Runs - Genomics 2007¶

AIDrun1¶

Run ID: AIDrun1
Participant: uamsterdam.meij
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: automatic
Run description: baseline

AIDrun2¶

Run ID: AIDrun2
Participant: uamsterdam.meij
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: automatic
Run description: Same as baseline (AIDrun1), but with more elaborate passage identification.

AIDrun3¶

Run ID: AIDrun3
Participant: uamsterdam.meij
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: automatic
Run description: baseline, re-ranked according to a language model of the entity-type assiciated with the topic.

asubaral1¶

Run ID: asubaral1
Participant: arizona-stateu.gonzalez
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: automatic
Run description: - finding relatedness between words in the passages and keywords in the corresponding question - keyword expansion by utilizing the terms appearing in the definitions of the keywords of the questions

asubaral2¶

Run ID: asubaral2
Participant: arizona-stateu.gonzalez
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: - using both Lucene and Indri indexing systems for retrieval - passage length is as minimal as possible - finding relatedness between words in the passages and keywords in the corresponding question - keyword expansion by utilizing the terms appearing in the definitions of the keywords of the questions

asubaral3¶

Run ID: asubaral3
Participant: arizona-stateu.gonzalez
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: interactive
Run description: Similar to our first run, except we interactively modified queries to improve the answers

biokiP¶

Run ID: biokiP
Participant: concordiau.bergler
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: interactive
Run description: Interactive selection of keyphrases and weights. Result span expanded to paragraph.

biokiS¶

Run ID: biokiS
Participant: concordiau.bergler
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: interactive
Run description: Interactive selection of keyphrases and weights. Result span expanded to sentence boundaries.

biokiST¶

Run ID: biokiST
Participant: concordiau.bergler
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: interactive
Run description: Interactive selection of keyphrases and weights. Entity-type matching on some queries. Result span expanded to sentence boundaries.

DUTgen1¶

Run ID: DUTgen1
Participant: dalianu.yang
Track: Genomics
Year: 2007
Submission: 7/14/2007
Type: interactive
Run description: Indri; named entity recognition; sentence-leval overlapped window; query expansion based on MeSH;post-processing using templates

DUTgen2¶

Run ID: DUTgen2
Participant: dalianu.yang
Track: Genomics
Year: 2007
Submission: 7/14/2007
Type: interactive
Run description: Indri; named entity recognition; sentence-leval overlapped window; query expansion based on MeSH;result combination

DUTgen3¶

Run ID: DUTgen3
Participant: dalianu.yang
Track: Genomics
Year: 2007
Submission: 7/14/2007
Type: automatic
Run description: BM25; named entity recognition; NP extraction for topics; sentence-leval overlapped window; query expansion based on MeSH; result combination;

EBI1Lucene¶

Run ID: EBI1Lucene
Participant: ebi.yepes
Track: Genomics
Year: 2007
Submission: 7/14/2007
Type: automatic
Run description: System based on a Lucene index that considers the spans as documents. The Lucene scoring function has been modified to better deal with large and small documents based on the article by Singhal about pivoted cosine normalization. A postprocessing of the spans has been done removing HTML without any content and by finding the zone with relevant information based on the similarity of the query and the sentences in the spans.

EBI2Fusion¶

Run ID: EBI2Fusion
Participant: ebi.yepes
Track: Genomics
Year: 2007
Submission: 7/14/2007
Type: automatic
Run description: This run is the fusion of two configurations of our system. Our system is based on a Lucene index that considers the spans as documents (configuration 1); in addition, query expansion and boosting of some spans based on the entities matched between the query and the spans can be done (configuration 2). The Lucene scoring function has been modified to better deal with large and small documents based on the article by Singhal about pivoted cosine normalization. In addition, the spans have been processed by removing HTML without any content and by finding the zone with relevant information based on the similarity of the query and the sentences in the spans.

EBI3Boosting¶

Run ID: EBI3Boosting
Participant: ebi.yepes
Track: Genomics
Year: 2007
Submission: 7/14/2007
Type: automatic
Run description: System based on a Lucene index that considers the spans as documents. The Lucene scoring function has been modified to better deal with large and small documents based on the article by Singhal about pivoted cosine normalization. A postprocessing of the spans has been done analyzing the entities in the query and in the span and boosting the spans based on the entities that are matched. In addition, the spans have been processed by removing HTML without any content and by finding the zone with relevant information based on the similarity of the query and the sentences in the spans.

fdgerun1¶

Run ID: fdgerun1
Participant: fudanu.niu
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: automatic
Run description: Automatically extract the relevant concepts of each topic, retrieval sentences according to those concepts.

fdgerun2¶

Run ID: fdgerun2
Participant: fudanu.niu
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: automatic
Run description: Automatically extract relevant concepts of each topic, combine the result of sentence retrieval and the one of context retrieval.

fdgerun3¶

Run ID: fdgerun3
Participant: fudanu.niu
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: manual
Run description: Score each sentence according to the concurrency of concept terms from different groups.

