b62ab0f57a5f4ebc78f8e2df
Run description: This run uses DPH + Bo1 query expansion. Semantic relevance score using paragraph vectors is also applied.

UCASSEM3¶

Run ID: UCASSEM3
Participant: UCAS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: d2a4602747551c547ceb8a12de761276
Run description: This run uses BM25 with KL query expansion. Semantic relevance score based on D2D similarity is also applied.

UCASSEM3a¶

Run ID: UCASSEM3a
Participant: UCAS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 175f461406e3d4b981af93783a63d370
Run description: This run uses BM25 + KL query expansion. Semantic relevance score using word embeddings is also applied.

UCASSEMUMLS¶

Run ID: UCASSEMUMLS
Participant: UCAS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 5443b12652f31ec452501b04bf1816ae
Run description: This run uses DPH with Bo1 and UMLS query expansion. Semantic relevance score based on D2D similarity using word embeddings is also applied.

UCASSEMUMLSa¶

Run ID: UCASSEMUMLSa
Participant: UCAS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: eae592f750c62fee97973d15af17e4fc
Run description: This run uses DPH + Bo1 and UMLS query expansion. Semantic relevance score using paragraph vectors is also applied.

UD_GU_CT_1¶

Run ID: UD_GU_CT_1
Participant: UD_GU_BioTM
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: 03298aa1b66540162989d3f57f722e9c
Run description: This run includes high-confidence results from our system. Trials with disease/gene appearing in important sections, e.g., title or inclusion criteria, are ranked higher. The trials are mostly interventional or observational/prospective. External resources used: disease ontology and mesh terms.

UD_GU_CT_2¶

Run ID: UD_GU_CT_2
Participant: UD_GU_BioTM
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: 87e39b086441f6db8a0090d23a8a42fe
Run description: This run adds more trials to the first run UD_GU_CT_1. The added trials are mostly observational/prospective and contain interventions like genetic, behavioral or device. External resources used: disease ontology and mesh terms.

UD_GU_CT_3¶

Run ID: UD_GU_CT_3
Participant: UD_GU_BioTM
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: 5be1b15b196b902cc152894ff25292f0
Run description: This run adds more trials to the second run UD_GU_CT_2. The added trials are mainly observational/retrospective and other special cases. External resources used: disease ontology and mesh terms.

UD_GU_CT_4¶

Run ID: UD_GU_CT_4
Participant: UD_GU_BioTM
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: c24b0c3cdc395072ae7b14b018ee9e44
Run description: This run adds more trials to the first run UD_GU_CT_1, using a less restricted removal rule. External resources used: disease ontology and mesh terms.

UD_GU_CT_5¶

Run ID: UD_GU_CT_5
Participant: UD_GU_BioTM
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: trials
MD5: 53a3666169752ee971a90a4fbf4974c8
Run description: This run adds more trials to the second run UD_GU_CT_2, using a less restricted removal rule. External resources used: disease ontology and mesh terms.

UD_GU_SA_1¶

Run ID: UD_GU_SA_1
Participant: UD_GU_BioTM
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: 119379fcd2a34f62231752b85fd0f1ca
Run description: A text mining pipeline was applied to extract information specific to precision medicine. Articles where genes and diseases were more frequent in title or conclusions, received higher ranking. External resources used: DNorm, GNormPlus and MeSH thesaurus.

UD_GU_SA_2¶

Run ID: UD_GU_SA_2
Participant: UD_GU_BioTM
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: 8561291554a66d632155ac3dad087933
Run description: A text mining pipeline was applied to extract information specific to precision medicine. Articles containing semantic relationship between variant and disease received higher ranking. External resources used: DNorm, GNormPlus and MeSH thesaurus.

UD_GU_SA_3¶

Run ID: UD_GU_SA_3
Participant: UD_GU_BioTM
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: e9518156a6f27b63e4997212e7922743
Run description: A text mining pipeline was applied to extract information specific to precision medicine. Extracted information is used to rank articles. External resources used: DNorm, GNormPlus and MeSH thesaurus.

UD_GU_SA_4¶

Run ID: UD_GU_SA_4
Participant: UD_GU_BioTM
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: a45bde5d6b39100f127ba53e8315813e
Run description: A text mining pipeline was applied to extract information specific to precision medicine. Articles containing exact or more specific mentions of disease received higher ranking. External resources used: DNorm, GNormPlus and MeSH thesaurus.

