Runs - Genomics 2004¶

akoike¶

Results | Participants | Input | Summary | Appendix

Run ID: akoike
Participant: u.tokyo
Track: Genomics
Year: 2004
Submission: 7/16/2004
Type: manual
Task: adhoc
Run description: Strong heuristic

akoyama¶

Results | Participants | Input | Summary | Appendix

Run ID: akoyama
Participant: u.tokyo
Track: Genomics
Year: 2004
Submission: 7/16/2004
Type: manual
Task: adhoc
Run description: Initial result screening

aliasiBase¶

Run ID: aliasiBase
Participant: alias-i
Track: Genomics
Year: 2004
Submission: 7/7/2004
Type: automatic
Task: adhoc
Run description: SYSTEM Jakarta Lucene 1.3 (jakarta.apache.org/lucene) LINGUISTICS Alias-i LingPipe (www.aliasi.com/lingpipe) INDEXING Tokenized with LingPipe IndoEuropeanTokenizer, lowercased, without a stop list. Combined document text into single field with boosts 4title, 1abstract, 2ordinary MESH, 4starred MESH, 1chemical names. Took 5 hours, producing 9GB index, including storing original titles, abstracts, MESH terms and chemical names. SEARCH Tokenized with LingPipe IndoEuropeanTokenizer, lowercased, removed stop terms. Used fields with boosts 4title, 2need, 1context. Lucene's standard TF/IDF implementation TF is square root of raw frequency; IDF is log(num docs/doc freq + 1) + 1; doc and query normalized to unit length. Took 15 minutes per result set of 50 topics with 1000 results/topic.

aliasiTerms¶

Run ID: aliasiTerms
Participant: alias-i
Track: Genomics
Year: 2004
Submission: 7/7/2004
Type: automatic
Task: adhoc
Run description: SYSTEM Jakarta Lucene 1.3 (jakarta.apache.org/lucene) LINGUISTICS Alias-i LingPipe (www.aliasi.com/lingpipe) INDEXING Tokenized with LingPipe IndoEuropeanTokenizer, lowercased, without a stop list. Combined document text into single field with boosts 4title, 1abstract, 2ordinary MESH, 4starred MESH, 1chemical names. Took 5 hours, producing 9GB index, including storing original titles, abstracts, MESH terms and chemical names. SEARCH Tokenized with LingPipe IndoEuropeanTokenizer, lowercased, removed stop terms. Used fields with boosts 4title, 2need, 1context. Lucene's standard TF/IDF implementation TF is square root of raw frequency; IDF is log(num docs/doc freq + 1) + 1; doc and query normalized to unit length. Took 15 minutes per result set of 50 topics with 1000 results/topic. TERMS Added all terms recognized by LingPipe's GENIA-corpus trained Genomics named entity annotator as quoted search terms with a boost of 4 (multiplied by field boost).

biotext1trge¶

Results | Participants | Input | Summary | Appendix

Run ID: biotext1trge
Participant: u.cberkeley.hearst
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Threshold change

BIOTEXT21¶

Results | Participants | Input | Summary | Appendix

Run ID: BIOTEXT21
Participant: u.cberkeley.hearst
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: annhi
Run description: GO code extraction, MGI before 2003

BIOTEXT22¶

Results | Participants | Input | Summary | Appendix

Run ID: BIOTEXT22
Participant: u.cberkeley.hearst
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: annhi
Run description: GO code extraction

BIOTEXT23¶

Results | Participants | Input | Summary | Appendix

Run ID: BIOTEXT23
Participant: u.cberkeley.hearst
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: annhi
Run description: GO identification + classification (SVM) + human

BIOTEXT24¶

Results | Participants | Input | Summary | Appendix

Run ID: BIOTEXT24
Participant: u.cberkeley.hearst
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: annhi
Run description: GO identification + classification (SVM)

BIOTEXT25¶

Results | Participants | Input | Summary | Appendix

Run ID: BIOTEXT25
Participant: u.cberkeley.hearst
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: annhi
Run description: GO identification + classification (SVM); window of 50

biotext2trge¶

Results | Participants | Input | Summary | Appendix

Run ID: biotext2trge
Participant: u.cberkeley.hearst
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Threshold change

biotext3trge¶

Results | Participants | Input | Summary | Appendix

Run ID: biotext3trge
Participant: u.cberkeley.hearst
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Threshold change

biotext4trge¶

Results | Participants | Input | Summary | Appendix

Run ID: biotext4trge
Participant: u.cberkeley.hearst
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Threshold change

biotext5trge¶

Results | Participants | Input | Summary | Appendix

Run ID: biotext5trge
Participant: u.cberkeley.hearst
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Threshold change

BioTextAdHoc¶

Results | Participants | Input | Summary | Appendix

Run ID: BioTextAdHoc
Participant: u.cberkeley.hearst
Track: Genomics
Year: 2004
Submission: 7/16/2004
Type: automatic
Task: adhoc
Run description: Features include MeSH descriptors, gene mappings, and gene families.

