Runs - Blog 2006¶

Abstract¶

Results | Participants | Proceedings | Appendix

Run ID: Abstract
Participant: robert-gordonu.harper
Track: Blog
Year: 2006
Submission: 9/1/2006
Task: open_task

blog06r1¶

Run ID: blog06r1
Participant: cmu.dir.callan
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: unigram presence features by Bayesian Logistic Regression

blog06r2¶

Run ID: blog06r2
Participant: cmu.dir.callan
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: bigram presence features by Bayesian Logistic Regression

blog06r3¶

Run ID: blog06r3
Participant: cmu.dir.callan
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: trigram presence features by Bayesian Logistic Regression

blog06r4¶

Run ID: blog06r4
Participant: cmu.dir.callan
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: unigram presence features by Bayesian Logistic Regression

blog06r5¶

Run ID: blog06r5
Participant: cmu.dir.callan
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: boigram presence features by Bayesian Logistic Regression

blog06r6¶

Run ID: blog06r6
Participant: cmu.dir.callan
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: trigram presence features by Bayesian Logistic Regression

csirabstract¶

Results | Participants | Proceedings | Appendix

Run ID: csirabstract
Participant: csiro-ict.mcarthur
Track: Blog
Year: 2006
Submission: 9/1/2006
Task: open_task

csirofull¶

Results | Participants | Proceedings | Appendix

Run ID: csirofull
Participant: csiro-ict.mcarthur
Track: Blog
Year: 2006
Submission: 10/3/2006
Task: open_task

ICT¶

Run ID: ICT
Participant: cas-iiis.tan
Track: Blog
Year: 2006
Submission: 7/28/2006
Type: automatic
Task: opinion
Run description: expand run

IIIS¶

Run ID: IIIS
Participant: cas-iiis.tan
Track: Blog
Year: 2006
Submission: 7/28/2006
Type: automatic
Task: opinion
Run description: base of line

mcwil1k¶

Run ID: mcwil1k
Participant: fudan.wu
Track: Blog
Year: 2006
Submission: 7/28/2006
Type: automatic
Task: opinion
Run description: title words from description and narrative no postprocess

mcwil2knl¶

Run ID: mcwil2knl
Participant: fudan.wu
Track: Blog
Year: 2006
Submission: 7/28/2006
Type: automatic
Task: opinion
Run description: title words from description and narrative

necabstract¶

Results | Participants | Proceedings | Appendix

Run ID: necabstract
Participant: nec-labs.tseng
Track: Blog
Year: 2006
Submission: 9/5/2006
Task: open_task

necnistfinal¶

Results | Participants | Proceedings | Appendix

Run ID: necnistfinal
Participant: nec-labs.tseng
Track: Blog
Year: 2006
Submission: 10/3/2006
Task: open_task

NII1¶

Run ID: NII1
Participant: nii-eguchi
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: (to be reported)

NII3¶

Run ID: NII3
Participant: nii-eguchi
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: (to be reported)

NII5¶

Run ID: NII5
Participant: nii-eguchi
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: (to be reported)

NII6¶

Run ID: NII6
Participant: nii-eguchi
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: (to be reported)

NII7¶

Run ID: NII7
Participant: nii-eguchi
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: (to be reported)

NIIabstract¶

Results | Participants | Proceedings | Appendix

Run ID: NIIabstract
Participant: nii-eguchi
Track: Blog
Year: 2006
Submission: 9/1/2006
Task: open_task

NIIfinal¶

Results | Participants | Proceedings | Appendix

Run ID: NIIfinal
Participant: nii-eguchi
Track: Blog
Year: 2006
Submission: 10/2/2006
Task: open_task

ParTiDesDmt2¶

Run ID: ParTiDesDmt2
Participant: umaryland.oard
Track: Blog
Year: 2006
Submission: 7/25/2006
Type: automatic
Task: opinion
Run description: Blog postings (i.e., permalinks) are segmented into paragraphs. Queries are automatically formulated using the title and description fields. Retrieved paragraphs are lematized and computed a sentiment probability score by checking a manually created sentiment lexicon and an automatically computed sentiment lexicon. Paragraphs below a heuristic threshold (15%) of sentiment probability are demoted 2 times of their positions in the rank list. (The automatically computed sentiment lexicon is collected by selecting the words with low IDFs from the paragraphs retrieved with the title fields only, and then using Turney's Pointwise Mutual Information approach.)

