Runs - Clinical Decision Support 2014¶

atigeo1¶

Run ID: atigeo1
Participant: atigeo
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: d425043d6b4617ab2d156d98ad6d3f3d
Run description: indri.age.none.norm.mesh.final_topics.20140725.indri.none.neg.none.mesh.final_topics.20140725 Uses NegEx & Metamap

atigeo2¶

Run ID: atigeo2
Participant: atigeo
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 44e3fce7b574bb8d83ff6e90e314504b
Run description: indri.age.none.norm.mesh.final_topics.20140725.indri.none.neg.none.mesh.final_topics.20140725 Uses NegEx & Metamap

atigeo3¶

Run ID: atigeo3
Participant: atigeo
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 1be8492588c9755c8d3c8e07abca3694
Run description: solr.age.neg.norm.none.final_topics.20140725.indri.age.neg.norm.mesh.final_topics.20140725 Uses NegEx & Metamap

atigeo4¶

Run ID: atigeo4
Participant: atigeo
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 0ca40fe17ce44175c4fbdd896784400d
Run description: solr.age.neg.norm.none.final_topics.20140725.indri.age.neg.norm.mesh.final_topics.20140725 Uses NegEx & Metamap

atigeo5¶

Run ID: atigeo5
Participant: atigeo
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 07c0e1bc09883c9553a7113fb425b1d3
Run description: indri.age.neg.norm.mesh.final_topics.20140725 Uses NegEx & Metamap

bacon¶

Run ID: bacon
Participant: BigPig
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: d51456d54cc5ea30090c4893aab3c6c5
Run description: metamap

BiTeMSIBtex1¶

Run ID: BiTeMSIBtex1
Participant: BiTeM_SIBtex
Track: Clinical Decision Support
Year: 2014
Submission: 7/22/2014
Type: automatic
Task: main
MD5: d73f5f6a3bc5eea9de1ba95aaaf11cd8
Run description: These are the strategies we used for run 1. - we indexed at the document level - we had one index - we added mesh terms to the text. Mesh terms were both imported from MEDLINE (and repeated 10 times), and locally mapped with a classical Rabin-Karp algorithm (repeated as many times as they were mapped). MeSH terms were also mapped and added in the queries. - we used an additional "MeSHtarget" strategy : MeSHtargets are aimed to represent the question type (diagnosis, test, or treatment). We exploited the UMLS Semantic Network in order to have three MeSH subsets for each type. Then, we added a specific term to the document ("meshtargetdiag", "meshtargettreat", "meshtargettest"), each time we mapped a mesh term belonging to one of these subsets. The query also contained the asked MeSHtargets (e.g. "meshtargetdiag" for queries from 1 to 10) repeated three times. - reranking: if the article type was review-article or case-report, its score had a +20% boost. - IR was performed with Terrier, stopwords and Porter stemming, weighting scheme BM25, and query expansion. All numbers were removed from the queries. External resources : MeSH.

BiTeMSIBtex2¶

Run ID: BiTeMSIBtex2
Participant: BiTeM_SIBtex
Track: Clinical Decision Support
Year: 2014
Submission: 7/22/2014
Type: automatic
Task: main
MD5: 503aa037b3c61763f8d9e7d19cbc6c79
Run description: These are the strategies we used for run 2. - same strategies than run 1, but we used two different indexes: one for text and meshtargets, the other with mesh terms and meshtargets. Queries contained respectively text and meshtargets, and mesh terms and meshtargets. Results from two indexes were merged by linear combination (same weight to both indexes). External resources : MeSH.

BiTeMSIBtex3¶

Run ID: BiTeMSIBtex3
Participant: BiTeM_SIBtex
Track: Clinical Decision Support
Year: 2014
Submission: 7/23/2014
Type: automatic
Task: main
MD5: f223e31a39cf01e260742431f53053d6
Run description: These are the strategies we used for run 3. - same strategies than run 2, but we indexed at the section level. One document per section was created. For MeSH terms retrieved from MEDLINE, they were repeated three times instead of ten. We retrieved sections but kept the document id. We kept the score of the first retrieved section. - the MeSHtargets strategy was not applied External resources : MeSH.

BiTeMSIBtex4¶

Run ID: BiTeMSIBtex4
Participant: BiTeM_SIBtex
Track: Clinical Decision Support
Year: 2014
Submission: 7/22/2014
Type: automatic
Task: main
MD5: 46d105f2c722d932c9a889e5b03a627f
Run description: These are the strategies we used for run 4. - same strategies than run 3, but the MeSHtargets strategy was applied External resources : MeSH.

