Runs - Clinical Decision Support 2016¶

AutoDes¶

Run ID: AutoDes
Participant: FDUDMIIP
Track: Clinical Decision Support
Year: 2016
Submission: 7/26/2016
Type: automatic
MD5: abd0de9ffd4080670096a4058e32b44f
Run description: We used language model based on indri

AutoNote¶

Run ID: AutoNote
Participant: FDUDMIIP
Track: Clinical Decision Support
Year: 2016
Submission: 7/26/2016
Type: automatic
MD5: 3159817cb469bbf73bad82aa2b8bb4c4
Run description: We used language mode based on Indri.

AutoSummary¶

Run ID: AutoSummary
Participant: FDUDMIIP
Track: Clinical Decision Support
Year: 2016
Submission: 7/26/2016
Type: automatic
MD5: 7f42f9ff5f7e370984196265e89402e9
Run description: We used language mode based on Indri.

AutoSummary1¶

Run ID: AutoSummary1
Participant: FDUDMIIP
Track: Clinical Decision Support
Year: 2016
Submission: 7/26/2016
Type: automatic
MD5: 376e1e164bb1a7157e1d384ecaef1639
Run description: We used language mode based on Indri.

cbnun1¶

Run ID: cbnun1
Participant: cbnu
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 1e268c7e50d6f250dfc67619e104e704
Run description: pseudo relevance feedback based on word embeddings and disease-centered clusters (topic type: notes)

cbnus1¶

Run ID: cbnus1
Participant: cbnu
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 92403e672b069f43ff561757310ad7c6
Run description: pseudo relevance feedback using word embeddings and disease-centered clusters (topic summary type)

cbnus2¶

Run ID: cbnus2
Participant: cbnu
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: a100ad484f4bc7a6057301903818dbdf
Run description: re-ranking based on word embeddings and disease-centered clusters (topic summary type)

CCNUDESR2¶

Run ID: CCNUDESR2
Participant: CCNU2016TREC
Track: Clinical Decision Support
Year: 2016
Submission: 7/25/2016
Type: automatic
MD5: 7ae671ea59f5c1ab2b1dc915c7b2585e
Run description: Modify the IDF of BM25 in the first step, and then use pseudo relevance feedback.

CCNUNOTER1¶

Run ID: CCNUNOTER1
Participant: CCNU2016TREC
Track: Clinical Decision Support
Year: 2016
Submission: 7/25/2016
Type: automatic
MD5: a94677907f303463696457795fd0804d
Run description: Modify the IDF of BM25 in the first step, and then use pseudo relevance feedback.

CCNUNOTER2¶

Run ID: CCNUNOTER2
Participant: CCNU2016TREC
Track: Clinical Decision Support
Year: 2016
Submission: 7/25/2016
Type: automatic
MD5: 95dddd38f12851dd88cd9a61bb1cca99
Run description: Modify the TF of MATF in the first step, and then use pseudo relevance feedback.

CCNUNOTER3¶

Run ID: CCNUNOTER3
Participant: CCNU2016TREC
Track: Clinical Decision Support
Year: 2016
Submission: 7/25/2016
Type: automatic
MD5: 620ac76cb87f822925185bc272c88529
Run description: Modify the TF of MATF and IDF of BM25 in the first step, and then combine them,in addition to, use pseudo relevance feedback.

CCNUSUMR1¶

Run ID: CCNUSUMR1
Participant: CCNU2016TREC
Track: Clinical Decision Support
Year: 2016
Submission: 7/25/2016
Type: automatic
MD5: 959052d0512b233b6cee218928cd5128
Run description: Firstly, modify BM25 and MATF respectively. Then, linearly combine them.

CSIROdSum¶

Run ID: CSIROdSum
Participant: CSIROmed
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: ad7584872955a583394bb5369f98ee1b
Run description: Description and summary fields were fed into Solr. Abstracts and titles were given boost over bodies of the paper. Matching phrases were boosted in ranking. Title and abstract metamap identified concepts were also indexed and searched.

CSIROmeta¶

Run ID: CSIROmeta
Participant: CSIROmed
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: decaf51db60c9db5457d28e2dc54fc28
Run description: Metamap was run over all three fields, and keywords were searched for in Solr. Abstracts and titles were given boost over bodies of the paper. Matching phrases were boosted in ranking. Title and abstract metamap identified concepts were also indexed and searched.

CSIROmnul¶

Run ID: CSIROmnul
Participant: CSIROmed
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: manual
MD5: 8dbca157b6ffcdec6e06840a3ea982e1
Run description: Used a Noun Chunks Extractor to extract keyphrases in each topic, the extracted key-phrases were checked against a dictionary of medical abbreviations and expanded if found. Synonyms of disease names were added based on either Mayoclinic or wikipedia entries of that disease (if applicable).

CSIROnote¶

Run ID: CSIROnote
Participant: CSIROmed
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 7acfaacdc26db48cbaf1a30f28d8f433
Run description: Supplied notes were santised and fed directly into the Solr search engine as queries. Abstracts and titles were given boost over bodies of the paper. Matching phrases were boosted in ranking. Title and abstract metamap identified concepts were also indexed and searched.