UD_GU_SA_5¶

Run ID: UD_GU_SA_5
Participant: UD_GU_BioTM
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: 9dedea0b8564aac8c0c47aabdfb2db49
Run description: A text mining pipeline was applied to extract information specific to precision medicine. Features from all previous four runs are combined to rank articles. External resources used: DNorm, GNormPlus and MeSH thesaurus.

udelT1Base¶

Run ID: udelT1Base
Participant: udel
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 49366fdfde5fe0282ff7628c86e18a8f
Run description: A baseline system for the university of delaware, information retrieval lab. In this run, we apply an initial filtering by removing all the empty documents (i.e., files with no abstracts) from the collection. In addition, we rely on Terrier 4.2 IR system and divergence from randomness retrieval model to retrieve the top abstracts for each topic.

udelT1Comb¶

Run ID: udelT1Comb
Participant: udel
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 5ad5dc4f8615f6ad9c5f65b9252282de
Run description: Another run by the university of delaware, information retrieval lab. In this run, we use Inverse square rank fusion (ISR) to combine the results from the other four runs: udelT1Base, udelT1PRF, udelT1Gene, and udelT1GeMeSH. Note that the underlying retrieval system for all those runs is Terrier.

udelT1GeMeSH¶

Run ID: udelT1GeMeSH
Participant: udel
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 2c0415493820cd3078dd362c898decf7
Run description: Another run by the university of delaware, information retrieval lab. In this run, we apply an initial filtering by removing all the empty documents (i.e., files with no abstracts) from the collection. In addition, we rely on Terrier 4.2 IR system and divergence from randomness (DFR) retrieval model to retrieve the top abstracts for each topic. Moreover, we experiment with gene disambiguation as well cancer disambiguation by expanding each gene and disease with their other known names that are used in biomedical literature. We rely on Genetics Home Reference by NLM for genes whereas we use the known MeSH for the cancer diseases.

udelT1Gene¶

Run ID: udelT1Gene
Participant: udel
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 37a6393e3d5308d1141ad85e4ed24090
Run description: Another run by the university of delaware, information retrieval lab. In this run, we apply an initial filtering by removing all the empty documents (i.e., files with no abstracts) from the collection. In addition, we rely on Terrier 4.2 IR system and divergence from randomness (DFR) retrieval model to retrieve the top abstracts for each topic. Moreover, we experiment with gene disambiguation by expanding each gene with its other known names that are used in biomedical literature. We rely on Genetics Home Reference by NLM to extract those other names.

udelT1PRF¶

Run ID: udelT1PRF
Participant: udel
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: abstracts
MD5: 026e02230d47a87992a36a48933e0327
Run description: Another run by the university of delaware, information retrieval lab. In this run, we apply an initial filtering by removing all the empty documents (i.e., files with no abstracts) from the collection. In addition, we rely on Terrier 4.2 IR system and divergence from randomness (DFR) retrieval model to retrieve the top abstracts for each topic. Moreover, we experiment with pseudo relevance feedback using KL model with 60 terms from the top 5 documents

udelT2Base¶

Results | Participants | Input | Summary | Appendix

Run ID: udelT2Base
Participant: udel
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 009c40c3edb9e931cf3e6b9952f6df7a
Run description: A baseline system by the university of delaware, information retrieval lab. In this run, we filter each trail document by removing exclusion criteria and negated contexts using NegEX. In addition, we rely on Terrier 4.2 IR system and divergence from randomness retrieval model (DFR) to retrieve the top trails for each patient.

udelT2Comb¶

Results | Participants | Input | Summary | Appendix

Run ID: udelT2Comb
Participant: udel
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 658d2d388ebbf09d34488963242c3876
Run description: Another run by the university of delaware, information retrieval lab. In this run, we use Inverse square rank fusion (ISR) to combine the results from the other four runs udelT2Base, udelT2PRF, udelT2Gene, and udelT2GeMeSH. Note that the underlying retrieval system for all of these runs is Terrier.