BiTeMSIBtex5¶

Run ID: BiTeMSIBtex5
Participant: BiTeM_SIBtex
Track: Clinical Decision Support
Year: 2014
Submission: 7/22/2014
Type: automatic
Task: main
MD5: 44d275ef8229cd857ae54aa7c8173198
Run description: These are the strategies we used for run 5. - same strategies than run 4, but we used a citation network reranking strategy. For each document, we extracted the citations that belongs to the collection. Then we reranked all the run: for each retrieved document i (and its score RSVi), all its cited documents (whether they already were retrieved or not) had a 0.2*RSVi boost. It means that documents that are often cited by a lot of top retrieved documents have a strong boost. External resources : MeSH.

BM25¶

Run ID: BM25
Participant: DawitAfshin
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: manual
Task: main
MD5: 4cf9091438d8207db7a0dcde7fab42f3
Run description: This method uses query expansion/optimization to get the retrieved results. Terrier version 3.6 was used to index and retrieve using BM25. The summary queries were used to retrieve files from the collection.

BM25EXP¶

Run ID: BM25EXP
Participant: DawitAfshin
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: manual
Task: main
MD5: d359041603a1b61cdb2699e916b1f135
Run description: This method uses query expansion/optimization to get the retrieved results. Terrier version 3.6 was used to index and retrieve using BM25. The summary queries were used to retrieve files from the collection. SNOMED CT, METAMAP, UMLS Metathesaurus were used to expand the query.

bolgogi¶

Run ID: bolgogi
Participant: BigPig
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 74aa406fab2f9e07f3742fcb6500e7dd
Run description: metamap, lucene-bm25similarity

DAIICTdqep¶

Run ID: DAIICTdqep
Participant: DA_IICT
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 2c5b209dfc6746689835ce4827ca94e2
Run description: The run DAIICTdqep is generated using parameter free automatic query expansion technique.

DAIICTdqer8¶

Run ID: DAIICTdqer8
Participant: DA_IICT
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 88fa69800f19bf40b091c3a5cf735e55
Run description: The run DAIICTdqer8 is generated using automatic query expansion technique rocchio with beta parameter as 0.8 .

DAIICTf¶

Run ID: DAIICTf
Participant: DA_IICT
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: fcd78f1fd15e136f072e677e6961e748
Run description: The submitted run DAIICTf is a fused result (by combSUM fusion method) of four different experiments with various automatic query expansion techniques.

DAIICTsqer8¶

Run ID: DAIICTsqer8
Participant: DA_IICT
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: d82eda003b6a69079a990b588f94fc79
Run description: The run DAIICTsqer8 is generated using automatic query expansion technique rocchio with beta parameter as 0.8 .

DAIICTzf¶

Run ID: DAIICTzf
Participant: DA_IICT
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 97e65a25d8d702816a6d12b2f289d51e
Run description: The submitted run DAIICTzf is a fused result (by z-fusion method) of four different experiments with various automatic query expansion techniques.

descript50ex¶

Run ID: descript50ex
Participant: CSEIITV
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: automatic
Task: main
MD5: c01d23c8fac85b47dd9bbb89c19f986a
Run description: Descriptions were used in this run. Only 50 documents were ranked in this run. For initial pruning of documents we used Lucene.

ecnuBig¶

Run ID: ecnuBig
Participant: ecnu
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: automatic
Task: main
MD5: 292f621b8cc9bb83fb7de64da38a093e
Run description: We retrieve by Terrier and Indri search engine with BM25, DFR-BM25, TF-IDF, BB2 and PL2 weight model. Then we combine all of our results together.

ecnuIndex¶

Run ID: ecnuIndex
Participant: ecnu
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: automatic
Task: main
MD5: d9a2d3d4de80dfa6bcef56433b46fbc8
Run description: We use Terrier search engine and BM25F weight model. We create index respectively for title, abstract, text, table, figure and reference. We retrieve with the six indexs and combine the results together in the end.

ecnuSmall¶

Run ID: ecnuSmall
Participant: ecnu
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: automatic
Task: main
MD5: e7e3e15c6e1310273f773097a0c59ab6
Run description: We retrieve with Terrier in BM25 weight model and Indri in TFIDF model. Then we combine the results together.

ecnuWeight¶

Run ID: ecnuWeight
Participant: ecnu
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: automatic
Task: main
MD5: b60a77ce777dc4cbbe8bdec2b0c21f1b
Run description: In query expansion, we automatic assign weight to the key words in query. Then we retrieve by Indri.

GuHNegProxL¶

Run ID: GuHNegProxL
Participant: Georgetown
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: a556047dded04abc14ed7dbd5e98e325
Run description: Georgetown Run 2: Lemur Indri search engine was used for indexing and retrieval (language modeling with Dirichlet smoothing), UMLS Metathesaurus and Meta Map were used to detect medical terms and provide synonyms. The Stanford NLP Parser was used to add nouns to the list of medical terms for topics that were classified as difficult. Idf filtering was applied. NegEx was used to detect negated phrases. Overly common words among queries were also detected by indexing queries and using an idf threshold. Weights of common terms among queries and of terms mapping from negated phrases were reduced.