CSIROsumm¶

Run ID: CSIROsumm
Participant: CSIROmed
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: feded3232bb507aee2ca32bae2535820
Run description: Metamap was run over summary fields. Original summary and keywords from the metamap output were fed into Solr. Abstracts and titles were given boost over bodies of the paper. Matching phrases were boosted in ranking. Title and abstract metamap identified concepts were also indexed and searched.

d2vCombIrit¶

Run ID: d2vCombIrit
Participant: IRIT
Track: Clinical Decision Support
Year: 2016
Submission: 7/26/2016
Type: automatic
MD5: cccfe8b21b86508cd66f2dc3d96ca602
Run description: Ranking with a neural network with takes in account the text, the description of concepts extracted from text (using MeSH). Scoring is combined with BM25 ranking.

d2vDescIrit¶

Run ID: d2vDescIrit
Participant: IRIT
Track: Clinical Decision Support
Year: 2016
Submission: 7/26/2016
Type: automatic
MD5: c87571711d127db5439cc374c94d7a6c
Run description: Ranking with a neural network with takes in account the text, the description of concepts extracted from text (using MeSH)

dacmmf¶

Run ID: dacmmf
Participant: HAUT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 61bb0b50814c56eefddfcc154a377204
Run description: Abstract-based. Medical Text Indexer for query expansion; Article classification; Multi-model fusion using description.

DAdescTM¶

Run ID: DAdescTM
Participant: DA_IICT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: a23c3a60425ac3d8863a9c7459f8b39b
Run description: Run using description as query and topics from topic modeling for pseudo relevance feedback.

DAnote¶

Run ID: DAnote
Participant: DA_IICT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 79fe9bbe978161d11f9d06343a2a1044
Run description: Run using note as query and In_expC2 retrieval model.

DAnoteRoc¶

Run ID: DAnoteRoc
Participant: DA_IICT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 214757e25830adbd11e4464e77a73540
Run description: Run using note as query and rocchio for pseudo relevance feedback.

DAnoteTM¶

Run ID: DAnoteTM
Participant: DA_IICT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 522ccada91472a638ca127ca9765f2ab
Run description: Run using note as query and topics from topic modeling for pseudo relevance feedback.

DAsummTM¶

Run ID: DAsummTM
Participant: DA_IICT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 61815e1d9f7e3bdbda4cae1e504c3381
Run description: Run using summary as query and topics from topic modeling for pseudo relevance feedback.

DDPHBo1CM¶

Run ID: DDPHBo1CM
Participant: IAII_PUT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 0dabe1822699a56338f219357ec3307a
Run description: Document retieval using Terrier (with DPH Bo1) and query expansion - keywords for topic type and MESH for medical terms in description.

DDPHBo1MWRe¶

Run ID: DDPHBo1MWRe
Participant: IAII_PUT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: b43950133e804301103c3b74ca893735
Run description: Hybrid mode - depending on score, different methods used; for most relevant - DPH Bo1 using description, then gradually added query expansions: using MESH and topic type keywords, then word2vec query expansion and finally removed description text leaving only query expansion.

descUIOWAS2¶

Run ID: descUIOWAS2
Participant: UIowaS
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 121dd1bb5ed885c2f242983beb553191
Run description: We start by processing topics2016.xml and extract the 30 topics. Given the type (note, summary, or description) of topic, we use only that text category as input in a particular run. We use Metamap to extract all the UMLS concepts in the text. We use a subset of these extracted concepts in creating the query per topic. We also extract the age and gender of the patient from the text and use it in the query. Lastly we use indri to run the query on pre-indexed pmc dataset to get the final ranked list.

DUTHaaRPF¶

Run ID: DUTHaaRPF
Participant: DUTH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: ba545db62ae96326413042cd3edda289
Run description: Using UMLS METAMAP mappings, the complete set of atoms for CUIs of interest were extracted and added to the SUMMARY in order to create a query. RPF with 3 Docs/10Terms was used (INDRI 5.8)

DUTHmaRPF¶

Run ID: DUTHmaRPF
Participant: DUTH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 9422e6b4db6d64b2dc53e05a8b34d000
Run description: Using UMLS METAMAP mappings, multiple atoms for CUIs of interest were extracted and added to the SUMMARY in order to create a query. RPF with 3 Docs/10Terms was used (INDRI 5.8)

DUTHsaRPF¶

Run ID: DUTHsaRPF
Participant: DUTH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: c0c3d731a11746b918e35799b0296b4c
Run description: Using UMLS METAMAP mappings, the first atom for CUIs of interest was extracted and added to the SUMMARY in order to create a query. RPF with 3 Docs/10Terms was used (INDRI 5.8)

ECNUmanual¶

Run ID: ECNUmanual
Participant: ECNU
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: manual
MD5: bbdc02c6a1a7a8fbc89a063dcdad033b
Run description: We invite an medical PhD student to give us the diagnosis of each topic by only using content in the note. We retrieve by utilizing Terrier with PL2 model.
Code: https://github.com/heyunh2015/cds2016.git

ECNUrun1¶

Run ID: ECNUrun1
Participant: ECNU
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: f226b53302c3edee3b313d47f1316b73
Run description: Put the original summay content into the Google engine, use the words which not only appear in the top 10 retrieval results but also appear in the MeSH dictionary to do the query expansion.
Code: https://github.com/heyunh2015/cds2016.git