ConversAuto¶

Run ID: ConversAuto
Participant: converspeech
Track: Genomics
Year: 2004
Submission: 7/14/2004
Type: automatic
Task: adhoc
Run description: Automatic Query terms and filter terms are constructed from the topics using automated term extraction. Query terms are combined to form queries using a federated biomedical and English-language ontology. Filter terms order and score the results of the searches, which are implemented using PubMed e-utilities.

ConversManu¶

Run ID: ConversManu
Participant: converspeech
Track: Genomics
Year: 2004
Submission: 7/14/2004
Type: manual
Task: adhoc
Run description: Manual Queries and filter terms generated automatically were modified in consultation with a biologist based on an understanding of the intended search for each topic. Three kinds of general modifications were made.

csusm¶

Run ID: csusm
Participant: u.sanmarcos
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: manual
Task: adhoc
Run description: Removed stop words from XML Medline files and topics file. Query constructed using title field, in some cases information from the need field. Word-based search, each term in the query is assigned an equivalent weight. If there is a match add term weight to totalweight. If totalweight is >= total# terms for query * .5 * weight then document is retrieved.

cuhkrun1¶

Results | Participants | Input | Summary | Appendix

Run ID: cuhkrun1
Participant: chinese.u.hongkong
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: annhi
Run description: SVM classification is used with term weighting scheme of tf-idf; and with cost ration automatically tuned.

cuhkrun2¶

Results | Participants | Input | Summary | Appendix

Run ID: cuhkrun2
Participant: chinese.u.hongkong
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: annhi
Run description: Target gene identification is conducted which SVM classification is used with term weight considering distance between the term and target gene. Also, with cost ration automatically tuned.

cuhkrun3¶

Results | Participants | Input | Summary | Appendix

Run ID: cuhkrun3
Participant: chinese.u.hongkong
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: annhi
Run description: Using SVM with term weighting tf-idf and with different set of features. Also, automatically tuning the cost ration.

DCUma¶

Run ID: DCUma
Participant: dubblincity.u
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: manual
Task: adhoc
Run description: We built a MeSH term index and it is the only index used by our system for this run. With the help of an expert, we manually build mesh queries from the topics.

DCUmatn1¶

Run ID: DCUmatn1
Participant: dubblincity.u
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: manual
Task: adhoc
Run description: Two indices were built for this run, one with MeSH descriptors and one with title/abstract terms. We used the title and need fields to automatically generate abstract/title query and produced a ranking with the BM25 algorithm. We then queried the MeSH index using MeSH queries generated manually from the topics with the help of an expert and obtained a second ranking. The two ranking were combined and only the top 1000 documents were retained for each topic.

dimacsAabsw1¶

Run ID: dimacsAabsw1
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: annhi
Run description: Title, subject and abstract of the article were used to represent document-gene pair. Text representation stemmed, logtf-idf, cosine normalization. A classifier, Bayesian Binary Regression (BBR) with Gaussian prior, was trained separately for each domain. The test results across domains were then combined into one file.

dimacsAg3mh¶

Run ID: dimacsAg3mh
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 8/29/2004
Type: automatic
Task: annhi
Run description: MeSH terms from MEDLINE abstracts were used to represent document-gene pairs. Each MeSH term was represented as a token. Text representation binary. A classifier, Bayesian Binary Regression (BBR) with Gaussian prior was trained separately for each domain. The test results across domains were then combined into one file.

dimacsAl3w¶

Run ID: dimacsAl3w
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 8/29/2004
Type: automatic
Task: annhi
Run description: Fulltext taken from journal articles' subject, title, abstract, and body fields were used to represent document-gene pairs. Text representation stemmed, logtf-idf and cosine normalization. A classifier, Bayesian Binary Regression (BBR) with Laplace prior was trained separately for each domain. The test results across domains were then combined into one file.

dimacsAp5w5¶

Run ID: dimacsAp5w5
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: annhi
Run description: Paragraphs containing gene symbol/description (taken from gtrain and from LocusLink) were used to represent document-gene pair. Text representation stemmed, logtf-idf, cosine normalization. Positive instances in the training were given weight 5. A classifier, Bayesian Binary Regression (BBR) with Gaussian prior, was trained separately for each domain. The test results across domains were then combined into one file.

dimacsAw20w5¶

Run ID: dimacsAw20w5
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: annhi
Run description: Join of windows of half-size 20 around gene symbol/description (taken from gtrain and from LocusLink) were used to represent document-gene pair. Text representation stemmed, logtf-idf, cosine normalization. Positive instances in the training were given weight 5. A classifier, Bayesian Binary Regression (BBR) with Gaussian prior, was trained separately for each domain. The test results across domains were then combined into one file.