udelT2GeMeSH¶

Results | Participants | Input | Summary | Appendix

Run ID: udelT2GeMeSH
Participant: udel
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: a76ed30acdac77e2a0e7d558b8b8b147
Run description: Another run by the university of delaware, information retrieval lab. In this run, we filter each trail document by removing exclusion criteria and negated contexts using NegEX. In addition, we rely on Terrier 4.2 IR system and divergence from randomness retrieval model (DFR) to retrieve the top trails for each patient. Moreover, we experiment with gene disambiguation as well cancer disambiguation by expanding each gene and disease with their other known names that are used in biomedical literature. We rely on Genetics Home Reference by NLM for gene expansion whereas we use the known MeSH for the expansion of cancer diseases.

udelT2Gene¶

Results | Participants | Input | Summary | Appendix

Run ID: udelT2Gene
Participant: udel
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: 4cd72f263e06fdadda0770693054dc20
Run description: Another run by the university of delaware, information retrieval lab. In this run, we filter each trail document by removing exclusion criteria and negated contexts using NegEX. In addition, we rely on Terrier 4.2 IR system and divergence from randomness retrieval model (DFR) to retrieve the top trails for each patient. Moreover, we experiment with gene disambiguation by expanding each gene with its other known names that are used in biomedical literature. We rely on Genetics Home Reference by NLM to extract those names.

udelT2PRF¶

Results | Participants | Input | Summary | Appendix

Run ID: udelT2PRF
Participant: udel
Track: Precision Medicine
Year: 2017
Submission: 7/31/2017
Type: automatic
Task: trials
MD5: e7d0313ca20caa794fc708b39f29e864
Run description: Another run by the university of delaware, information retrieval lab. In this run, we filter each trail document by removing exclusion criteria and negated contexts using NegEX. In addition, we rely on Terrier 4.2 IR system and divergence from randomness retrieval model (DFR) to retrieve the top trails for each patient. Furthermore, we experiment with pseudo relevance feedback using KL model with 40 terms from the top 2 documents.

UDInfoPMCT10¶

Run ID: UDInfoPMCT10
Participant: udel_fang
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 11f2ea93d8e4e1179323f03bbd8cafbb
Run description: Combine the results from both term based representation and concept based representation

UDInfoPMCT3¶

Run ID: UDInfoPMCT3
Participant: udel_fang
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 4d6784ddda300299b3c21ca5d1e0c8ec
Run description: Search the disease, gene, and other information in detailed_description, inclusion, keyword, condition_browse, intervention_browse. Filter the results based on gender and age information

UDInfoPMCT4¶

Run ID: UDInfoPMCT4
Participant: udel_fang
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: f02ff2632dc87e6ce5630870b59d2ec7
Run description: Search the disease, gene, and other information in detailed_description, inclusion, keyword, condition_browse, intervention_browse with query expansion

UDInfoPMCT6¶

Run ID: UDInfoPMCT6
Participant: udel_fang
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: b679975caf7cf3e4f41bc540d3e4d332
Run description: Concept based. Search the disease, gene, and other information in detailed_description, inclusion, keyword, condition_browse, intervention_browse.

UDInfoPMCT8¶

Run ID: UDInfoPMCT8
Participant: udel_fang
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 9f8ea3b36622604f15effbcd7cd47872
Run description: Search the disease, gene, and other information in concept based representation. and filter the results using demographic data in term based representation.

UDInfoPMSA2¶

Run ID: UDInfoPMSA2
Participant: udel_fang
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: a24194e62086edea429ae0668fedd743
Run description: Search using disease, gene, and other separately. The weight on disease is 0.5, gene is 0.3, and other is 0.2

UDInfoPMSA3¶

Run ID: UDInfoPMSA3
Participant: udel_fang
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 48d2e2bcd97be4caeffeda643923c8b4
Run description: Search the disease, gene, and other information. Expand the query using the alias and top 10 terms from summary extracted from Gene Card. The weight on original terms is 0.8, and expansion term is 0.2

UDInfoPMSA5¶

Run ID: UDInfoPMSA5
Participant: udel_fang
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 8fa5ddfba3fb4db70ead85ff31078a7a
Run description: Concept based. Search using disease, gene, and other separately. The weight on disease is 0.5, gene is 0.3, and other is 0.2

UDInfoPMSA6¶

Run ID: UDInfoPMSA6
Participant: udel_fang
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 88797af5b366e970b8a16c94f20d628b
Run description: Concept based. Search using disease, and gene. Applied Unified and Balance method on the identified concepts

UDInfoPMSA7¶

Run ID: UDInfoPMSA7
Participant: udel_fang
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 1aae15cd83699d70300c11b9b13af188
Run description: Combine the results from both term based representation and concept based representation.