ECNUrun3¶

Run ID: ECNUrun3
Participant: ECNU
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: feb7541c65d4b041caad9440fdbfa665
Run description: Query expansion with pesudo relevance feedback and terms extracted from Google engine Combine the results from Language model, BM25 model and log-logistic DFR model
Code: https://github.com/heyunh2015/cds2016.git

ECNUrun4¶

Run ID: ECNUrun4
Participant: ECNU
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: f6aa61738c1640d78f850fabcaf606eb
Run description: Utilizing the KDBA system to find out the query terms which appear in DBPedia. Then we apply the MeSH to search those words, and the results are processed and added to query as the expansion terms. Finally, we retrieve with terrier.
Code: https://github.com/heyunh2015/cds2016.git

ECNUrun5¶

Run ID: ECNUrun5
Participant: ECNU
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 97e9ed406dca089d7e2f9307ad5492c3
Run description: Put the original summay content into the Google engine, use the words which not only appear in the top 10 retrieval results but also appear in the MeSH dictionary to do the query expansion.
Code: https://github.com/heyunh2015/cds2016.git

ETHDescRR¶

Run ID: ETHDescRR
Participant: ETH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 769ebe24c6a8d9f471e3d34cb440ce1f
Run description: We use a range of query expansion and re-ranking features based on MeSH.

ETHNote¶

Run ID: ETHNote
Participant: ETH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 1179e30af96b9ddaf1e110921b80177c
Run description: We use a range of query expansion and re-ranking features based on MeSH.

ETHNoteRR¶

Run ID: ETHNoteRR
Participant: ETH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: c5377e6711975f56074353c58b427e4b
Run description: We use a range of query expansion and re-ranking features based on MeSH.

ETHSumm¶

Run ID: ETHSumm
Participant: ETH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 5f7797be61fb4826df9528696642b77f
Run description: We use a range of query expansion and re-ranking features based on MeSH.

ETHSummRR¶

Run ID: ETHSummRR
Participant: ETH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 6f0364939a6a415b5a5988207777dc82
Run description: We use a range of query expansion and re-ranking features based on MeSH.

lsbn¶

Run ID: lsbn
Participant: hany-miner
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: e6b047cfebd8e528e4d0a06ed36c9411
Run description: Key features of this run were BiGram,BM25,Child vs Adult Filter. No SVM feature is added in this run.

lssbd¶

Run ID: lssbd
Participant: hany-miner
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 78fb4b63f35f53b0a93a1cc3d8bf98f7
Run description: Key features of this run were BiGram,BM25,SVM,Child vs Adult Filter. First, SVM (LIBSVM)is used to classify the three different topics, then used Lucene to search documents. Synonyms were added and the source was from Metathesaurus. After retreiving relevant documents, AgeRnageFilter find whether topic and documents are related to child or adult and filter ones do not match.

lssbn¶

Run ID: lssbn
Participant: hany-miner
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 03f0800e33c9ea407171c43be1a7eb09
Run description: Key features of this run were BiGram,BM25,SVM,Child vs Adult Filter. First, SVM (LIBSVM)is used to classify the three different topics, then used Lucene to search documents. Synonyms were added and the source was from Metathesaurus. After retreiving relevant documents, AgeRnageFilter find whether topic and documents are related to child or adult and filter ones do not match.

lssbs¶

Run ID: lssbs
Participant: hany-miner
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 18a71748c9a35f688f6e3cc89d35bfbc
Run description: Key features of this run were BiGram,BM25,SVM,Child vs Adult Filter. First, SVM (LIBSVM)is used to classify the three different topics, then used Lucene to search documents. Synonyms were added and the source was from Metathesaurus. After retreiving relevant documents, AgeRnageFilter find whether topic and documents are related to child or adult and filter ones do not match.

LucBase¶

Run ID: LucBase
Participant: SCIAICLTeam
Track: Clinical Decision Support
Year: 2016
Submission: 7/22/2016
Type: automatic
MD5: 93643863bf3348fa94623f6b088c6a4b
Run description: This run used Lucene to index the corpus. We passed the topic summary as the query and returned the top 20 documents for each topic.

LucNote¶

Run ID: LucNote
Participant: SCIAICLTeam
Track: Clinical Decision Support
Year: 2016
Submission: 7/22/2016
Type: automatic
MD5: 2f85f67576580067b976529525d43e3a
Run description: This run used Lucene to find documents for each topic. We passed in the note field of each topic as queries and returned the top 20 documents.

LucNoteFrame¶

Run ID: LucNoteFrame
Participant: SCIAICLTeam
Track: Clinical Decision Support
Year: 2016
Submission: 7/26/2016
Type: automatic
MD5: 724fce981b491e1e1d74d01efffb75b0
Run description: First, the corpus is searched using Lucene for the top 30 documents for each topic note. These documents are then passed into our framing program, which compares elements of each document to the topics and narrows the documents into a top 10 list. The run also uses UMLS Metamap and the Stanford POS tagger to assist in document parsing.