GenTeaBB¶

Run ID: GenTeaBB
Participant: uhosp-geneva.ruch
Track: Genomics
Year: 2007
Submission: 7/13/2007
Type: automatic
Run description: Same as GenTeam1 but with Boolean boosting.

GenTeam1¶

Run ID: GenTeam1
Participant: uhosp-geneva.ruch
Track: Genomics
Year: 2007
Submission: 7/13/2007
Type: automatic
Run description: Basic run using easyIR as IR engine (dtu.dtn, Porter).

GenTeaPA¶

Run ID: GenTeaPA
Participant: uhosp-geneva.ruch
Track: Genomics
Year: 2007
Submission: 7/14/2007
Type: automatic
Run description: Same run as GenTeaBB, but with passage selection based on assesing the density of semantic targets.

HFmanual¶

Results | Participants | Input | Summary | Appendix

Run ID: HFmanual
Participant: umd-collegepark.oard
Track: Genomics
Year: 2007
Submission: 7/13/2007
Type: manual
Run description: Hongfang's run

hltcairo1¶

Results | Participants | Input | Summary | Appendix

Run ID: hltcairo1
Participant: ibm.cairo
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: First 25 results from the search engine

hltcairo2¶

Results | Participants | Input | Summary | Appendix

Run ID: hltcairo2
Participant: ibm.cairo
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: First 50 results from the search engine

icbdoc¶

Results | Participants | Input | Summary | Appendix

Run ID: icbdoc
Participant: weill-med-cornellu
Track: Genomics
Year: 2007
Submission: 7/13/2007
Type: manual
Run description: Rank fusion of seven different search techniques implemented in Twease (includes both automatic and manual runs). Optimized for document MAP.

icbpassage¶

Results | Participants | Input | Summary | Appendix

Run ID: icbpassage
Participant: weill-med-cornellu
Track: Genomics
Year: 2007
Submission: 7/13/2007
Type: manual
Run description: Rank fusion of seven different search techniques implemented in Twease (includes both automatic and manual runs). Passages are marked as a post-processing step after document retrieval and fusion. At most, ten passages are included per document. Optimized for passage MAP.

icbtwease¶

Results | Participants | Input | Summary | Appendix

Run ID: icbtwease
Participant: weill-med-cornellu
Track: Genomics
Year: 2007
Submission: 7/13/2007
Type: manual
Run description: Manual run with minimal interval semantics performed with Twease using a slider value of 80. This run was performed with the same Twease software version deployed at Twease.org as of July 2007.

iitx1r2¶

Run ID: iitx1r2
Participant: iit.urbain
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: MST passage extraction by concept sc = ( 1.0result.getPassConceptSCNorm() + 0.1result.getSentConceptSCNorm() + 1.0result.getPassConceptIdfSumNorm() + 0.1result.getSentConceptIdfSumNorm())

iitx2r2¶

Run ID: iitx2r2
Participant: iit.urbain
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: MST passage extraction by concept sc = ( 1.0result.getPassConceptSCNorm() + 0.1result.getSentConceptSCNorm() + 1.0result.getPassConceptIdfSumNorm() + 0.1result.getSentConceptIdfSumNorm()) With sentence boosting dependency grammar, sumidf, nconcepts

iitx3r2¶

Run ID: iitx3r2
Participant: iit.urbain
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: MST passage extraction by concept sc =((passConceptSCNorm+sentConceptSCNorm+passConceptIdfSumNorm+sentConceptIdfSumNorm)/4)

IRn¶

Run ID: IRn
Participant: ualicante.terol
Track: Genomics
Year: 2007
Submission: 7/17/2007
Type: automatic
Run description: This run has been performed by applying the Information Retrieval technique based on passages. The passages are composed of four sentences. The indexing of the document collection applies the Okapi measure.

kyoto1¶

Run ID: kyoto1
Participant: kyotou.wan
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: automatic
Run description: Paragraph-level impact-based retrieval combined with a probabilistic model for term co-occurrence. Passages scored using a variant of TF/IDF, but results are ranked using only the IR system's scores.

kyoto2¶

Run ID: kyoto2
Participant: kyotou.wan
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: Paragraph-level impact-based retrieval combined with a probabilistic model for term co-occurrence. Passages scored using a variant of TF/IDF, but more results were used and then ranked using only the PM's scores.

kyoto3¶

Run ID: kyoto3
Participant: kyotou.wan
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: Paragraph-level impact-based retrieval combined with a probabilistic model for term co-occurrence. Passages scored using only the probabilistic model and final ranking determined using equal weight on both systems.