dimacsTfl9d¶

Run ID: dimacsTfl9d
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: triage
Run description: Training and test set documents were prefiltered to obtain documents with MeSH term Mice. Prefiltered training set was used to construct a classifier. Article title, abstract and MEDLINE MeSH terms were used for document representation. Text representation stemmed, logtf-idf, cosine normalization. A classifier, Bayesian Binary Regression (BBR) with Laplace prior, was trained, and used to make predictions with maximum expected effectiveness threshold tuning.

dimacsTfl9w¶

Run ID: dimacsTfl9w
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: triage
Run description: Training and test set documents were prefiltered to obtain documents with MeSH term Mice. Prefiltered training set was used to construct a classifier. Fulltext obtained from journal articles' title, abstract and body fields were used for document representation. Text representation stemmed, logtf-idf, cosine normalization. A classifier, Bayesian Binary Regression (BBR) with Laplace prior, was trained, and used to make predictions with maximum expected effectiveness threshold tuning.

dimacsTl9md¶

Run ID: dimacsTl9md
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: triage
Run description: Article title, abstract and MEDLINE MeSH terms were used for document representation. Text representation stemmed, logtf-idf, cosine normalization. A classifier, Bayesian Binary Regression (BBR) with Laplace prior, was trained and used to make predictions with maximum expected effectiveness threshold tuning.

dimacsTl9mhg¶

Run ID: dimacsTl9mhg
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: triage
Run description: For each document, MeSH terms and GENBANK references were taken from the corresponding MEDLINE abstracts, and each MeSH term and organism from the corresponding GENBANK entry was represented as a token. Text representation binary. A classifier, Bayesian Binary Regression (BBR) with Laplace prior, was trained and used to make predictions with maximum expected effectiveness threshold tuning.

dimacsTl9w¶

Run ID: dimacsTl9w
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: triage
Run description: Fulltext obtained from journal articles' subject, title, abstract and body fields were used for document representation. Text representation stemmed, logtf-idf, cosine normalization. A classifier, Bayesian Binary Regression (BBR) with Laplace prior, was trained, and used to make predictions with maximum expected effectiveness threshold tuning.

edinauto2¶

Run ID: edinauto2
Participant: u.edinburgh.sinclair
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: Lucene retrieval engine; queries weighted according to Web counts and expanded with noun phrases and synonyms

edinauto5¶

Run ID: edinauto5
Participant: u.edinburgh.sinclair
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: Lucene retrieval engine; queries weighted according to Web counts and expanded with noun phrases and synonyms

edis2¶

Run ID: edis2
Participant: u.edinburgh.sinclair
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Mice relevance based on Discussion sections and MH and RN headings, with triage decision based on Discussion section only.

eint2¶

Run ID: eint2
Participant: u.edinburgh.sinclair
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Based on introduction and MH and RN headings only

EMCTNOT1¶

Run ID: EMCTNOT1
Participant: tno.kraaij
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Documents were indexed for three different ontologies using concept recognition (Collexis). MesH terms were also encoded using semantic groups. This resulted in 4 different feature representations for which 5 different classifiers were trained. The output of these classifiers was combined in a meta-classifier.

emet2¶

Run ID: emet2
Participant: u.edinburgh.sinclair
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Based on Methods sections and MH and RN headings only

epub2¶

Run ID: epub2
Participant: u.edinburgh.sinclair
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Based on title, abstract and MH and RN headings only

eres2¶

Run ID: eres2
Participant: u.edinburgh.sinclair
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Mice relevance based on Results sections and MH and RN headings, with triage decision based on Results section only.

geneteam1¶

Run ID: geneteam1
Participant: u.hospital.geneva
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Triage with bigrams only

geneteam2¶

Run ID: geneteam2
Participant: u.hospital.geneva
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Triage with bigrams and monograms

geneteam3¶

Run ID: geneteam3
Participant: u.hospital.geneva
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Triage with only monograms

geneteamA1¶

Run ID: geneteamA1
Participant: u.hospital.geneva
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: sentence selection V1 categorizer "categMatcher" threshold selection

geneteamA2¶

Run ID: geneteamA2
Participant: u.hospital.geneva
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: sentence selection V1 categorizer "GOMAP" threshold selection

geneteamA3¶

Run ID: geneteamA3
Participant: u.hospital.geneva
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: sentence selection V2 categorizer "categMatcher" threshold selection

geneteamA4¶

Run ID: geneteamA4
Participant: u.hospital.geneva
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: sentence selection V2 categorizer "GOMAP" threshold selection

geneteamA5¶

Run ID: geneteamA5
Participant: u.hospital.geneva
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: Use of naive bayes info gain feature selection sentence selection V2

GUCbase¶

Run ID: GUCbase
Participant: german.u.cairo
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: Baseline run all possible combinations generated

GUCir30¶

Run ID: GUCir30
Participant: german.u.cairo
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: Lemur Text segments with gene names only Threshold for + Okapi score 30 w/ Backup