LucWeight¶

Run ID: LucWeight
Participant: SCIAICLTeam
Track: Clinical Decision Support
Year: 2016
Submission: 7/26/2016
Type: automatic
MD5: afed83af2876331768e67a176244b993
Run description: This run combined Lucene with machine learning. The weights of symptoms found through machine learning were used to help boost certain symptoms in the Lucene queries.

LucWghtFrame¶

Run ID: LucWghtFrame
Participant: SCIAICLTeam
Track: Clinical Decision Support
Year: 2016
Submission: 7/25/2016
Type: automatic
MD5: a187d83add609187019039916cea4a67
Run description: First, the corpus is searched using a Lucene for the top 30 documents for each topic summary. These documents are then passed into our framing program, which compares elements of each document to the topics and narrows the documents into a top 10 list. The run also uses UMLS Metamap and the Stanford POS tagger to assist in document parsing.

ManualRun¶

Run ID: ManualRun
Participant: nch_risi
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: manual
MD5: 45edaaa42c16535bf0e5d6afcfbe495d
Run description: A group of human experts worked together to come up with keywords and use google to search. We have discrepancy between our googled list of PMCs and provided PMCIDs. But we only include provided PMCIDs only. This eliminates about 50% of relevant entries.

mayoad¶

Run ID: mayoad
Participant: MayoNLPTeam
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 0584a4c1bbeb83fb58b19d73c7d5867f
Run description: Using the description field in each topic, we applied an ensemble model of (1) a Part-of-Speech query term weighting model; (2) a Markov Random Field Model of extracted medical phrases; and (3) Query Expansion using MeSH terms.

mayoan¶

Run ID: mayoan
Participant: MayoNLPTeam
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: baca97e1af7bc89fa2a666938c5b5507
Run description: Using the note field in each topic, we applied an ensemble model of (1) a Part-of-Speech query term weighting model; (2) a Markov Random Field Model of extracted medical phrases (by MedTagger); and (3) Query Expansion using MeSH terms.

mayoas¶

Run ID: mayoas
Participant: MayoNLPTeam
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: acfb19e713ef24c6aa341ef16f9dd9dd
Run description: Using the summary field in each topic, we applied an ensemble model of (1) a Part-of-Speech query term weighting model; (2) a Markov Random Field Model of extracted medical phrases; and (3) Query Expansion using MeSH terms.

mayomd¶

Run ID: mayomd
Participant: MayoNLPTeam
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: manual
MD5: b0120655be72bcb2dc69997a32c3acd1
Run description: Using the description field in each topic, we applied an ensemble model of (1) a Part-of-Speech query term weighting model; (2) a Markov Random Field Model of manually extracted medical phrases; and (3) Query Expansion using MeSH terms.

mayomn¶

Run ID: mayomn
Participant: MayoNLPTeam
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: manual
MD5: 7ce484cb254c5b55265b3aa5c5530589
Run description: Using the note field in each topic, we applied an ensemble model of (1) a Part-of-Speech query term weighting model; (2) a Markov Random Field Model of manually extracted medical phrases; and (3) Query Expansion using MeSH terms.

MRKPrfNote¶

Run ID: MRKPrfNote
Participant: MERCKKGAA
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 88cda94ce45ab409dada2f0e8ce6c7d5
Run description: Search with Notes on title, abstract, body

MRKSumCln¶

Run ID: MRKSumCln
Participant: MERCKKGAA
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 6260f8ef885554e8fc1a6f6dfed2b0a5
Run description: Summaries search on title, abstract, body with pseudo relevance feedback. Summaries have been cleaned to include only noun phrases.

MRKUmlsSolr¶

Run ID: MRKUmlsSolr
Participant: MERCKKGAA
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 9b15463be7b7e1398be68e2be7314fbe
Run description: Search using summaries in title, abstract, body. Enable Pseudo relevance feedback, enable UMLS query expansion.

MrkUmlsXgb¶

Run ID: MrkUmlsXgb
Participant: MERCKKGAA
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 3ce43374863b75d4c2a5a136c0db4cff
Run description: Search using summaries in title, abstract, body. Enable Pseudo relevance feedback, enable UMLS query expansion, enable gradient boosting-based reranking. Enable Wikipedia-based Word vectors

nacmmf¶

Run ID: nacmmf
Participant: HAUT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 00b3ff2373e422164a3c55baeb47cc4b
Run description: Abstract-based. Medical Text Indexer for query expansion; Article classification; Multi-model fusioni using note.

NDPHBo1C¶

Run ID: NDPHBo1C
Participant: IAII_PUT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 7ffa7ef63b7344c7666e128c2af65c3b
Run description: Document retieval using Terrier (with DPH Bo1) and query expansion - keywords for topic type.