LHNCBC¶

Run ID: LHNCBC
Participant: nlm.aronson
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: automatic
Run description: An automatic run based on LHC's Essie search engine and for which results are reranked based on relationships extracted from Essie results using MetaMap and SemRep.

MuMshFd¶

Run ID: MuMshFd
Participant: umelbourne.stokes
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: Automatic query expansion with entities and ontological terms, but without passage reduction and reranking.

MuMshFdRsc¶

Run ID: MuMshFdRsc
Participant: umelbourne.stokes
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: Automatic query expansion with entities and ontological terms, followed by passage reduction and reranking.

MuMshNfdRsc¶

Run ID: MuMshNfdRsc
Participant: umelbourne.stokes
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: Automatic query expansion with ontological terms only, followed by passage reduction and reranking.

ncbi2007a¶

Results | Participants | Input | Summary | Appendix

Run ID: ncbi2007a
Participant: umd-collegepark.oard
Track: Genomics
Year: 2007
Submission: 7/13/2007
Type: automatic
Run description: Reranked Essie hits from NCBI

ncbi2007b¶

Results | Participants | Input | Summary | Appendix

Run ID: ncbi2007b
Participant: umd-collegepark.oard
Track: Genomics
Year: 2007
Submission: 7/13/2007
Type: automatic
Run description: run generated by Larry

NLMfusion¶

Run ID: NLMfusion
Participant: nlm.aronson
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: automatic
Run description: An automatic run obtained by applying fusion to the LHNCBC run, a Terrier run, an NCBI Themes run, an INDRI run and an easyIR run.

NLMinter¶

Run ID: NLMinter
Participant: nlm.aronson
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: interactive
Run description: An interactive run based on an interactively created filter applied to the NLMfusion run.

OHSUQA¶

Run ID: OHSUQA
Participant: ohsu.hersh
Track: Genomics
Year: 2007
Submission: 7/13/2007
Type: automatic
Run description: Two stage query generation with MESH and gene synonym expansion, and entity-specific keywords. Lucene maximal passage index, TF*IDF. MMTX based sentence entity count passage trimming.

OHSUQASUB¶

Run ID: OHSUQASUB
Participant: ohsu.hersh
Track: Genomics
Year: 2007
Submission: 7/13/2007
Type: automatic
Run description: Two stage query generation with MESH and substances expansion, and entity-specific keywords. Lucene maximal passage index, TF*IDF. MMTX based sentence entity count passage trimming.

OHSUQASUBEX¶

Run ID: OHSUQASUBEX
Participant: ohsu.hersh
Track: Genomics
Year: 2007
Submission: 7/13/2007
Type: automatic
Run description: Two stage query generation with MESH and substances expansion, and entity-specific keywords. Lucene maximal passage index, TF*IDF. MMTX based sentence entity count passage trimming.

TsingHua3¶

Results | Participants | Input | Summary | Appendix

Run ID: TsingHua3
Participant: tsinghuau.zhang
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: (run3)Machine learning and dictionary based NE recognition, BM2500, Treble passage retrieval.

TsingHua4¶

Results | Participants | Input | Summary | Appendix

Run ID: TsingHua4
Participant: tsinghuau.zhang
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: (run4) Machine learning and dictionary based NE, sigma local df for weighting, Treble passage retrieval

TsingHua5¶

Results | Participants | Input | Summary | Appendix

Run ID: TsingHua5
Participant: tsinghuau.zhang
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: (run5_new) strictly generated dictionary for NE, max local df for weighting, Treble passage retrieval, reduction

UBexp1¶

Results | Participants | Input | Summary | Appendix

Run ID: UBexp1
Participant: ubuffalo.ruiz
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: automatic
Run description: Automatic runs generated with Indri. Queries were build automatically by expanding them with MetaMap and discarding common terms. Gene and proteins names were expanded using Gene Ontology. Query formulated using synonyms to represent the expanded terms and multiword phrases when appropriate. Reference sections were discarded by restricting results to those passages that did not have the word "Medline".