GUCir50¶

Run ID: GUCir50
Participant: german.u.cairo
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: Lemur Text segments with gene names only Threshold for + Okapi score 50 w/ Backup

GUClin1260¶

Run ID: GUClin1260
Participant: german.u.cairo
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Linear SVM model top 1260 returned documents

GUClin1700¶

Run ID: GUClin1700
Participant: german.u.cairo
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Linear SVM model top 1700 returned documents

GUCply1260¶

Run ID: GUCply1260
Participant: german.u.cairo
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Polynomial SVM model top 1260 returned documents captions only

GUCply1700¶

Run ID: GUCply1700
Participant: german.u.cairo
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Polynomial SVM model top 1700 returned documents captions only

GUCsvm0¶

Run ID: GUCsvm0
Participant: german.u.cairo
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: Linear SVM Text segments with gene names only Threshold for + 0.0

GUCsvm5¶

Run ID: GUCsvm5
Participant: german.u.cairo
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: Linear SVM Text segments with gene names only Threshold for + -5.0

GUCwdply2000¶

Run ID: GUCwdply2000
Participant: german.u.cairo
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Polynomial SVM model top 2000 returned documents full documents

IBMIRLver1¶

Results | Participants | Input | Summary | Appendix

Run ID: IBMIRLver1
Participant: ibm.india
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: used bioannotator, followed by classifier

iowarun1¶

Run ID: iowarun1
Participant: u.iowa
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: Iowa team's first run 390 lines

iowarun2¶

Run ID: iowarun2
Participant: u.iowa
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: Iowa's team run 2 305 lines

iowarun3¶

Run ID: iowarun3
Participant: u.iowa
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: Iowa's team run 3 447 lines

iowarun4¶

Run ID: iowarun4
Participant: u.iowa
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: annhi
Run description: Used a retrieval approach based on citation sentences.

KoikeyaHi1¶

Results | Participants | Input | Summary | Appendix

Run ID: KoikeyaHi1
Participant: u.tokyo
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: manual
Task: annhi
Run description: Efficient structural analyses of sentences.

KoikeyaHiev1¶

Results | Participants | Input | Summary | Appendix

Run ID: KoikeyaHiev1
Participant: u.tokyo
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: manual
Task: annhiev
Run description: Efficient structural analyses of sentences.

KoikeyaTri1¶

Results | Participants | Input | Summary | Appendix

Run ID: KoikeyaTri1
Participant: u.tokyo
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: manual
Task: triage
Run description: Efficient structural analyses of sentences.

KoikeyaTri2¶

Results | Participants | Input | Summary | Appendix

Run ID: KoikeyaTri2
Participant: u.tokyo
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: manual
Task: triage
Run description: Efficient structural analyses of sentences. Extended version.

KoikeyaTri3¶

Results | Participants | Input | Summary | Appendix

Run ID: KoikeyaTri3
Participant: u.tokyo
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: manual
Task: triage
Run description: Efficient structural analyses of sentences. Extended and high-rank documents version.

lga1¶

Run ID: lga1
Participant: indiana.u.seki
Track: Genomics
Year: 2004
Submission: 7/16/2004
Type: automatic
Task: adhoc
Run description: query expansion and pseudo-relevance feedback

lga2¶

Run ID: lga2
Participant: indiana.u.seki
Track: Genomics
Year: 2004
Submission: 7/16/2004
Type: automatic
Task: adhoc
Run description: query expansion

lgcab1¶

Run ID: lgcab1
Participant: indiana.u.seki
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhiev
Run description: knn+svd

lgcab2¶

Run ID: lgcab2
Participant: indiana.u.seki
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhiev
Run description: knn

lgcad1¶

Run ID: lgcad1
Participant: indiana.u.seki
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: knn+svd

lgcad2¶

Run ID: lgcad2
Participant: indiana.u.seki
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: knn

lgct1¶

Run ID: lgct1
Participant: indiana.u.seki
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Naive Bayes Classifier Gene Filter Mesh Terms

lgct2¶

Run ID: lgct2
Participant: indiana.u.seki
Track: Genomics
Year: 2004
Submission: 8/31/2004
Type: automatic
Task: triage
Run description: Naive Bayes Classifier Gene Filter Mesh Terms

LHCUMDSE¶

Run ID: LHCUMDSE
Participant: nlm.umd.ul
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: Search engine (SE) queries were formulated by extracting information from the task queries gene names recognized by ABGene, the corresponding MeSH terms for the gene names, phrases/concepts identified by MetaMap, organisms extracted using the NCBI Taxonomy, and other useful words extracted based on their form and TFIDF score. The extracted information was translated into SE syntax and weighted, and the results were filtered to those identified as likely genetically related by Journal Descriptor Indexing (JDI).