UKY_AGG¶

Run ID: UKY_AGG
Participant: UKNLP
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: abstracts
MD5: c21bc366116ec1747968e2128a6e8038
Run description: solr/lucene, ulms + mesh on demand, aggregated results of previous runs
Code: https://github.com/romanegloo/trec2017pm

UKY_BASE¶

Run ID: UKY_BASE
Participant: UKNLP
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: abstracts
MD5: 658d52b88f171e9f2c50b57ff314a6c0
Run description: solr/lucene bm25, basic umls query expansion
Code: https://github.com/romanegloo/trec2017pm

UKY_CJT¶

Run ID: UKY_CJT
Participant: UKNLP
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: abstracts
MD5: 48a0866cb5ab9673c78ed8a94399a1f7
Run description: solr/lucene bm25, re-rank with the conjunctive results on top
Code: https://github.com/romanegloo/trec2017pm

UKY_COM¶

Run ID: UKY_COM
Participant: UKNLP
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: abstracts
MD5: ebb6e0ce4c7807069db23f18c0a80e8a
Run description: query expansion with mesh on demand results, re-ranked with conjunctive matches
Code: https://github.com/romanegloo/trec2017pm

UKY_MAN¶

Run ID: UKY_MAN
Participant: UKNLP
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: manual
Task: abstracts
MD5: 25dea569db35808769f2944b9e9fdc0d
Run description: selectively chosen queries by topics from the previous experiments
Code: https://github.com/romanegloo/trec2017pm

UKY_T2¶

Run ID: UKY_T2
Participant: UKNLP
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: trials
MD5: 69c2ad8064e901855456e06cdb7e4e25
Run description: same as RUN2 on articles
Code: https://github.com/romanegloo/trec2017pm

UKY_T3¶

Run ID: UKY_T3
Participant: UKNLP
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: trials
MD5: ac9422b8ad60e04f3c2fa149b942a316
Run description: same as RUN3 on articles
Code: https://github.com/romanegloo/trec2017pm

UKY_T4¶

Run ID: UKY_T4
Participant: UKNLP
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: trials
MD5: f7e180a4433c0178e3518002d81342f6
Run description: same as RUN4 on articles
Code: https://github.com/romanegloo/trec2017pm

UKY_TRL¶

Run ID: UKY_TRL
Participant: UKNLP
Track: Precision Medicine
Year: 2017
Submission: 7/30/2017
Type: automatic
Task: trials
MD5: 562e48e94f8958d94c263ba76fb42dd5
Run description: basic umls atoms based query expansion on trials
Code: https://github.com/romanegloo/trec2017pm

UNTIIACTDW¶

Run ID: UNTIIACTDW
Participant: UNTIIA
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 266d12fed2ddfd9fc5a960373d7657cd
Run description: Used Terrier In_expC2 weighting model with pseudo relevance feedback of top 3 documents and 50 terms within each document; Queries were constructed with all terms in the topics but terms in "Disease" tag were weighted higher (3.0).

UNTIIACTGA¶

Run ID: UNTIIACTGA
Participant: UNTIIA
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: a81d4f92e2fa2fd37ab38e55b677f66e
Run description: Used Terrier In_expC2 weighting model with pseudo relevance feedback of top 3 documents and 50 terms within each document; Queries were constructed with all terms in the topics but terms in "Disease" tag were weighted higher (3.0), added UMLS terms for the disease name and genome data, and generic treatment, ALSO, specific treatment terms were added to each query with respect to their disease name which were extracted from the cancer.org portal. ALSO, gene alias terms obtained from PubMed portal were added to the query terms.

UNTIIACTIS¶

Run ID: UNTIIACTIS
Participant: UNTIIA
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 536146532a2f1ede0bbab5325a4cd610
Run description: Used Terrier In_expC2 weighting model with pseudo relevance feedback of top 3 documents and 50 terms within each document; Queries were constructed with all terms in the topics but terms in "Disease" tag were weighted higher (3.0), added UMLS terms for the disease name and genome data, and generic treatment, ALSO, specific treatment terms were added to each query with respect to their disease name which were extracted from the cancer.org portal.