ParTiDesDmt3¶

Run ID: ParTiDesDmt3
Participant: umaryland.oard
Track: Blog
Year: 2006
Submission: 7/25/2006
Type: automatic
Task: opinion
Run description: Blog postings (i.e., permalinks) are segmented into paragraphs. Queries are automatically formulated using the title and description fields. Retrieved paragraphs are lematized and computed a sentiment probability score by checking a manually created sentiment lexicon and an automatically computed sentiment lexicon. Paragraphs below a heuristic threshold (15%) of sentiment probability are demoted 3 times of their positions in the rank list. (The automatically computed sentiment lexicon is collected by selecting the words with low IDFs from the paragraphs retrieved with the title fields only, and then using Turney's Pointwise Mutual Information approach.)

ParTitDef¶

Run ID: ParTitDef
Participant: umaryland.oard
Track: Blog
Year: 2006
Submission: 7/25/2006
Type: automatic
Task: opinion
Run description: Default run on paragraphs with the title fields only. Blog postings (i.e., permalinks) are segmented into paragraphs. Queries are automatically formulated using the title fiels only. No opinion detection is performed.

ParTitDesDef¶

Run ID: ParTitDesDef
Participant: umaryland.oard
Track: Blog
Year: 2006
Submission: 7/25/2006
Type: automatic
Task: opinion
Run description: Default run for retrieving paragraphs using title and description fields. Blog postings (i.e., permalinks) are segmented into paragraphs and retrieval is performed on the paragraphs. Queries are automatically formulated using the title and description fields. No opinion detection is performed.

PasTitDesDef¶

Run ID: PasTitDesDef
Participant: umaryland.oard
Track: Blog
Year: 2006
Submission: 7/25/2006
Type: automatic
Task: opinion
Run description: Default run for retrieving passages using the title and description fiels. Blog postings (i.e., permalinks) are segmented into passages by sliding a window with a fixed size. Queries are automatically formulated using the title and description fields. No opinion detection is performed.

pisaBlDes¶

Run ID: pisaBlDes
Participant: upisa.attardi
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: Index obtained from permalinks, replacing body from content in corresponding link if present, filtering out decorative HTML elements, filtering spam documents according to a blacklist.

pisaBlDesOp6¶

Run ID: pisaBlDesOp6
Participant: upisa.attardi
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: Use proximity 6 to an opinionated word. Index obtained from permalinks, replacing body from content in corresponding link if present, filtering out decorative HTML elements, filtering spam documents according to a blacklist.

pisaBlTit¶

Run ID: pisaBlTit
Participant: upisa.attardi
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: Index obtained from permalinks, replacing body from content in corresponding link if present, filtering out decorative HTML elements, filtering spam documents according to a blacklist.

pisaBlTitOp6¶

Run ID: pisaBlTitOp6
Participant: upisa.attardi
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: Use query with proximity 6 to an opinionated word. Index obtained from permalinks, replacing body from content in corresponding link if present, filtering out decorative HTML elements, filtering spam documents according to a blacklist.

rguBL¶

Run ID: rguBL
Participant: robert-gordonu.harper
Track: Blog
Year: 2006
Submission: 8/11/2006
Type: automatic
Task: opinion
Run description: Standard language modelling retrieval model expanding the query with opinionated terms.

rguOPN¶

Run ID: rguOPN
Participant: robert-gordonu.harper
Track: Blog
Year: 2006
Submission: 8/11/2006
Type: automatic
Task: opinion
Run description: standard language modelling framework extended using the proportion of opinionated terms (taken from a list) in documents as part of the score.

rguSBJ¶

Run ID: rguSBJ
Participant: robert-gordonu.harper
Track: Blog
Year: 2006
Submission: 8/11/2006
Type: automatic
Task: opinion
Run description: standard language modelling framework extended using the proportion of subjective terms in documents as part of the score

THUBLOGMF¶

Run ID: THUBLOGMF
Participant: tsinghuau.zhang
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: step1 search on permalinks and feeds for relevant (multi-field model) Step2 use a dictionary for opinion judgement on step1 result files context information. (multi-field model) Final result use combined similarities on that of step1 and step2.