GuHSINeg¶

Run ID: GuHSINeg
Participant: Georgetown
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 447aac3b7f003b4271b403d2bd776307
Run description: Georgetown Run 1: Lemur Indri search engine was used for indexing and retrieval (language modeling with Dirichlet smoothing), UMLS Metathesaurus and Meta Map were used to detect medical terms and provide synonyms. The Stanford NLP Parser was used to add nouns to the list of medical terms for topics that were classified as difficult. Idf filtering was applied. Then, a second level stricter filtering of the medical terms was applied using the Meta Map classifications. NegEx was used to detect negated phrases. Weights of terms mapping from negated phrases were reduced.

GuHSINegL¶

Run ID: GuHSINegL
Participant: Georgetown
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: 2e85e2368c629c5cb6f348ae1af2ccb3
Run description: Georgetown Run 5 Same approach as run 1, but further reduced weight of terms mapped from negated phrases.

GuHSNegProxH¶

Run ID: GuHSNegProxH
Participant: Georgetown
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 140292e61f7878679913347e8236b1de
Run description: Georgetown Run 3: Lemur Indri search engine was used for indexing and retrieval (language modeling with Dirichlet smoothing), UMLS Metathesaurus and Meta Map were used to detect medical terms and provide synonyms. The Stanford NLP Parser was used to add nouns to the list of medical terms for topics that were classified as difficult. Idf filtering was applied. Then, a second level stricter filtering of the medical terms was applied using the Meta Map classifications. NegEx was used to detect negated phrases. Overly common words among queries were also detected by indexing queries and using an idf threshold. Weights of common terms among queries and of terms mapping from negated phrases were reduced.

GuHSNegProxL¶

Run ID: GuHSNegProxL
Participant: Georgetown
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 50273b9424fbba9d56558613a8704026
Run description: Georgetown Run 4: Lemur Indri search engine was used for indexing and retrieval (language modeling with Dirichlet smoothing), UMLS Metathesaurus and Meta Map were used to detect medical terms and provide synonyms. The Stanford NLP Parser was used to add nouns to the list of medical terms for topics that were classified as difficult. Idf filtering was applied. NegEx was used to detect negated phrases. Overly common words among queries were also detected by indexing queries and using an idf threshold. Weights of common terms among queries and of terms mapping from negated phrases were reduced. This differs from run 3 by using a lower threshold for words common among queries.

hltcoe5drf¶

Run ID: hltcoe5drf
Participant: hltcoe
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: 45bca907669bc407e6a5847e34e56890
Run description: JHU HAIRCUT system - statistical language model. Padded, word-spanning character 5-grams as tokens. Relevance Feedback. No domain customization. No resources. Descriptions instead of summaries from topics.

hltcoe5s¶

Run ID: hltcoe5s
Participant: hltcoe
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: 1636d9761eb62940f75ed438662f6dca
Run description: JHU HAIRCUT system - statistical language model. Padded, word-spanning character 5-grams as tokens. No relevance Feedback employed. No domain customization. No resources. (Note: all three submitted 5-gram runs suffered from a build error where approx. 3.5% of the corpus was mistakenly not indexed. At submission time we're rebuilding the index, but it'll be some hours still before corrected runs are available. The word-based run used the full text corpus, excepting two problematic documents our parser didn't like.)

hltcoe5srf¶

Run ID: hltcoe5srf
Participant: hltcoe
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: d4caee60eb21bb259fb3f8f45303af15
Run description: JHU HAIRCUT system - statistical language model. Padded, word-spanning character 5-grams as tokens. Relevance Feedback. No domain customization. No resources.

hltcoewsrf¶

Run ID: hltcoewsrf
Participant: hltcoe
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: b7119072746b7661b8737aa8eca9bfb7
Run description: JHU HAIRCUT system - statistical language model. Words as tokens. Relevance Feedback. No domain customization. No resources.

icd¶

Run ID: icd
Participant: cuhk_sls
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: b6c14694e76aa08eec867636740d6ab9
Run description: Words not found in the ICD-10 data were used to filter out terms in the query file. External resources are Terrier and ICD-10.

icdqe¶

Run ID: icdqe
Participant: cuhk_sls
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: ca0936fe98d700d654cdf993af535dc8
Run description: ICD-10 was indexed so that each term is made into a document. Queries were first submitted to this ICD-10-based system to locate terms that can be appended to the original query. These modified queries were then used on the original document collection. External resources: Terrier and ICD-10.

InL2c1¶

Run ID: InL2c1
Participant: DawitAfshin
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: manual
Task: main
MD5: 5678968bc065e6dcd7d06297a9716f53
Run description: This method uses query expansion/optimization to get the retrieved results. Terrier version 3.6 was used to index and retrieve using InL2c1. The summary queries were used to retrieve files from the collection.