UNTIIACTLQ¶

Run ID: UNTIIACTLQ
Participant: UNTIIA
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 6b4611b1e0aa8de469233b6879f65aa3
Run description: Used Terrier In_expC2 weighting model with pseudo relevance feedback of top 3 documents and 50 terms within each document; Queries were constructed with all terms in the topics but terms in "Disease" tag were weighted higher (3.0). Used logical queries. Synonyms of disease names and expansions of Genome data were added to the query terms.

UNTIIACTSY¶

Run ID: UNTIIACTSY
Participant: UNTIIA
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 8d59c8e5c6b3dc20f30171f2360418f4
Run description: Used Terrier In_expC2 weighting model with pseudo relevance feedback of top 3 documents and 50 terms within each document; Queries were constructed with all terms in the topics but terms in "Disease" tag were weighted higher (3.0). ALSO, UMLS terms for the disease name and genome data, and generic treatment were added to the query.

UNTIIADW¶

Run ID: UNTIIADW
Participant: UNTIIA
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 4884c075ccd4221498d9fd1234749b82
Run description: Used Terrier In_expC2 weighting model with pseudo relevance feedback of top 3 documents and 50 terms within each document, queries were constructed with all terms in the topics but terms in "Disease" tag were weighted higher (3.0). Reranked the retrieved documents based on the availability of disease terms, genome data and treatment related terms.

UNTIIAGA¶

Run ID: UNTIIAGA
Participant: UNTIIA
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: de205a71fabe3cbbf125e0d8a8ac4d8e
Run description: Used Terrier In_expC2 weighting model with pseudo relevance feedback of top 3 documents and 50 terms within each document; Queries were constructed with all terms in the topics but terms in "Disease" tag were weighted higher (3.0), added UMLS terms for the disease name and genome data, and generic treatment, ALSO, specific treatment terms were added to each query with respect to their disease name which were extracted from the cancer.org portal. ALSO, gene alias terms obtained from PubMed portal were added to the query terms; Reranked the retrieved documents based on the availability of disease terms, genome data and treatment related terms.

UNTIIAIS¶

Run ID: UNTIIAIS
Participant: UNTIIA
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: f5876b4cb4936b5ef76cad527088652a
Run description: Used Terrier In_expC2 weighting model with pseudo relevance feedback of top 3 documents and 50 terms within each document; Queries were constructed with all terms in the topics but terms in "Disease" tag were weighted higher (3.0), added UMLS terms for the disease name and genome data, and generic treatment, ALSO, specific treatment terms are added to each query with respect to their disease name which are extracted from the cancer.org portal; Reranked the retrieved documents based on the availability of disease terms, genome data and treatment related terms.

UNTIIALQ¶

Run ID: UNTIIALQ
Participant: UNTIIA
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 3067ad6f6a9cb958bc50db8dbe66616c
Run description: Used Terrier In_expC2 weighting model with pseudo relevance feedback of top 3 documents and 50 terms within each document; Queries were constructed with all terms in the topics but terms in "Disease" tag were weighted higher (3.0). Used logical queries. Synonyms of disease names and expansions of Genome data were added to the query terms; Reranked the retrieved documents based on the availability of disease terms, genome data and treatment related terms.

UNTIIASY¶

Run ID: UNTIIASY
Participant: UNTIIA
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 95eced4ee107321a12a04886644e1568
Run description: Used Terrier In_expC2 weighting model with pseudo relevance feedback of top 3 documents and 50 terms within each document, queries were constructed with all terms in the topics but terms in "Disease" tag were weighted higher (3.0), added UMLS terms for the disease name and genome data, and generic treatment . Reranked the retrieved documents based on the availability of disease terms, genome data and treatment related terms.

UTDHLTAF¶

Run ID: UTDHLTAF
Participant: UTDHLTRI
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 90a4740a7f271d9e5e72a6d3c17f7bb5
Run description: Hybrid Question Answering and Information Retrieval approach representing each topic as a set of aspects represented by different queries. Relies on Lucene and incorporates drug/gene information from COSMIC, DGIdb, and FdaLabels.