NDPHBo1CM¶

Run ID: NDPHBo1CM
Participant: IAII_PUT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 6e738ac07c820b8e1263cd621feb9a48
Run description: Document retieval using Terrier (with DPH Bo1) and query expansion - keywords for topic type and MESH for medical terms in note.

nkuRun1¶

Run ID: nkuRun1
Participant: NKU
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 114135a4ab958a63f5c92be75cb44487
Run description: We extract the main part from articles and build a new collection. We indexed and retrieved with Terrier system. We combined the results which were retrieved by BM25 model and In_expB2 model. And we extend the type with some synonyms.

nkuRun2¶

Run ID: nkuRun2
Participant: NKU
Track: Clinical Decision Support
Year: 2016
Submission: 7/21/2016
Type: automatic
MD5: 22553a7e8865b1ab2d7b93730cd2c04e
Run description: We extract the main part from articles and build a new collection. We indexed and retrieved with Terrier system using BM25 weighting model and Pseudo relevance feedback. And we extend the type with some synonyms.

nkuRun3¶

Run ID: nkuRun3
Participant: NKU
Track: Clinical Decision Support
Year: 2016
Submission: 7/19/2016
Type: automatic
MD5: 8f56fb556b3aa0322e049c66f1111bdb
Run description: We extract the main part from articles and build a new collection. We indexed and retrieved with Terrier system using In_expB2 weighting model and Pseudo relevance feedback.

nkuRun4¶

Run ID: nkuRun4
Participant: NKU
Track: Clinical Decision Support
Year: 2016
Submission: 7/21/2016
Type: automatic
MD5: 3f051209af7e3d54362e716c3ade5e72
Run description: We extract the main part from articles and build a new collection. We indexed and retrieved with Terrier system using BM25 weighting model and Pseudo relevance feedback. And we extend the type with some synonyms.

nkuRun5¶

Run ID: nkuRun5
Participant: NKU
Track: Clinical Decision Support
Year: 2016
Submission: 7/19/2016
Type: automatic
MD5: 8b73a9cf4282201ba67800c93df61eac
Run description: We extract the main part from articles and build a new collection. We indexed and retrieved with Terrier system using In_expB2 weighting model and Pseudo relevance feedback.

NLMrun1¶

Run ID: NLMrun1
Participant: NLM_NIH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: dd63ab6d896e0b8a6d653d4fdc5b5a52
Run description: NLMrun1: an automatic run that uses the summaries and combines the results of two IR models In expB2 and TF-IDF.

NLMrun2¶

Run ID: NLMrun2
Participant: NLM_NIH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 29d0b2f68be1adf0247fba8b58a4582c
Run description: NLMrun2: uses MeSH for query expansion and BM25 for IR.

NLMrun3¶

Run ID: NLMrun3
Participant: NLM_NIH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: aa39caf7b48b96a7b3582fc8b3b0f7e8
Run description: NLMrun3: uses the IR model BM25 and the MeSH terms extracted from the summaries.

NLMrun4¶

Run ID: NLMrun4
Participant: NLM_NIH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 051ba28fc761f8c8b63d41b873da4233
Run description: NLMrun4: uses MeSH terms extracted from the notes and BM25 model for IR.

NLMrun5¶

Run ID: NLMrun5
Participant: NLM_NIH
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: b5fb363c135e7554a199efe385fc715c
Run description: NLMrun5 uses MeSH terms extracted from the notes and combines two IR models TF-IDF and In expB2.

NoteES¶

Run ID: NoteES
Participant: nch_risi
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: f29d0108abd6c27e3313184e5431a22a
Run description: UMLS to provide keywords. ES to rank using notes only. Last submission.

noteUIOWAS1¶

Run ID: noteUIOWAS1
Participant: UIowaS
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 4f0d59ee6e71563fe296408a908f6e95
Run description: We start by processing topics2016.xml and extract the 30 topics. Given the type (note, summary, or description) of topic, we use only that text category as input in a particular run. We use Metamap to extract all the UMLS concepts in the text. We use a subset of these extracted concepts in creating the query per topic. We also extract the age and gender of the patient from the text and use it in the query. Lastly we use indri to run the query on pre-indexed pmc dataset to get the final ranked list.

prna1sum¶

Run ID: prna1sum
Participant: prna
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 309f9bc29be3b063d5cfde54fab1a43f
Run description: We used an automated multiple-steps driven method to extract relevant biomedical articles corresponding to each given topic. We performed clinical concepts extraction with ontology mapping for identifying important IDF-weighted topical keywords from the given topic summaries, which were used to extract relevant diagnoses, test, and treatment concepts from Wikipedia clinical medicine category articles embedded into a knowledge graph architecture. Ultimately, the Wiki concepts related to relevant diagnoses, tests, and treatments were used in mapping pertinent biomedical articles, which were further filtered by named entity and demographic information, and ordered by publication date and importance in relation to the extracted Wiki keywords.

prna2desc¶

Run ID: prna2desc
Participant: prna
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 96c6f62ff161e664dbdebb67dd3f75e8
Run description: We used an automated multiple-steps driven method to extract relevant biomedical articles corresponding to each given topic. We performed clinical concepts extraction with ontology mapping for identifying important IDF-weighted topical keywords from the given topic descriptions, which were used to extract relevant diagnoses concepts from Wikipedia clinical medicine category articles embedded into a knowledge graph architecture. Ultimately, the Wiki concepts related to relevant diagnoses, tests, and treatments were used in mapping pertinent biomedical articles, which were further filtered by named entity and demographic information, and ordered by publication date and importance in relation to the extracted Wiki keywords.