InL2c1EXP¶

Run ID: InL2c1EXP
Participant: DawitAfshin
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: manual
Task: main
MD5: d0efafb485c54ea746a46d035a78b38d
Run description: This method uses query expansion/optimization to get the retrieved results. Terrier version 3.6 was used to index and retrieve using InL2c1. The summary queries were used to retrieve files from the collection. SNOMED CT, METAMAP, UMLS Metathesaurus were used to expand the query.

IRGURUN1¶

Run ID: IRGURUN1
Participant: georgetown_ir
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 2ee914bdb73589f4e846270ad3e6fb4c
Run description: The documents were indexed using ElasticSearch (a modern implementation of Lucene) and retrieved Divergence from Randomness (DRF) model. The original query is both expanded using Pseudo Relevance Feedback and contracted by filtering query terms based on their presence on health related Wikipedia pages. The retrieved results where classified into treatment, test or diagnosis using a SVM trained with a set of manually annotated papers randomly sampled from the collection. Terms were used as features (except stopwords), as well as MetaMap concepts in the title and in the abstract (when available). Finally, results were clustered and reranked

IRGURUN2¶

Run ID: IRGURUN2
Participant: georgetown_ir
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 74a5d15a21501a1e442c725e5e0f6911
Run description: The documents were indexed using ElasticSearch (a modern implementation of Lucene) and retrieved Divergence from Randomness (DRF) model. The original query is both expanded using Pseudo Relevance Feedback and contracted by filtering query terms based on their presence on health related Wikipedia pages. For those queries asking for a test or a diagnosis, biographical information were extracted from the retrieved documents and the query via pattern matching and used to rerank the list of results. For treatments, a SVM was trained with a set of manually annotated papers randomly sampled from the collection. We kept as features those terms whose stemmed version matched the stemmed version of a set of seed words related to treatments.

IRGURUN3¶

Run ID: IRGURUN3
Participant: georgetown_ir
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 285a693c2328012959fac08d764d44ed
Run description: The documents were indexed using ElasticSearch (a modern implementation of Lucene) and retrieved Divergence from Randomness (DRF) model. The original query is both expanded using Pseudo Relevance Feedback and contracted by filtering query terms based on their presence on health related Wikipedia pages. Documents where reranked based on the presence of the terms treatment, test and diagnosis in them. Different sections were weighted by the normalized inverse of their length.

IRGURUN4¶

Run ID: IRGURUN4
Participant: georgetown_ir
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 1ea7fde5f0ac3ba20a875ffc82ab3210
Run description: The documents were indexed using ElasticSearch (a modern implementation of Lucene) and retrieved Divergence from Randomness (DRF) model. The original query is both expanded using Pseudo Relevance Feedback and contracted by filtering query terms based on their presence on health related Wikipedia pages. Documents where reranked based on how many MetaMap concepts they shared with the original query.

IRGURUN5¶

Run ID: IRGURUN5
Participant: georgetown_ir
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 1c56347d6dd86a94bc9d6231404f3271
Run description: Fusion retrieval system that combines the methods used in the other four submitted runs to generate a list of ranked results for each query. Each system is weighted equally.

KISTI01¶

Run ID: KISTI01
Participant: KISTI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: ca9f4c9d5735cc1ca8933f9cc01b01e3
Run description: Simple baseline using Lucene's language model

KISTI02¶

Run ID: KISTI02
Participant: KISTI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 951ae5cf4c1a8d491ab3422fe737244b
Run description: Reranking using PRF by considering abbreviations based on the baseline

KISTI03¶

Run ID: KISTI03
Participant: KISTI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 97edc8fdc6abfae565761c5c3a906dea
Run description: Reranking using ESA with ICD10-hierarchy

KISTI04¶

Run ID: KISTI04
Participant: KISTI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: a29cbde487b9fa2202fe8643e1e82a24
Run description: Reranking using MetaMap

KISTI05¶

Run ID: KISTI05
Participant: KISTI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: c89116a89593961646a58d11167bdd43
Run description: Reranking using MetaMap

manual¶

Run ID: manual
Participant: cuhk_sls
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: manual
Task: main
MD5: ff5f94b8a57a3658e3f51e13cfcdfcab
Run description: Entirely manual modification of the queries by having a domain expert select words that she deemed important in understanding the case. Would be an interesting baseline to our other techniques which either add terms or remove terms automatically. External resources is just Terrier.