MeijiHilG¶

Run ID: MeijiHilG
Participant: meiji.u
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: query expansion depended on context

NLMA1¶

Run ID: NLMA1
Participant: nlm.umd.ul
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: A machine learning method which is a variation of decision list learning. NLMA1 is obtained using the whole training set (T1=0.5, T2=50)

NLMA2¶

Run ID: NLMA2
Participant: nlm.umd.ul
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: annhi
Run description: A machine learning method which is a variation of decision list learning. NLMA2 is obtained with an additional bootstrapping step, for (PMID, gene) pairs with a summation that is larger than T3 (i.e., 100), add them into the training set, and redo the categorization (with a threshold T1=0.8, T2=80).

NLMT21¶

Run ID: NLMT21
Participant: nlm.umd.ul
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Using all training data, create theme in + with background + or - Query test set with theme Keep if score > 0

NLMT22¶

Run ID: NLMT22
Participant: nlm.umd.ul
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Split into three sets, JBC,JCB,PNAS Create theme for each set in + with background + or - Split test set by journal Query each set by its theme Combine sco > 0 for all 3 journals

NLMT2ADA¶

Run ID: NLMT2ADA
Participant: nlm.umd.ul
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: ADA machine learning algorithm using 3-fold cross validation on train

NLMT2BAYES¶

Run ID: NLMT2BAYES
Participant: nlm.umd.ul
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: BAYES machine learning algorithm using 3-fold cross validation on train

NLMT2SVM¶

Run ID: NLMT2SVM
Participant: nlm.umd.ul
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: SVM machine learning algorithm using 3-fold cross validation on train

NTU2v3N1¶

Run ID: NTU2v3N1
Participant: ntu.chen
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: triage
Run description: title + abstract + fig/tbl captions

NTU3v3N1¶

Run ID: NTU3v3N1
Participant: ntu.chen
Track: Genomics
Year: 2004
Submission: 8/29/2004
Type: automatic
Task: triage
Run description: atl + abs + fig/tbl captions + body(if no abs)

NTU3v3N1c2¶

Run ID: NTU3v3N1c2
Participant: ntu.chen
Track: Genomics
Year: 2004
Submission: 8/29/2004
Type: automatic
Task: triage
Run description: atl + abs + fig/tbl captions + body(if no abs) different parameters

NTU4v3N1416¶

Run ID: NTU4v3N1416
Participant: ntu.chen
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: atl + abs + fig/tbl captions + BODY (if no abs) or result/discussion/conclusion Sections in BODY (if has abs)

nusbird2004a¶

Run ID: nusbird2004a
Participant: mlg.nus
Track: Genomics
Year: 2004
Submission: 8/26/2004
Type: automatic
Task: triage
Run description: SVM

nusbird2004b¶

Run ID: nusbird2004b
Participant: mlg.nus
Track: Genomics
Year: 2004
Submission: 8/29/2004
Type: automatic
Task: triage
Run description: SVM

nusbird2004c¶

Run ID: nusbird2004c
Participant: mlg.nus
Track: Genomics
Year: 2004
Submission: 8/29/2004
Type: automatic
Task: triage
Run description: SVM, with additional positive training examples.

nusbird2004d¶

Run ID: nusbird2004d
Participant: mlg.nus
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: SVM

nusbird2004e¶

Run ID: nusbird2004e
Participant: mlg.nus
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: SVM

OHSUAll¶

Run ID: OHSUAll
Participant: ohsu.hersh
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: Automatic query generation, processed by Lucene using TF-IDF without normalization

OHSUNBAYES¶

Run ID: OHSUNBAYES
Participant: ohsu.hersh
Track: Genomics
Year: 2004
Submission: 8/24/2004
Type: automatic
Task: triage
Run description: CHI SQUARE 0.025 FULL FEATURE SET NAIVE BAYES CLASSIFIER

OHSUNeeds¶

Run ID: OHSUNeeds
Participant: ohsu.hersh
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: Automatic query generation, processed by Lucene using TF-IDF without normalization

OHSUSVMJ20¶

Run ID: OHSUSVMJ20
Participant: ohsu.hersh
Track: Genomics
Year: 2004
Submission: 8/24/2004
Type: automatic
Task: triage
Run description: CHI SQUARE 0.025 FULL FEATURE SET SVMLIGHT j=20.0 CLASSIFIER

OHSUVP¶

Run ID: OHSUVP
Participant: ohsu.hersh
Track: Genomics
Year: 2004
Submission: 8/24/2004
Type: automatic
Task: triage
Run description: CHI SQUARE 0.025 FULL FEATURE SET VOTING PERCEPTRON CLASSIFIER LINEAR KERNEL FN LEARNING RATE 20.0 FP LEARNING RATE 1.0

PD50501¶

Run ID: PD50501
Participant: u.padova
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: MySQL has been employed as IR engine. PD50501 has been obtained by using a new query expansion technique. After retrieving 20 documents, all the symbols are selected together with the keywords occurring in the 10-word window built around each symbol. The 50 most frequent keywords are selected and linked to the symbols co-occurring in the text windows. This way, a bi-partite graph is generated where symbols "point-to" frequent and close keywords.SPLIT (see PDTNsmp4 description) is then performed to disclose the mutual reinforcement relationship between symbols and keywords. The most "authoritative" keywords are frequently pointed-to by the symbols which frequently point-to "authoritative" keywords. The most "authoritative" keywords are added to expand the query which is in turn used to retrieve 1000 documents.