UBHFmanual¶

Results | Participants | Input | Summary | Appendix

Run ID: UBHFmanual
Participant: ubuffalo.ruiz
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: manual
Run description: Queries were expanded using publicly available resources. This list was manually filtered to discard ambiguous names of gene and proteins.

uchsc1¶

Results | Participants | Input | Summary | Appendix

Run ID: uchsc1
Participant: ucolorado.cohen
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: manual
Run description: Queries were manually expanded and individual terms were assigned weights. Lists of terms matching keyword classes were included in the queries; those terms recieved equal weights. The queries were submitted to the Indri search engine of the Lemur toolkit. Post-processing included filtering out passages that did not contain genes, mutations or biological substances, according to query type.

uchsc2¶

Results | Participants | Input | Summary | Appendix

Run ID: uchsc2
Participant: ucolorado.cohen
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: interactive
Run description: Queries were manually expanded and individual terms were assigned weights based on MeSH distance. Additionally, salient biomedical predicates were also expanded for 5 of the queries. Lists of terms matching keyword classes were included in the queries; those terms recieved equal weights. The queries were submitted to the Indri search engine of the Lemur toolkit. Post-processing included filtering out passages that did not contain genes, mutations or biological substances, according to query type.

UICGenRun1¶

Run ID: UICGenRun1
Participant: uillinois.chicago.zhou
Track: Genomics
Year: 2007
Submission: 7/15/2007
Type: automatic
Run description: Utilize UMLS to get some of the entities.

UICGenRun2¶

Run ID: UICGenRun2
Participant: uillinois.chicago.zhou
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: Do not differentiate the importance of entities in passages as long as some entity presents in passages.

UIowa07Gen01¶

Results | Participants | Input | Summary | Appendix

Run ID: UIowa07Gen01
Participant: uiowa.eichman
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: manual
Run description: title of reference identified from logical document structure

UIUCconj¶

Run ID: UIUCconj
Participant: uc-zhai
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: automatic run

UIUCrelfb¶

Run ID: UIUCrelfb
Participant: uc-zhai
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: interactive
Run description: Interactive run with relevance feedback

UIUCsyn¶

Run ID: UIUCsyn
Participant: uc-zhai
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: automatic run with synonym expansion

UniNE1¶

Run ID: UniNE1
Participant: uneuchatel.savoy
Track: Genomics
Year: 2007
Submission: 7/9/2007
Type: automatic
Run description: Retrieval based on Divergence from randomness. Query expansion using forms generated from query words. The length of a passage is delimited by the
tag.

UniNE2¶

Run ID: UniNE2
Participant: uneuchatel.savoy
Track: Genomics
Year: 2007
Submission: 7/9/2007
Type: automatic
Run description: Data fusion of three IR systems. 1 Retrieval based on Okapi model with query expansion using forms generated from query words. 2 Retrieval based on Okapi model, using only the original query words. Re-ranking based on distance between query words and entity in the query. 3 Retrieval based on Divergence from randomness. Query expansion using forms generated from query words. Each passage is a sentence.

UniNE3¶

Run ID: UniNE3
Participant: uneuchatel.savoy
Track: Genomics
Year: 2007
Submission: 7/9/2007
Type: automatic
Run description: Data fusion of three IR systems 1- Retrieval based on Divergence from randomness. Query expansion using forms generated from query words and word variant generation for entity and query terms. 2 Retrieval based on Okapi model with query expansion using forms generated from query words. 3 Retrieval based on Divergence from randomness. Query expansion using forms generated from query words. Re-ranking based on distance between query words and entity in the query. Each passage is delimited by the
tag.

UTEMC1¶

Run ID: UTEMC1
Participant: tno-ict
Track: Genomics
Year: 2007
Submission: 7/9/2007
Type: automatic
Run description: UMLS-based thesaurus in combination with language-modeling. Run optimized for aspect-retrieval.

UTEMC2¶

Run ID: UTEMC2
Participant: tno-ict
Track: Genomics
Year: 2007
Submission: 7/9/2007
Type: automatic
Run description: UMLS-based thesaurus in combination with language-modeling. Run optimized for precision.

york07ga1¶

Run ID: york07ga1
Participant: yorku.huang
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: No query expansion. Use only terms extracted from the raw topics for retrieval. Use BM25 for term weighting in structured queries. Use Okapi to build word-based index.

york07ga2¶

Run ID: york07ga2
Participant: yorku.huang
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: Expand query terms for 11 gene-related topics by using Entrez Gene. Use BM25 for term weighting in structured queries. Use Okapi to build sentence-based index.

york07ga3¶

Run ID: york07ga3
Participant: yorku.huang
Track: Genomics
Year: 2007
Submission: 7/16/2007
Type: automatic
Run description: Expand query terms for all the topics by using UMLS. Use BM25 for term weighting in structured queries. Use Okapi to build word-based index.