MERCK1¶

Run ID: MERCK1
Participant: Merck_DA
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 60e66ba755648abc67f0d29e795eef56
Run description: NER with MER skill cartridge Indexing with Luxid Query with raw summary text

MERCK2¶

Run ID: MERCK2
Participant: Merck_DA
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 2324cf4191a043e65fb8d90c0367e138
Run description: 1. Named entity recognition with MER skill cartridge 2. Indexing with Luxid 3. Query with pre-indexed entities

MERCK3¶

Run ID: MERCK3
Participant: Merck_DA
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: a4cc825f0216aa0e471dcbc2bb8fcd44
Run description: NER with MER skill cartridge Indexing with Luxid Query with pre-indexed entities and restrict to "Case reports"

mesh¶

Run ID: mesh
Participant: cuhk_sls
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: d34c74994d2ec3e9605991a44f9233a2
Run description: Words not found in the MeSH data were used to filter out terms in the query file. External resources are Terrier and MeSH.

MIIjmab¶

Run ID: MIIjmab
Participant: UCLA_MII
Track: Clinical Decision Support
Year: 2014
Submission: 7/25/2014
Type: automatic
Task: main
MD5: f4b7417925701f7a129aa673408c4bde
Run description: Lucene English Analyzer, Jelinek-Mercer similarity metric, indexed with unigrams and UMLS CUIs. Queries were composed automatically, expanding query based on population attributes, temporal keywords, and clinical question type. Searched titles and abstracts only.

MIIjmboost¶

Run ID: MIIjmboost
Participant: UCLA_MII
Track: Clinical Decision Support
Year: 2014
Submission: 7/25/2014
Type: automatic
Task: main
MD5: acf395af51a0083d4e4dbbfbe0ab2197
Run description: Lucene English Analyzer, Jelinek-Mercer similarity metric, indexed with unigrams and UMLS CUIs. Queries were composed automatically, expanding query based on population attributes, temporal keywords, and clinical question type. Boosted some semantic types in the query: population, symptoms, findings, diseases.

MIIjmignore¶

Run ID: MIIjmignore
Participant: UCLA_MII
Track: Clinical Decision Support
Year: 2014
Submission: 7/25/2014
Type: automatic
Task: main
MD5: 7a6f12af73ae6a2ee2a783d81be92027
Run description: Lucene English Analyzer, Jelinek-Mercer similarity metric, indexed with unigrams and UMLS CUIs. Queries were composed automatically, expanding query based on population attributes, temporal keywords, and clinical question type.

MIItfauto¶

Run ID: MIItfauto
Participant: UCLA_MII
Track: Clinical Decision Support
Year: 2014
Submission: 7/25/2014
Type: automatic
Task: main
MD5: 11a785195461ab578d07e62d07f47c72
Run description: Lucene Standard Analyzer, default similarity metric, indexed with unigrams and UMLS CUIs. Queries were composed automatically, expanding query based on population attributes, temporal keywords, and clinical question type.

MIItfman¶

Run ID: MIItfman
Participant: UCLA_MII
Track: Clinical Decision Support
Year: 2014
Submission: 7/25/2014
Type: manual
Task: main
MD5: 8be1fce1151b42e7b3489e24c6d1e53c
Run description: Lucene Standard Analyzer, default similarity metric, indexed with unigrams and UMLS CUIs. Queries were composed manually by a clinician. Explicit mention of diagnosis in query was allowed for "treatment" and "test" questions.

myrun¶

Run ID: myrun
Participant: IKMLAB
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: b2b812044c963986095e1edce8a648b9
Run description: We use tfidf to count every documents score with topic.

NOVASEARCH1¶

Run ID: NOVASEARCH1
Participant: NovaSearch
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 6f987373e17c8a6b7a2ed6aa13e2aa0a
Run description: Indexing with Lucene. BM25L retrieval function, query expansion with MeSH and pseudo-relevance feedback with top results. External data: MeSH thesaurus for query expansion

NOVASEARCH2¶

Run ID: NOVASEARCH2
Participant: NovaSearch
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 77cf7b5f814b0c263788dce16865b65d
Run description: Indexing with Lucene. BM25L retrieval function, query expansion with MeSH and pseudo-relevance feedback with top results. Article Weighting for question type based on MeSH term count by category. External data: MeSH thesaurus for query expansion and weighting

NOVASEARCH3¶

Run ID: NOVASEARCH3
Participant: NovaSearch
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: c9db4b6e8800ad6d690ae1801fcb2c5e
Run description: Late fusion of NOVASEARCH1 and NOVASEARCH2 runs using Reciprocal Rank Fusion (RRF). External data: MeSH thesaurus for query expansion and weighting

NOVASEARCH4¶

Run ID: NOVASEARCH4
Participant: NovaSearch
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 4797596bce52631be3b6f4032e7bcba4
Run description: Combination of BM25L, BM25+, Language Models, and TFIDF using RRF. Individual runs with query expansion with MeSH and pseudo-relevance feedback with top results and article Weighting for question type based on MeSH term count by category. External data: MeSH thesaurus for query expansion and weighting.