PDTNsmp4¶

Run ID: PDTNsmp4
Participant: u.padova
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: MySQL has been employed as IR engine. PDTNsmp4 has been obtained by indexing the collection using SPLIT, a stemming algorithm proposed and tested for some European languages. SPLIT infers directly the word formation rules from the corpus of words in the collection by exploiting a mutual reinforcement between the prefixes and suffixes of words. The hypothesis that SPLIT is effective has been tested in this domain-specific context, in which word formation rules can be different from those of common English.

pllsgen4a1¶

Run ID: pllsgen4a1
Participant: patolis.fujita
Track: Genomics
Year: 2004
Submission: 7/14/2004
Type: automatic
Task: adhoc
Run description: Word based indexing, no stemmer, Inquery stopwords, TF*IDF,BM25TF, MESH expansion,LocusLink Expansion, Pseudo feedback.

pllsgen4a2¶

Run ID: pllsgen4a2
Participant: patolis.fujita
Track: Genomics
Year: 2004
Submission: 7/14/2004
Type: automatic
Task: adhoc
Run description: Word based indexing, porter stemmer, Inquery stopwords, TF*IDF,BM25TF, MESH expansion,LocusLink Expansion, Pseudo feedback.

pllsgen4t1¶

Run ID: pllsgen4t1
Participant: patolis.fujita
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: triage
Run description: SVM Classifier using features from fulltext words, MeSH terms, Gene names, weighted by log-TF.

pllsgen4t2¶

Run ID: pllsgen4t2
Participant: patolis.fujita
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: triage
Run description: SVM Classifier using features from fulltext words, MeSH terms, Gene names, weighted by log-TF.

pllsgen4t3¶

Run ID: pllsgen4t3
Participant: patolis.fujita
Track: Genomics
Year: 2004
Submission: 8/27/2004
Type: automatic
Task: triage
Run description: SVM Classifier using features from fulltext words, MeSH terms, Gene names, weighted by log-TF.

pllsgen4t4¶

Run ID: pllsgen4t4
Participant: patolis.fujita
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: SVM classifier using features from fulltext words, MeSH terms and Gene names, weighted by log-TF

pllsgen4t5¶

Run ID: pllsgen4t5
Participant: patolis.fujita
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: SVM classifier using features from fulltext words, MeSH terms and Gene names, weighted by log-TF

PSE¶

Run ID: PSE
Participant: german.u.cairo
Track: Genomics
Year: 2004
Submission: 7/16/2004
Type: automatic
Task: adhoc
Run description: no query expansion, no-stemming, hyphenated expressions, OKAPI-BM25, PSE

RMITa¶

Run ID: RMITa
Participant: rmit.scholer
Track: Genomics
Year: 2004
Submission: 7/14/2004
Type: automatic
Task: adhoc
Run description: Plain text search using 'title', and 'need' fields from the query, and all fields from the Medline data. Searches were conducted using indexes from the Zettaire search system.

RMITb¶

Run ID: RMITb
Participant: rmit.scholer
Track: Genomics
Year: 2004
Submission: 7/14/2004
Type: automatic
Task: adhoc
Run description: Novel technique using o 'title', and 'need' fields from the queries o all fields from Medline data o term expansion o species-based search-space partitioning Searches were conducted using indexes from the Zettaire search system.

run1¶

Results | Participants | Input | Summary | Appendix

Run ID: run1
Participant: utwente
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: manual
Task: adhoc
Run description: We just recently joined the genomics track, because of this our system couldnt run all queries yet. the ones it could run are included in our results. The titles and abstracts from the medline records were used.

rutgersGAH1¶

Run ID: rutgersGAH1
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 7/14/2004
Type: automatic
Task: adhoc
Run description: We used MG retrieval system to index the MEDLINE article collection. We used title, abstract, chemical names and mesh terms sections of the articles to index the collection. We constructed queries from topic titles and needs. We performed the retrieval as a two-stage process. First we identifed the relevant documents by MG, and then we performed boolean full-text on the results using MySQL.

rutgersGAH2¶

Run ID: rutgersGAH2
Participant: rutgers.dayanik
Track: Genomics
Year: 2004
Submission: 7/14/2004
Type: automatic
Task: adhoc
Run description: We used MG retrieval system to index the MEDLINE article collection. We used title, abstract, chemical names and mesh terms sections of the articles to index the collection. We constructed queries from topic titles and needs. We computed a relevance score for ranking using the frequency and inverse document frequency of query terms as well as the proximity of topic title query terms.