prna3note¶

Run ID: prna3note
Participant: prna
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 2e6566db9beb5152980ec61a1600ede4
Run description: We used an automated multiple-steps driven method to extract relevant biomedical articles corresponding to each given topic. We performed clinical concepts extraction with ontology mapping for identifying important IDF-weighted topical keywords from the given topic notes, which were used to extract relevant diagnoses concepts from Wikipedia clinical medicine category articles. Ultimately, the Wiki concepts related to relevant diagnoses, tests, and treatments were used in mapping pertinent biomedical articles, which were further filtered by named entity and demographic information, and ordered by publication date and importance in relation to the extracted Wiki keywords.

prna4note¶

Run ID: prna4note
Participant: prna
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 6a0455caa94523f547fa22d50a85ff6b
Run description: We used an automated multiple-steps driven method to extract relevant biomedical articles corresponding to each given topic. We performed clinical concepts extraction with ontology mapping for identifying important IDF-weighted topical keywords from the given topic notes, which were used to extract relevant diagnoses concepts from Wikipedia clinical medicine category articles embedded into a knowledge graph architecture. Ultimately, the Wiki concepts related to relevant diagnoses, tests, and treatments were used in mapping pertinent biomedical articles, which were further filtered by named entity and demographic information, and ordered by publication date and importance in relation to the extracted Wiki keywords.

prna5note¶

Run ID: prna5note
Participant: prna
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 0ae2cd7f2606e39403e2e7b84fc03518
Run description: We used an automated multiple-steps driven method to extract relevant biomedical articles corresponding to each given topic. For each topic note, we predict the differential diagnoses concepts using a deep learning model trained on a collection of MIMIC II notes and Wikipedia clinical medicine category articles. Ultimately, the Wiki concepts related to relevant diagnoses, tests, and treatments were used in mapping pertinent biomedical articles, which were further filtered by named entity and demographic information, and ordered by publication date and importance in relation to the extracted Wiki keywords.

RONE¶

Run ID: RONE
Participant: hany-miner
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: ca6975a8b73227f4d6475a252885dff6
Run description: 5. This run used the vector space model described in article (http://www.r-bloggers.com/build-a-search-engine-in-20-minutes-or-less/), which a basic search model using R. Term Frequency Document was created first for all the 1.25 million documents and the Description text from the list of topics was used for querying the vector space.

run1¶

Run ID: run1
Participant: iris
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 5a8c4378015c502aa414891791239fa1
Run description: extract UMLS terms, predict the diagnosis with wikipedia, expand org query with prediction

run2¶

Run ID: run2
Participant: iris
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 2115de0c3af6893fb2af593da8ce5968
Run description: extract UMLS terms, predict the diagnosis with wikipedia, expand org query with prediction

run3¶

Run ID: run3
Participant: iris
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: ed053ef2d4340fe4dd69274e15776c9c
Run description: extract UMLS terms, predict the diagnosis with wikipedia, expand org query with prediction

run4¶

Run ID: run4
Participant: iris
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: manual
MD5: dd7959713e78eb91f62f6faf21dca413
Run description: extract UMLS terms, predict the diagnosis with wikipedia, expand org query with prediction

run5¶

Run ID: run5
Participant: iris
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: manual
MD5: 5f20385412899d9b77a8b5a66da3ba2c
Run description: extract UMLS terms, predict the diagnosis with wikipedia, expand org query with prediction

sacmmf¶

Run ID: sacmmf
Participant: HAUT
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 5bee7038479ef6e666340e1f033c6329
Run description: Abstract based. Medical Text Indexer for query expansion; Article classification; Multi-model fusion Using summary.

SDPHBo1NE¶

Run ID: SDPHBo1NE
Participant: IAII_PUT
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 222a6d3c9df6a2dc9b4ca0749864c278
Run description: This is a baseline Terrier search with DPH Bo1 option and no additional query expansions.

SumClsRerank¶

Run ID: SumClsRerank
Participant: nch_risi
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: ff5f4f611d1c296db4e9b8e56b44fa2d
Run description: UMLS for keywords extraction and expansion. Three categories of keywords: previous, present and undecided. Elastic search to provide an initially ranked list candidates. Logistic regression classifier, trained over 2014-2015 data, for reranking.

SumCmbRank¶

Run ID: SumCmbRank
Participant: nch_risi
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: f6e6e8cdbbe5ec1acb6b608baee21264
Run description: UMLS for keywords extraction and expansion. Three categories of keywords: previous, present and undecided. Elastic search to provide an initially ranked list candidates. Logistic regression classifier, trained over 2014-2015 data, for reranking. Using the average rank of ES and classification as the final rank.

SumES¶

Run ID: SumES
Participant: nch_risi
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 83849e80c13eb32a8c64834d38ceff28
Run description: UMLS for keywords extraction and expansion. Three categories of keywords: previous, present and undecided. Elastic search to provide an initially ranked list candidates. This is ES ranking only. PM25, boosting title, abstract and body: 2,4,1. Boost previous, present and undecided 2(1),3,2

summUIOWAS3¶

Run ID: summUIOWAS3
Participant: UIowaS
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 72941ae515b67d5b1a1c9a9afded5b05
Run description: We start by processing topics2016.xml and extract the 30 topics. Given the type (note, summary, or description) of topic, we use only that text category as input in a particular run. We use Metamap to extract all the UMLS concepts in the text. We use a subset of these extracted concepts in creating the query per topic. We also extract the age and gender of the patient from the text and use it in the query. Lastly we use indri to run the query on pre-indexed pmc dataset to get the final ranked list.