NOVASEARCH5¶

Run ID: NOVASEARCH5
Participant: NovaSearch
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 0c6ec72df33c40f0bc42a2d28ec0d215
Run description: Indexing with Lucene. BM25L retrieval function, query expansion with MeSH and pseudo-relevance feedback with top results from journals with the highest impact factor. Article Weighting for question type based on MeSH term count by category. External data: MeSH thesaurus for query expansion and weighting and JCR Impact Factors List 2013 for pseudo relevance feedback

ohsuAbstDef¶

Run ID: ohsuAbstDef
Participant: OHSU
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: manual
Task: main
MD5: 681d5b25295332390c44700a8ef6a0c0
Run description: Manually constructed queries using our search interface, run only over the abstracts of the articles.

ohsuBodyDef¶

Run ID: ohsuBodyDef
Participant: OHSU
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: manual
Task: main
MD5: 2954885cc5357576b2f1ebb204a44c5d
Run description: Manually constructed queries using our search interface, run over the full text of the articles.

ohsuOrigAbst¶

Run ID: ohsuOrigAbst
Participant: OHSU
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: manual
Task: main
MD5: 9b63c63abbab319a7eb4d9f097cf4321
Run description: First attempt at queries, with no knowledge of article body text, run only on abstracts.

ohsuOrigBody¶

Run ID: ohsuOrigBody
Participant: OHSU
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: manual
Task: main
MD5: 597afb71397e152260d348d2b2740eb7
Run description: First attempt at queries, with no knowledge of article body text, run only on body text. Does blindly indexing article full text help?

origexp¶

Run ID: origexp
Participant: cuhk_sls
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: a8ff2bd56b1b77189dae8f0b7e213f61
Run description: The three types of cases and their synonyms were added to each query. Synonyms were obtained from WordNet. External resources are Terrier and WordNet.

prise1¶

Run ID: prise1
Participant: super_kxlab
Track: Clinical Decision Support
Year: 2014
Submission: 7/15/2014
Type: automatic
Task: main
MD5: bdba59670206af2aa0deaae15c9b1ea5
Run description: step 1. preprocess. step 2. extrat keywords. step 3. compute tf-idf. step 4. compute cosine similarity. step 5. rank. The third column "docno" is the number we found in the article under the tag "pmcid".

prna1¶

Run ID: prna1
Participant: Philips
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: 1a293c276f64b516cc1690b7daac88d1
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 topical keywords, which were used to extract clinical concepts from relevant Wiki articles. The Wiki concepts were used in mapping pertinent biomedical articles, which were further filtered by named entity information, and ordered by publication date and importance in relation to the extracted Wiki keywords.

Run1BoWC¶

Run ID: Run1BoWC
Participant: LIMSI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 330c9eaa27750f3057463050d20e41b2
Run description: Baseline run: 1. A manually-constructed MeSH query for each dimension (diagnosis/test/treatment) was sent to Pubmed. Based on the results we constructed 3 different subcorpora from the given corpus. 2. We then performed a plain-text-based retrieval based on BM25 in Terrier for each topic in its relevant subcorpus.

Run2MeSHDi¶

Run ID: Run2MeSHDi
Participant: LIMSI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 31ccb1a3e31bcf60efe94c05bbbf79f7
Run description: 1. We build a symptom-checker type system based on disease-symptom association information extracted from OrphaNet and Disease Symptom Knowledge Database. 2. For each topic we retrieved 5 disease hypotheses which were converted to their corresponding MeSH tems. 3. For each topic we queried PubMed with a query constructed of the MeSH terms of the retrieved diseases and generic (manually-constructed) query for the relevant dimension (diagnosis/test/treatment).

Run3MeSHDiCa¶

Run ID: Run3MeSHDiCa
Participant: LIMSI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 5d845a389941ffafc4319db6905195a3
Run description: 1. We build a symptom-checker type system based on disease-symptom association information extracted from OrphaNet and Disease Symptom Knowledge Database. 2. For each topic we retrieved 5 disease hypotheses which were converted to their corresponding MeSH tems. 3. For each topic we queried PubMed with a query constructed of the MeSH terms of the retrieved diseases, MeSH terms extracted from the topics (with MetaMap and Restrict-to-Mesh) and generic (manually-constructed) query for the relevant dimension (diagnosis/test/treatment).

Run4BoWDiCa¶

Run ID: Run4BoWDiCa
Participant: LIMSI
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: 7b76afb684f7d8570c367cf767733abb
Run description: 1. We build a symptom-checker type system based on disease-symptom association information extracted from OrphaNet and Disease Symptom Knowledge Database. 2. For each topic we retrieved 5 disease hypotheses whose raw-text name variants (extracted from UMLS, OrphaNet and Disease Symptom Knowledge Database) were combined with the topic summary to form bag-of-words queries. 3. We performed plain-text-based retrieval using Terrier (BM25 scoring).