shefauto1¶

Run ID: shefauto1
Participant: u.sheffield.gaizauskas
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: Use the UMLS to extend synonyms for each keywords. Use LT chunker to find noun phrases and abbrievations.

shefauto2¶

Run ID: shefauto2
Participant: u.sheffield.gaizauskas
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: No synonyms are used. Use LT chunker to find noun phrases and abbrievations.

tgnNecaux¶

Run ID: tgnNecaux
Participant: tarragon
Track: Genomics
Year: 2004
Submission: 7/12/2004
Type: automatic
Task: adhoc
Run description: Variant run. Eliminates noise phrases, then uses on all fields and combines using , with CONTEXT field as the auxiliary.

tgnSplit¶

Run ID: tgnSplit
Participant: tarragon
Track: Genomics
Year: 2004
Submission: 7/12/2004
Type: automatic
Task: adhoc
Run description: Base run. Eliminates noise phrases, then uses on all fields and combines using .

THIRcat01¶

Run ID: THIRcat01
Participant: tsinghua.ma
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: balanced between +/- subsets

THIRcat02¶

Run ID: THIRcat02
Participant: tsinghua.ma
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: kmeans clustering, low threshold

THIRcat03¶

Run ID: THIRcat03
Participant: tsinghua.ma
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: using GO-item related words to expand the feature directory

THIRcat04¶

Run ID: THIRcat04
Participant: tsinghua.ma
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: kmeans clustering, high threshold

THIRcat05¶

Run ID: THIRcat05
Participant: tsinghua.ma
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: bool weight instead of TFIDF weight

THUIRgen01¶

Run ID: THUIRgen01
Participant: tsinghua.ma
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: manual
Task: adhoc
Run description: Use our IR systerm named TMiner.Recognize term in non_text fields,in particular MeSH,Other Term (OT),Gene Symbol(GS) and substance name(RN).Give different weights to weights the field of title(TI) and MeSH.Use relevance feedback technology. All above processes are done autaomatiacally. Manually select gene names and expend queries by an abbreviations dictionary.

THUIRgen02¶

Run ID: THUIRgen02
Participant: tsinghua.ma
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: Use our IR systerm named TMiner Recognize term in non_text fields,in particular MeSH,Other Term (OT),Gene Symbol(GS) and substance name(RN).Give different weights to weights the fields of title(TI) and MeSH.Use relevance feedback technology.Query Expansion (QE) uses an abbreviation dictionary. All processes are done automatically.

tnog2¶

Run ID: tnog2
Participant: tno.kraaij
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: Baseline run, using only title and abstract fields. LM based retrieval model

tnog3¶

Run ID: tnog3
Participant: tno.kraaij
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: Combination run baseline run tnog2 interpolated with a run based on MeSH queries. MeSH queries were automatically derived from the top 3 documents in the tnog2 run.

tq0¶

Run ID: tq0
Participant: nlm.umd.ul
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: TexTool, a text neighboring program based on TFIDF vector retrieval, was applied to the queries after removing common stopwords (the, we, etc.) from a list of 313, entering the title twice and the need and context once, to find 1,000 similar abstracts.

TRICSUSM¶

Run ID: TRICSUSM
Participant: u.sanmarcos
Track: Genomics
Year: 2004
Submission: 8/26/2004
Type: manual
Task: triage
Run description: We have designed a decision tree learning algorithm for the text categorization. The set of atributes is based on the keywords, the contents of the glosref tag, and font-change tags. We searched in the abstract for such attributes to determine whether to select the text or not.

UBgtNormJM1¶

Run ID: UBgtNormJM1
Participant: suny.buffalo
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: Document indexing Language models using linear interpolation smoothing, Query construction topic expanded with NLM's MetaMap, common phrases discarded before retrieval.

UIowaGN1¶

Run ID: UIowaGN1
Participant: u.iowa
Track: Genomics
Year: 2004
Submission: 7/16/2004
Type: automatic
Task: adhoc
Run description: We use all fields in the topics. We automatically expand synonyms using LocusLink. We then conduct retrieval using Lucy/Zettair search engine and rank documents by relevance score. Phrases are in the queries.

utaauto¶

Run ID: utaauto
Participant: u.tampere
Track: Genomics
Year: 2004
Submission: 7/7/2004
Type: automatic
Task: adhoc
Run description: -InQuery retrieval system -Title, Abstract, MH, RN, and GS fields indexed -Text- and MH-field based retrieval -Removal of non-topic words from the topics using average term frequency statistics -Identification of phrases in topics using a collocation method -Pseudorelevance feedback MH terms added to the first queries -Weighting of query keys whose document frequency is low (structural weighting)

utamanu¶

Run ID: utamanu
Participant: u.tampere
Track: Genomics
Year: 2004
Submission: 7/10/2004
Type: manual
Task: adhoc
Run description: -InQuery retrieval system -Title, Abstract, MH, RN, and GS fields indexed -Text- and MH-field based retrieval -The formulation of Boolean queries -New terms from MeSH, genome databases, and dictionaries -Queries were reformulated using MH terms from the top documents of the initial search