THUBLOGMFNO¶

Run ID: THUBLOGMFNO
Participant: tsinghuau.zhang
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: search on permalinks and feeds (mul-field model) for relevant. no further opinion judgement.

THUBLOGPFS¶

Run ID: THUBLOGPFS
Participant: tsinghuau.zhang
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: step1 search on permalinks and feeds for relevant (PFS model) Step2 use a dictionary for opinion judgement on step1 result files context information. (PFS model) Final result use combined similarities on that of step1 and step2.

THUBLOGPMCT¶

Run ID: THUBLOGPMCT
Participant: tsinghuau.zhang
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: step1 search on permalinks for relevant Step2 use a dictionary for opinion judgement on step1 result files context information. Final result use combined similarities on that of step1 and step2.

THUBLOGPMFU¶

Run ID: THUBLOGPMFU
Participant: tsinghuau.zhang
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: step1 search on permalinks for relevant, long query Step2 use a dictionary for opinion judgement on step1 result files with full-text information. Final result use combined similarities on that of step1 and step2.

UABas11¶

Run ID: UABas11
Participant: umaryland-bc.finin
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: Baseline, uncleaned collection, 7 feature SVM, 1 1 weighting

UAEX11¶

Run ID: UAEX11
Participant: umaryland-bc.finin
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: Experiment, semicleaned collection, 11 feature SVM, 1 1 weighting

UAEX12¶

Run ID: UAEX12
Participant: umaryland-bc.finin
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: Experiment, semicleaned collection, 11 feature SVM, 1 2 weighting

UAEX13¶

Run ID: UAEX13
Participant: umaryland-bc.finin
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: Experiment, semicleaned collection, 11 feature SVM, 1 3 weighting

UAEX21¶

Run ID: UAEX21
Participant: umaryland-bc.finin
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: Experiment, semicleaned collection, 11 feature SVM, 2 1 weighting

UALR06a260r2¶

Run ID: UALR06a260r2
Participant: uarkansas-littlerock.joshi
Track: Blog
Year: 2006
Submission: 7/28/2006
Type: automatic
Task: opinion
Run description: We extracted features from 254 training blog-posts (labelled data). Then the features were ranked by Information Gain Feature selection technique and top 260 features were selected for using SVMs (Support Vector Machines) . The queries were run only on title field (automatic run) and has training accuracy of around 74%.

UALR06a500r4¶

Run ID: UALR06a500r4
Participant: uarkansas-littlerock.joshi
Track: Blog
Year: 2006
Submission: 7/28/2006
Type: automatic
Task: opinion
Run description: We manually created a set of 500 likely features (mostly adjectives and adverbs). These features were used for training SVMs (Support Vector Machines). The queries were run only on title field (automatic run). SVM has training accuracy of around 80% over 300 samples. Iterative procedure technique allowed for faster convergence of training of SVMs whuile maintaining the accuracy.

UALR06m260r3¶

Run ID: UALR06m260r3
Participant: uarkansas-littlerock.joshi
Track: Blog
Year: 2006
Submission: 7/28/2006
Type: manual
Task: opinion
Run description: We extracted features from 254 training blog-posts (labelled data). Then the features were ranked by Information Gain Feature selection technique and top 260 features were selected for using SVMs (Support Vector Machines) . The queries were run only on title and description fields (manual run). SVM has training accuracy of around 74% over 300 samples.

UALR06m500r5¶

Run ID: UALR06m500r5
Participant: uarkansas-littlerock.joshi
Track: Blog
Year: 2006
Submission: 7/28/2006
Type: manual
Task: opinion
Run description: We manually created a set of 500 likely features (mostly adjectives and adverbs). These features were used for training SVMs (Support Vector Machines). The queries were run on title as well as description fields (manual run). SVM has training accuracy of around 80% over 300 samples. Iterative procedure technique allowed for faster convergence of training of SVMs whuile maintaining the accuracy.

UALR06m5Tr1¶

Run ID: UALR06m5Tr1
Participant: uarkansas-littlerock.joshi
Track: Blog
Year: 2006
Submission: 7/28/2006
Type: manual
Task: opinion
Run description: We extracted the most generic features such as ratio of adjectives to the total number of words per blog-post, no. of adverbs to the total no. words per blog-post etc. The interest here is to find statistical generic features that are independent of the actual content of the blog post.