Run5BoWDiCaS¶

Run ID: Run5BoWDiCaS
Participant: LIMSI
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: b3a1ec1b4ba0f67772f6e8178d591e84
Run description: 1. We build a symptom-checker type system based on disease-symptom association information extracted from OrphaNet and Disease Symptom Knowledge Database. 2. For each topic we retrieved 5 disease hypotheses whose raw-text name variants (extracted from UMLS, OrphaNet and Disease Symptom Knowledge Database) as well as their symptoms (also name variants) were combined with the topic summary to form bag-of-words queries. 3. We performed plain-text-based retrieval using Terrier (BM25 scoring).

runSystem2¶

Run ID: runSystem2
Participant: ir.cs.sfsu
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 674eed97cb4f0d3ffdc607f3b613f823
Run description: The run takes in a given description and uses MetaMap API to categorize the semantic types. Various semantic types are given different weight. Some semantic types are also expanded through the use of MESH. The query is generated through the use of Indri Retrieval Model. For the type treatment and test, various treatment names and test names have been gathered through webMD, which is used for the post processing, including removal and adjustment of scores retrieved.

samgyupsal¶

Run ID: samgyupsal
Participant: BigPig
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: ef019168df9a06589173fe5383864e87
Run description: metamap, luence-bm25similarity

SNUMedinfo1¶

Run ID: SNUMedinfo1
Participant: SNUMedinfo
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: automatic
Task: main
MD5: bc7cf01aa9dd83c3a909be8663346149
Run description: Query expansion MeSH MEDLINE Task classification SVM Clinical Hedges Database

SNUMedinfo2¶

Run ID: SNUMedinfo2
Participant: SNUMedinfo
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: automatic
Task: main
MD5: e758660b9cb605a0130f9e230707dfa8
Run description: Query expansion MeSH MEDLINE Task classification SVM Clinical Hedges Database

SNUMedinfo3¶

Run ID: SNUMedinfo3
Participant: SNUMedinfo
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: automatic
Task: main
MD5: 186af407b7b9c46b66f8bf2695160b4a
Run description: Query expansion MeSH MEDLINE Task classification SVM Clinical Hedges Database

SNUMedinfo4¶

Run ID: SNUMedinfo4
Participant: SNUMedinfo
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: automatic
Task: main
MD5: 5afe2708cf4bd85e451a8615bcc8446d
Run description: Query expansion MeSH MEDLINE Task classification SVM Clinical Hedges Database

SNUMedinfo6¶

Run ID: SNUMedinfo6
Participant: SNUMedinfo
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: automatic
Task: main
MD5: 1477543594f22043e769517e0d8f7658
Run description: Query expansion MeSH MEDLINE Task classification SVM Clinical Hedges Database

summary50ex¶

Run ID: summary50ex
Participant: CSEIITV
Track: Clinical Decision Support
Year: 2014
Submission: 7/27/2014
Type: automatic
Task: main
MD5: 4a690c27ca9ebf1622eafc6b1c7702f4
Run description: Summaries were used in this run. Only 50 documents were ranked in this run. For initial pruning of documents we used Lucene.

tudorComb1¶

Run ID: tudorComb1
Participant: HENRI_TUDOR_LUX
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: b00bcf3232fac727a51c6c9c4d38aaed
Run description: + ignore low idf terms + query terms in description field + combination of three state-of-the-art weighting model namely BM25, LGD and In_expB2 + query expansion using 30 expanded terms within top-20 documents.

tudorComb2¶

Run ID: tudorComb2
Participant: HENRI_TUDOR_LUX
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: 5ffcaf8c48aec1e926afee4bf90369ee
Run description: + ignore low idf terms + query terms in description field + combination of three state-of-the-art weighting model namely BM25, LGD and In_expB2 + query expansion using 30 expanded terms within top-20 documents.

tudorComb3¶

Run ID: tudorComb3
Participant: HENRI_TUDOR_LUX
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: 27ee3d7aa2a630e025dcc759d4b8d4a6
Run description: + consider low idf terms + query terms in description field + combination of three state-of-the-art weighting model namely BM25, LGD and In_expB2 + query expansion using 30 expanded terms within top-20 documents.

tudorComb4¶

Run ID: tudorComb4
Participant: HENRI_TUDOR_LUX
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: automatic
Task: main
MD5: 8815aa4e8b48650ca54e3b328c96ee1f
Run description: + consider low idf terms + query terms in summary field + combination of three state-of-the-art weighting model namely BM25, LGD and In_expB2 + query expansion using 30 expanded terms within top-20 documents.