uwmtDg04n¶

Run ID: uwmtDg04n
Participant: u.waterloo.clarke
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: From the initial topic, after the removal of stopwords, we have created a plain query (without any expansions) and an expanded query that used data from the Medstract project, LocusLink and the EuGenes database to find acronym expansions and aliases/synonyms for gene names and proteins. We have run these two queries and analyzed the top documents returned in order to validate the possible expansions found in the databases and to find additional terms resulting from permutations of terms (such as "nf.kappa.b" and "kappa.b.nf"). In the second pass, we have run the resulting query (validated expansions/aliases) and scored the documents using Okapi. For the third pass, we computed additional expansion terms based on statistical feedback (obtained by analyzing the top documents returned) and Google feedback (analyzing the passages for the top 20 documents returned by Google for the unexpanded "need-only" query). We then executed the query (expanded by the feedback terms) and scored the documents using Okapi. In the last pass, we used the species names found in the topic to re-rank the documents. In general, if, for instance, "mice" occurs in the need of a topic, all documents containing "mouse", "mice", or "mus.musculus" are getting a better rank.

uwmtDg04tn¶

Run ID: uwmtDg04tn
Participant: u.waterloo.clarke
Track: Genomics
Year: 2004
Submission: 7/15/2004
Type: automatic
Task: adhoc
Run description: From the initial topic, after the removal of stopwords, we have created a plain query (without any expansions) and an expanded query that used data from the Medstract project, LocusLink and the EuGenes database to find acronym expansions and aliases/synonyms for gene names and proteins. We have run these two queries and analyzed the top documents returned in order to validate the possible expansions found in the databases and to find additional terms resulting from permutations of terms (such as "nf.kappa.b" and "kappa.b.nf"). In the second pass, we have run the resulting query (validated expansions/aliases) and scored the documents using Okapi. For the third pass, we computed additional expansion terms based on statistical feedback (obtained by analyzing the top documents returned) and Google feedback (analyzing the passages for the top 20 documents returned by Google for the unexpanded "need-only" query). We then executed the query (expanded by the feedback terms) and scored the documents using Okapi. In the last pass, we used the species names found in the topic to re-rank the documents. In general, if, for instance, "mice" occurs in the need of a topic, all documents containing "mouse", "mice", or "mus.musculus" are getting a better rank.

wdtriage1¶

Run ID: wdtriage1
Participant: indiana.u.yang
Track: Genomics
Year: 2004
Submission: 8/30/2004
Type: automatic
Task: triage
Run description: Categorizing by Rainbow with Bayes

wdvqlx1¶

Run ID: wdvqlx1
Participant: indiana.u.yang
Track: Genomics
Year: 2004
Submission: 7/16/2004
Type: automatic
Task: adhoc
Run description: query expansion with nouns and phrases

wdvqlxa1¶

Run ID: wdvqlxa1
Participant: indiana.u.yang
Track: Genomics
Year: 2004
Submission: 7/16/2004
Type: automatic
Task: adhoc
Run description: query expansion with Gene Ontology synonyms, nouns, and phrases

wiscW¶

Run ID: wiscW
Participant: u.wisconsin
Track: Genomics
Year: 2004
Submission: 8/20/2004
Type: automatic
Task: annhi
Run description: Statistical model using words only as features.

wiscWR¶

Run ID: wiscWR
Participant: u.wisconsin
Track: Genomics
Year: 2004
Submission: 8/20/2004
Type: automatic
Task: annhi
Run description: Statistical model using words plus learned syntactic/contextual rules as features.

wiscWRT¶

Run ID: wiscWRT
Participant: u.wisconsin
Track: Genomics
Year: 2004
Submission: 8/20/2004
Type: automatic
Task: annhi
Run description: Statistical model using words, learned syntactic rules, and induced GO-specific terms as features.

wiscWT¶

Run ID: wiscWT
Participant: u.wisconsin
Track: Genomics
Year: 2004
Submission: 8/20/2004
Type: automatic
Task: annhi
Run description: Statistical model using words plus induced GO-specific terms as features.

york04g1¶

Run ID: york04g1
Participant: york.u
Track: Genomics
Year: 2004
Submission: 7/16/2004
Type: automatic
Task: adhoc
Run description: 1. Retrieval using probability model 2. Weighting using BM25 3. Indexing using Okapi2.31 4. Relevance feedback using Okapi 5. Using machine learning techniques for query expansion

york04g2¶

Run ID: york04g2
Participant: york.u
Track: Genomics
Year: 2004
Submission: 7/16/2004
Type: manual
Task: adhoc
Run description: 1. Retireval using probabilistic model 2. Weighting using BM25 3. Indexing using Okapi2.31 4. Relevance feedback using Okapi 5. Using machine learning techniques for query expansion