UTDHLTAFT¶

Run ID: UTDHLTAFT
Participant: UTDHLTRI
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: d2eba0a6cc16add47d426613fcc3d5fe
Run description: Hybrid Question Answering and Information Retrieval approach representing each topic as a set of aspects represented by different queries. Relies on Lucene and incorporates drug/gene information from COSMIC, DGIdb, and FdaLabels.

UTDHLTFF¶

Run ID: UTDHLTFF
Participant: UTDHLTRI
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 24917edc3d49d07cf4d6a863a8c36d25
Run description: Rank fusion approach representing each topic representing each topic as a set of aspects represented by different queries and relying on different similarities. Relies on Lucene and incorporates drug/gene information from COSMIC, DGIdb, and FdaLabels.

UTDHLTFFT¶

Run ID: UTDHLTFFT
Participant: UTDHLTRI
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: b8d316da4a60a40e06749a92b52595b6
Run description: Rank fusion approach representing each topic representing each topic as a set of aspects represented by different queries and relying on different similarities. Relies on Lucene and incorporates drug/gene information from COSMIC, DGIdb, and FdaLabels.

UTDHLTJQ¶

Run ID: UTDHLTJQ
Participant: UTDHLTRI
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 62475ebb5c6ecefa9551f2730878a4ef
Run description: Hybrid Question Answering and Information Retrieval approach representing each topic with a single large query. Relies on Lucene and incorporates drug/gene information from COSMIC, DGIdb, and FdaLabels.

UTDHLTJQT¶

Run ID: UTDHLTJQT
Participant: UTDHLTRI
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: f319f6897da151662b5b4dae858ac261
Run description: Hybrid Question Answering and Information Retrieval approach representing each topic with a single large query. Relies on Lucene and incorporates drug/gene information from COSMIC, DGIdb, and FdaLabels.

UTDHLTSF¶

Run ID: UTDHLTSF
Participant: UTDHLTRI
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 60e5c07850a002894766a19a3e275f10
Run description: Rank fusion approach representing each topic with a single large query and considering multiple relevance functions. Relies on Lucene and incorporates drug/gene information from COSMIC, DGIdb, and FdaLabels.

UTDHLTSFT¶

Run ID: UTDHLTSFT
Participant: UTDHLTRI
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 65502ff3eb62a777d9c1d5f8a87edc07
Run description: Rank fusion approach representing each topic with a single large query and considering multiple relevance functions. Relies on Lucene and incorporates drug/gene information from COSMIC, DGIdb, and FdaLabels.

UTDHLTSQ¶

Run ID: UTDHLTSQ
Participant: UTDHLTRI
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: abstracts
MD5: 7c051c6715c733ba43764112972d09a0
Run description: Simple Information Retrieval approach without considering any external resources. Relies on Lucene.

UTDHLTSQT¶

Run ID: UTDHLTSQT
Participant: UTDHLTRI
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 26af0af6fc0fbf4968f25033a0c3c232
Run description: Simple Information Retrieval approach without considering any external resources. Relies on Lucene.

UWMSOIS0¶

Results | Participants | Input | Summary | Appendix

Run ID: UWMSOIS0
Participant: UWMSOIS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 5f4f14db0c1024afc90bdf51d49716d4
Run description: Base run for the Clinical Trials. All field text from original topics (topics2017.xml) have been used as queries. Bayesian smoothing with Dirichlet Prior and Porter stemmer had been set up as default for the retrieval in Terrier: e.g. query 1: "Liposarcoma CDK4 Amplification 38-year-old male GERD"