UDelInfoCDS1¶

Run ID: UDelInfoCDS1
Participant: udel_fang
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 7501f50d1ec6820be5eb05f259edfe00
Run description: Use the note query. Only keep the root noun phrases as the query. External resources: CTAKEs, MetaMap

UDelInfoCDS2¶

Run ID: UDelInfoCDS2
Participant: udel_fang
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 588533b76cf05751e4edab9b97618908
Run description: Use the note query. Keep the root noun phrases as the query, add the UMLS expansion for the abbreviations. External Resources: CTAKEs, MetaMap

UDelInfoCDS3¶

Run ID: UDelInfoCDS3
Participant: udel_fang
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: fbfb5c682ed02895ce5650e946e84dba
Run description: Use the note query. Keep the root noun phrases, add the UMLS expansion and MediLexicon expansion together. External Resources: CTAKEs, MetaMap, MediLexicon

UDelInfoCDS4¶

Run ID: UDelInfoCDS4
Participant: udel_fang
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 67b2e7b34720e9b6db25345956df1829
Run description: Use the description query. Only keep the root noun phrases as the query. External resources: CTAKEs, MetaMap

UDelInfoCDS5¶

Run ID: UDelInfoCDS5
Participant: udel_fang
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: bf7778e8dadba5c938b1b9b5f9b5a2d2
Run description: Use the summary query. Pseudo Relevance Feedback is applied to the original query.

udelNB¶

Run ID: udelNB
Participant: udel
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 5fab8e7bea08281418d26faa2884a024
Run description: A baseline run by the information retrieval lab at the University of Delaware. It uses the note version of queries and it is produced by Terrier IR platform with In_expB2 model and KL divergence for query expansion.

udelNRef¶

Run ID: udelNRef
Participant: udel
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 0f7a8b33cad4ec53d92ac6b2e227b6ee
Run description: A run by the information retrieval lab at the University of Delaware. It uses the note version of queries and it is produced by Terrier IR platform with In_expB2 model and KL divergence for query expansion. In addition, this run uses an automatic reformulation approach based on term re-weighing such that terms that are non relevant to the clinical query are removed whereas other terms are either boosted (due to their importance) or have their weights reduced (because they are less important).

udelSB¶

Run ID: udelSB
Participant: udel
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 84577981fd52d920e21ee7f877800a25
Run description: A baseline run by the information retrieval lab at the University of Delaware. It uses the summary version of queries and produced by Terrier IR platform with In_expB2 model and KL divergence for query expansion.

udelSDI¶

Run ID: udelSDI
Participant: udel
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 0f170ecd3f0d6196bbd3851a7c3925c3
Run description: A run by the information retrieval lab at the University of Delaware. It uses the summary version of queries and it is produced by Terrier IR platform with In_expB2 model and KL divergence for query expansion. In addition, this run uses external resource (mayo clinic) to inffer the possible diagnosis for the last 20 topics (tests and treatments) by expanding the 5 most relevant diagnoses for each clinical case.

udelSRef¶

Run ID: udelSRef
Participant: udel
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 26cf81e9ed8d13260616e9f2579279f3
Run description: A run by the information retrieval lab at the University of Delaware. It uses the summary version of queries and it is produced by Terrier IR platform with In_expB2 model and KL divergence for query expansion. In addition, this run uses an automatic reformulation approach based on term re-weighing such that terms that are non relevant to the clinical query are removed whereas other terms are either boosted (due to their importance) or have their weights reduced (because they are less important).

UNTIIANA¶

Run ID: UNTIIANA
Participant: UNTIIA
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: ef959b10c9bb88d5f9d76cbd1f882733
Run description: This run uses Note section. Terrier IR system is used for indexing and retrieval and Python regular expressions are rigorously used in preprocessing the PubMed data-set.

UNTIIANM¶

Run ID: UNTIIANM
Participant: UNTIIA
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: manual
MD5: 1d1594978793127030fa69db5b670cd0
Run description: This run is generated using the key terms in the Note section. The key terms are extracted manually by a medical expert. Additionally we used Terrier IR system for indexing and retrieval and UMLS Meta-thesaurus for getting the full form of acronyms and abbreviations in the note section.

UNTIIANMERG¶

Run ID: UNTIIANMERG
Participant: UNTIIA
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 74312f7df9b4182d159545f2818c398b
Run description: This run is an automatic run generated using the Note section of the topics. We used Terrier IR system for indexing and retrieval. For this run results of 5 different weighing models are merged together and re-ranked based on their normalized similarity score.

UNTIIASA¶

Run ID: UNTIIASA
Participant: UNTIIA
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 2590878f6861e367f950f50f491c3fd2
Run description: This run uses summery section of the topic as query, terrier IR system is used for indexing and retrieval. Additionally Python Regular expressions are extensively used in preprocessing the PubMed articles.