UAmsB06All¶

Run ID: UAmsB06All
Participant: uamsterdam.ilps
Track: Blog
Year: 2006
Submission: 7/29/2006
Type: automatic
Task: opinion
Run description: LM retrieval plus reweighting based on an "opinionatedness" level, on "spam level", on indegrees at the post and feed level, on a temporal distribution of the query words in the corpus, and on occurrence of the query as a phrase in the anchor text/tags/title. This combines the other submissions (plus some more data).

UAmsB06Base¶

Run ID: UAmsB06Base
Participant: uamsterdam.ilps
Track: Blog
Year: 2006
Submission: 7/29/2006
Type: automatic
Task: opinion
Run description: Baseline run vanilla language-modeling retrieval on content only.

UAmsB06L¶

Run ID: UAmsB06L
Participant: uamsterdam.ilps
Track: Blog
Year: 2006
Submission: 7/29/2006
Type: automatic
Task: opinion
Run description: LM retrieval, plus reweighting on log indegree both at feed level and post level

UAmsB06O¶

Run ID: UAmsB06O
Participant: uamsterdam.ilps
Track: Blog
Year: 2006
Submission: 7/29/2006
Type: automatic
Task: opinion
Run description: LM retrieval, plus reweighting based on an "opinionatedness" level assigned using a dictionary (General Inquirer) and a topical-sentence filtering mechanism

UAmsB06S¶

Run ID: UAmsB06S
Participant: uamsterdam.ilps
Track: Blog
Year: 2006
Submission: 7/29/2006
Type: automatic
Task: opinion
Run description: LM retrieval, plus reweighting based on a spam classification

uaplabstract¶

Results | Participants | Proceedings | Appendix

Run ID: uaplabstract
Participant: umaryland-bc.finin
Track: Blog
Year: 2006
Submission: 9/1/2006
Task: open_task

uaplfinal¶

Results | Participants | Proceedings | Appendix

Run ID: uaplfinal
Participant: umaryland-bc.finin
Track: Blog
Year: 2006
Submission: 10/3/2006
Task: open_task

ucscauto¶

Run ID: ucscauto
Participant: ucalifornia.sc.zhang
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: This is an automatic run, and the title field of each topic is used to retrieve documents from original permalinks.

ucscdesc¶

Run ID: ucscdesc
Participant: ucalifornia.sc.zhang
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: This run uses the description field of each topic for document retrieval. The retrieved documents are classified by a Bayesian Logistic Regression classifier.

ucscrelfb¶

Run ID: ucscrelfb
Participant: ucalifornia.sc.zhang
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: manual
Task: opinion
Run description: This run uses the description field of each topic for document retrieval. The retrieved documents are classified by a Bayesian Logistic Regression classifier, and the classified results are then given to the search engine as relevance feedback. The final search results after relevance feedback are submitted.

uicsr¶

Run ID: uicsr
Participant: uillinois.chicago.zhang
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: TREC 06 BLOG, Wei Zhang, UIC, NoA IR rank

uicst¶

Run ID: uicst
Participant: uillinois.chicago.zhang
Track: Blog
Year: 2006
Submission: 7/30/2006
Type: automatic
Task: opinion
Run description: TREC 06 BLOG, Wei Zhang, UIC, NoA STC rank

woqln2¶

Run ID: woqln2
Participant: indianau.yang
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: long query (emphasized nouns), without noise reduction, okapi weight, on-topic and opinion post-retrieval reranking

woqs2¶

Run ID: woqs2
Participant: indianau.yang
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: short query, without noise reduction, okapi weight, on-topic and opinion post-retrieval reranking

wxoqf2¶

Run ID: wxoqf2
Participant: indianau.yang
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: fusion of short and long query (emphasized nouns), with noise reduction, okapi weight, on-topic and opinion post-retrieval reranking

wxoqln2¶

Run ID: wxoqln2
Participant: indianau.yang
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: long query (emphasized nouns), with noise reduction, okapi weight, on-topic and opinion post-retrieval reranking

wxoqs2¶

Run ID: wxoqs2
Participant: indianau.yang
Track: Blog
Year: 2006
Submission: 7/31/2006
Type: automatic
Task: opinion
Run description: short query, with noise reduction, okapi weight, on-topic and opinion post-retrieval reranking