UWMSOIS1¶

Results | Participants | Input | Summary | Appendix

Run ID: UWMSOIS1
Participant: UWMSOIS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 2947fa5f03cad63d3cd47b4c7bd5cb63
Run description: Query Expansion (QE) using LDA: the first LDA topic word representing the first (top 1) document that was retrieved on the 2016 CDS Track dataset (2016 PMC snapshot: full-text documents) by each 2017 PM Track query. The expanded query was used as a new query for the Clinical Trials dataset. 1. A LDA model with 800 topics was trained based on the 2016 PMC snapshot (Dec. 4) - ftp://ftp.ncbi.nlm.nih.gov/pub/pmc. Before training, all full-text documents in the 2016 PMC snapshot (Dec. 4) were represented by only MeSH words that are included in the "MH" field (MESH) in the 2017 Mesh descriptor file (https://www.nlm.nih.gov/mesh/download_mesh.html). We considered only MeSH terms, which might be Unigram or Multi-gram words. If a MeSH includes more than one word, each word included in the MeSH term was not considered. Only a complete MeSH term described in the MeSH was considered as a term. Meanwhile, if a MeSH is made of a word, the word was regarded as a MeSH term. The Gensim module (Python-based, https://radimrehurek.com/gensim/) was used to generate a LDA model. 2. The 2016 CDS track dataset (http://www.trec-cds.org/2016.html) was indexed by Terrier in order to get a top 1 retrieved document for 2017 PM track query. 30 top 1 full-text documents for 30 queries of the PM Track were used to predict the first LDA topic for each top 1 document, which has the largest topic probability representing the document. The first topic word for the predicted topic was added to the original query. E.g. QE 1: "Liposarcoma CDK4 Amplification 38-year-old male GERD recurrence"(recurrence was added).

UWMSOIS2¶

Results | Participants | Input | Summary | Appendix

Run ID: UWMSOIS2
Participant: UWMSOIS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 25963abde8336562bdb688ad30ce08fe
Run description: TP * norm (ITP): Query Expansion (QE) using the normalization of inverted LDA topic probability using maximum and minimum values of the sums of each topic probability. The LDA model and 30 top 1 documents were used in the same way of the UWMSOIS1. The predicted topic probability for each documents was weighted by the normalized form of inverse topic probability: tp (Topic Probabilty for the top 1 document) * Norm (Inverted Topic Probability), where Norm (Inverted Topic Probability) = (ITP(i) of the document -min(ITPs)) / (max(ITPs)-min(ITPs)), ITP(i) of the document = inverse value of the sum of the topic i probability values for all documents in the collection, when the topic i has the biggest probability value in representing the document. Max(ITPs) means the maximum value of all ITPs for all topics. This is a similar concept of TF* normalization (IDF).

UWMSOIS3¶

Results | Participants | Input | Summary | Appendix

Run ID: UWMSOIS3
Participant: UWMSOIS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 0ab74937fdc52efc879dd91a1e4ea364
Run description: TP * I(C)TP : Query Expansion (QE) using the inverted LDA collection-level topic probability. The LDA model and 30 top 1 documents were used in the same way of the UWMSOIS1. The predicted topic probability for each documents was weighted by the Inverse Collection Topic Probability: tp (Topic Probabilty for the document) * Inverted Collection Topic Probability (log (sum of probability values for all topics/sum of probability for the first document topic)). This is a similar concept of TF* (ICF, CF: the word frequency in the collection). The log base is 10. One difference between TP and CF might be that the maximum value of TP cannot exceed 1, so the impact of relatively big TP values in a few documents might not be that critical(that affect the bias by a few documents) as much as the large number of term frequency in a few documents can does.

UWMSOIS4¶

Results | Participants | Input | Summary | Appendix

Run ID: UWMSOIS4
Participant: UWMSOIS
Track: Precision Medicine
Year: 2017
Submission: 8/1/2017
Type: automatic
Task: trials
MD5: 310a9de19e7febeb94b6039438306ff0
Run description: TP * ITF: Query Expansion (QE) using the inverted LDA topic frequency. The LDA model and 30 top 1 documents were used in the same way of the UWMSOIS1. The predicted topic probability for each documents was weighted by the Inverse Topic Frequency: tp (Topic Probability for the document) * Inverted Topic Frequency (log (sum of all frequency values for all topics/sum of the frequency for the document topic)). This is a similar concept of TF* IDF.