UNTIIASMERG¶

Run ID: UNTIIASMERG
Participant: UNTIIA
Track: Clinical Decision Support
Year: 2016
Submission: 7/27/2016
Type: automatic
MD5: 502aa5cbf947a82e9ce53d0ced8d4567
Run description: This run uses the summary section of the topics and we used Terrier IR system for indexing and retrieval. The ranking of this run is a merge of results from 5 different weighing models. The re-ranking is based on the normalized score after merging.

UWM0¶

Run ID: UWM0
Participant: UWM
Track: Clinical Decision Support
Year: 2016
Submission: 7/20/2016
Type: automatic
MD5: 597d88ffd42b860d53a3e6ef04783666
Run description: Base run. The summary fields from original topics (http://trec-cds.appspot.com/topics2016.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.

UWM1¶

Run ID: UWM1
Participant: UWM
Track: Clinical Decision Support
Year: 2016
Submission: 7/20/2016
Type: automatic
MD5: 8b0426e8ee4e06dccb8bfc4c4a6eac50
Run description: Query expansion with the words of reference titles. Top 5 retrieved documents for each topic were chosen from base run. The frequency of the words that appeared in the reference titles for top 5 retrieved documents has been counted. Frequently occurred words, which are included in the reference titles of each document, have been selected as the words for query expansion in addition to the base run query. , Selected words ordered by frequency (descending order) have been added up to maximum 10. The words with same lowest frequency have been removed in case that the total number of words exceeds 10. The optimal number of top retrieved documents, 5, and the optimal number of words, 10 have been chosen by referring to the results based on 2015 CDS Track dataset. In order to collect data of references and citing articles for each retrieved document, Europe PCM articles REST(Representational State Transfer)ful APIs ( http://europepmc.org/RestfulWebService) have been used. Stop words have been removed.

UWM2¶

Run ID: UWM2
Participant: UWM
Track: Clinical Decision Support
Year: 2016
Submission: 7/20/2016
Type: automatic
MD5: b7b7e7b58a74355116cbc2dca17583e6
Run description: Query expansion with the words of titles whose articles are citing top 5 retrieved documents for each topic. In a similar fashion to UWM1, Top 5 retrieved documents for each topic were chosen from base run. The frequency of the words that appeared in the titles has been counted. Frequently occurred words, which are included in the titles of documents that each retrieved document is cited by, have been selected as the words for query expansion in addition to the base run query. Some queries are same to original queries if they are not cited by other articles. Selected words ordered by frequency (descending order) have been added up to maximum 10. The words with same lowest frequency have been removed in case that the total number of words exceeds 10. The optimal number of top retrieved documents, 5, and the optimal number of words, 10 have been chosen by referring to the results based on 2015 CDS Track dataset. In order to collect data of references and citing articles for each retrieved document, Europe PCM articles REST(Representational State Transfer)ful APIs ( http://europepmc.org/RestfulWebService) have been used. Stop words have been removed.

WHUIRGroup1¶

Run ID: WHUIRGroup1
Participant: WHUIRGroup
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 913643215657666c9e6f547a108e74ad
Run description: We just use the nxml2txt tools process the PubMed documents. The use the Indri to Index. For the topic, we choose the notes field. We use smartStopWords both in indics and queries. Last, we use the Indri Language Model to get the result.
Code: https://github.com/spyysalo/nxml2txt

WHUIRGroup2¶

Run ID: WHUIRGroup2
Participant: WHUIRGroup
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 07e941d97b8192545d8a199958845108
Run description: We use the Indri to Index the PubMed documents. The field we used are "abstract","title" and "body". We use smartStopWords both in indics and queries. Then we use word2vec to expand the queries. We used the expanded queries and Indri LM model to get the run result.

WHUIRGroup4¶

Run ID: WHUIRGroup4
Participant: WHUIRGroup
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: c554b0f6e9a5443c5ce22ede4ebe0fc9
Run description: We just use the nxml2txt tools process the PubMed documents. Then use the Indri to Index gettingIndex1. We use the Indri to Index the PubMed documents. The field we used are "abstract","title" and "body".Then use the Indri to Index getting Index2. We use smartStopWords both in indics and queries. We used these Indics and BM25,TFIDF and LM to query the summaries query,get 6 runs. For each run,we can get the q-d's score and rank.We used these numbers as features.There is 12 features for each q-d pair. We used RankLib to train model and predict the result. The training data is CDS2015topics running in these two indics. The Learning Method is LambdaMART.
Code: https://sourceforge.net/p/lemur/wiki/RankLib/ https://github.com/spyysalo/nxml2txt

WHUIRGroup5¶

Run ID: WHUIRGroup5
Participant: WHUIRGroup
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: 61b235ca94cc1025b5c78b3bb670cbe2
Run description: We use the same way with WHHUIRGroup2 runs.Just replace the queries.In this run,we use notes instead of summaries.

WHUIRGroup6¶

Run ID: WHUIRGroup6
Participant: WHUIRGroup
Track: Clinical Decision Support
Year: 2016
Submission: 7/28/2016
Type: automatic
MD5: f27db28fa80b0565c0db1799fb60a1f3
Run description: There is something wrong with WHUIRGroup3. This is a new Edition of WHUIRGroup3. Hope to replace the former one. Thanks!
Code: https://sourceforge.net/p/lemur/wiki/RankLib/ https://github.com/spyysalo/nxml2txt