tudorCombm¶

Run ID: tudorCombm
Participant: HENRI_TUDOR_LUX
Track: Clinical Decision Support
Year: 2014
Submission: 7/29/2014
Type: manual
Task: main
MD5: 76fd38064e37b51574c4de5a511e79c2
Run description: + ignore low idf terms + query terms in descrption field + combination of three state-of-the-art weighting model namely BM25, LGD and In_expB2 + query expansion using 30 expanded terms within top-20 documents. + query expansion with long form of abbrevitations (manual) + manual query weighting: give a more important weight for medical query terms

TUW1¶

Run ID: TUW1
Participant: TUW
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: ccc0723e5e567b2c62e71b80603bec2a
Run description: In this run, Indri is used as the base search engine system. We tested different combinations of query formulation and indexes (18 at all). Metamao and Freebase were used in some variations.

TUW2¶

Run ID: TUW2
Participant: TUW
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 2c9eb4d5b3e71d6f86e9f04906849385
Run description: In this run, Lucene is used as the base search engine system. We tested different combinations of query formulation and indexes (18 at all). Metamap and Freebase were used in some variations.

TUW3¶

Run ID: TUW3
Participant: TUW
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: cadbe5e48836dcf02a7a78a729e3460e
Run description: In this run, Xapian is used as the base search engine system. We tested different combinations of query formulation and indexes (18 at all). Metamap and Freebase were used in some variations.

TUW4¶

Run ID: TUW4
Participant: TUW
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 3a2d0b49000b99b80d4b623719fe5315
Run description: In this run, the previous 3 runs were combined. We tested different combinations of query formulation and indexes (18 for each). Metamap and Freebase were used in some variations.

TUW5¶

Run ID: TUW5
Participant: TUW
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: dd2da902fa6307fa428e6ea924e04f60
Run description: In this run, we experiment with word2vec. No external resource are used.

UDInfoCDS1¶

Run ID: UDInfoCDS1
Participant: udel_fang
Track: Clinical Decision Support
Year: 2014
Submission: 7/17/2014
Type: automatic
Task: main
MD5: a73e0665d4a61adad592ed7ae007b52c
Run description: This run is a concept based run. The case databases website is also used to generate the query expansion. External resources: MetaMap, Case Databases.

UDInfoCDS2¶

Run ID: UDInfoCDS2
Participant: udel_fang
Track: Clinical Decision Support
Year: 2014
Submission: 7/17/2014
Type: automatic
Task: main
MD5: ed4fdd62462fc58a7248eb65c67752f9
Run description: This run is a concept based run. The case databases website and UMLS are used to generate the query expansion. External resources: MetaMap, Case Databases, UMLS.

UDInfoCDS3¶

Run ID: UDInfoCDS3
Participant: udel_fang
Track: Clinical Decision Support
Year: 2014
Submission: 7/17/2014
Type: automatic
Task: main
MD5: 0b0ab55884e381f8f56cd06e79ca5d5b
Run description: This run is a term based run. The case databases website is used to generate the query expansion. External resources: Case Databases.

UDInfoCDS4¶

Run ID: UDInfoCDS4
Participant: udel_fang
Track: Clinical Decision Support
Year: 2014
Submission: 7/17/2014
Type: automatic
Task: main
MD5: 36558f792b76c5c9d28f3b600f5c8474
Run description: This run is a concept based run. The UMLS database is used to generate the query expansion. External resources: UMLS.

UDInfoCDS5¶

Run ID: UDInfoCDS5
Participant: udel_fang
Track: Clinical Decision Support
Year: 2014
Submission: 7/17/2014
Type: automatic
Task: main
MD5: 9657efbda86c4e76308c77f9e8c48f1a
Run description: This run is a term based run. Only the summary query is used to generate the results. No external resource is used.

UTD0BL¶

Run ID: UTD0BL
Participant: UTDHLTRI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: f4ea6256d289b655efb9b7e6494a63e6
Run description: Baseline run, BM25 retrieval model using Wikipedia-inspired keywords.

UTD1QE¶

Run ID: UTD1QE
Participant: UTDHLTRI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 512f705fc45491c8077d6a040e4535c9
Run description: BM25 retrieval model incorporating Wikipedia, UMLS, and SNOMED information for Query Expansion.

UTD2LDA¶

Run ID: UTD2LDA
Participant: UTDHLTRI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: d9f60a8c1e119a7c8704e36424dc42ed
Run description: Matrix-based cosine similarity on LDA(k=100)-mapped topics and documents.

UTD3W2VE¶

Run ID: UTD3W2VE
Participant: UTDHLTRI
Track: Clinical Decision Support
Year: 2014
Submission: 7/28/2014
Type: automatic
Task: main
MD5: 1cd10ae0006063f9c56e15efeb573c65
Run description: BM25-based retrieval model using Google's Word2Vec topic modelling technique for query expansion.