UWMSOIS5¶

Run ID: UWMSOIS5
Participant: UWMSOIS
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: b3ef5a9a25cb3ebb07af4ec604ef5678
Run description: Base run for the PubMed documents set. All field text from original topics (topics2017.xml) have been used as queries. Bayesian smoothing with Dirichlet Prior and Porter stemmer had been set up as default for the retrieval in Terrier: e.g. query 1: "Liposarcoma CDK4 Amplification 38-year-old male GERD"

UWMSOIS6¶

Run ID: UWMSOIS6
Participant: UWMSOIS
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: 11d639844dbad033554529aed1499eea
Run description: Query Expansion (QE) using LDA: the first LDA topic word representing the first (top 1) document that was retrieved on the 2016 CDS Track dataset (2016 PMC snapshot: full-text documents) by each 2017 PM Track query. The expanded query was used as a new query(UWMSOIS6) for the MEDLINE data set. 1. A LDA model with 800 topics was trained based on the 2016 PMC snapshot (Dec. 4) - ftp://ftp.ncbi.nlm.nih.gov/pub/pmc. Before training, all full-text documents in the 2016 PMC snapshot (Dec. 4) were represented by only MeSH words that are included in the "MH" field (MESH) in the 2017 Mesh descriptor file (https://www.nlm.nih.gov/mesh/download_mesh.html). We considered only MeSH terms, which might be Unigram or Multi-gram words. If a MeSH includes more than one word, each word included in the MeSH term was not considered. Only a complete MeSH term described in the MeSH was considered as a term. Meanwhile, if a MeSH is made of a word, the word was regarded as a MeSH term. The Gensim module (Python-based, https://radimrehurek.com/gensim/) was used to generate a LDA model. 2. The 2016 CDS Track dataset (http://www.trec-cds.org/2016.html) was indexed by Terrier in order to get a top 1 retrieved document for 2017 PM Track query. 30 top 1 full-text documents for 30 queries of the PM Track were used to predict the first LDA topic for each top 1 document, which has the largest topic probability representing the document. The first topic word for the predicted topic was added to the original query. E.g. QE 1: "Liposarcoma CDK4 Amplification 38-year-old male GERD recurrence"(recurrence was added).

UWMSOIS7¶

Run ID: UWMSOIS7
Participant: UWMSOIS
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: fa291a85f2a1f79bb214864bfbcf0da2
Run description: TP * norm (ITP): Query Expansion (QE) using the normalization of inverted LDA topic probability using maximum and minimum values of the sums of each topic probability. The LDA model and 30 top 1 documents were used in the same way of the UWMSOIS1 & UWMSOIS6. The predicted topic probability for each documents was weighted by the normalized form of inverse topic probability: tp (Topic Probability for the top 1 document) * Norm (Inverted Topic Probability), where Norm (Inverted Topic Probability) = (ITP(i) of the document -min(ITPs)) / (max(ITPs)-min(ITPs)), ITP(i) of the document = inverse value of the sum of the topic i probability values for all documents in the collection, when the topic i has the biggest probability value in representing the document. Max(ITPs) means the maximum value of all ITPs for all topics. This is a similar concept of TF* normalization (IDF).

UWMSOIS8¶

Run ID: UWMSOIS8
Participant: UWMSOIS
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: 6bccbea1120a5f04b07c957c23a23388
Run description: TP * I(C)TP : Query Expansion (QE) using the inverted LDA collection-level topic probability. The LDA model and 30 top 1 documents were used in the same way of the UWMSOIS1 & UWMSOIS6. The predicted topic probability for each documents was weighted by the Inverse Collection Topic Probability: tp (Topic Probability for the document) * Inverted Collection Topic Probability (log (sum of probability values for all topics/sum of probability for the first document topic)). This is a similar concept of TF* (ICF, CF: the word frequency in the collection). The log base is 10. One difference between TP and CF might be that the maximum value of TP cannot exceed 1, so the impact of relatively big TP values in a few documents might not be that critical(that affect the bias by a few documents) as much as the large number of term frequency in a few documents can does.

UWMSOIS9¶

Run ID: UWMSOIS9
Participant: UWMSOIS
Track: Precision Medicine
Year: 2017
Submission: 8/2/2017
Type: automatic
Task: abstracts
MD5: 362d096aff6293af29a81b2c7457c1e8
Run description: TP * ITF: Query Expansion (QE) using the inverted LDA topic frequency. The LDA model and 30 top 1 documents were used in the same way of the UWMSOIS1 & UWMSOIS6. The predicted topic probability for each documents was weighted by the Inverse Topic Frequency: tp (Topic Probability for the document) * Inverted Topic Frequency (log (sum of all frequency values for all topics/sum of the frequency for the document topic)). This is a similar concept of TF* IDF.