Runs - Blog 2009¶

BIT09P¶

Run ID: BIT09P
Participant: BIT
Track: Blog
Year: 2009
Submission: 8/30/2009
Type: automatic
Task: feed
MD5: 5f68c75a879601f7830e97eaab34de69
Run description: BIT09P is the same as BIT09PH, except it only uses permalinks indexed corpus.

BIT09PH¶

Run ID: BIT09PH
Participant: BIT
Track: Blog
Year: 2009
Submission: 8/30/2009
Type: automatic
Task: feed
MD5: 2aa506357a64c51094b9dca691a397f9
Run description: BIT09PH uses wikipedia corpus to get relevant feedback documents for query model, and use an opinionated lexicon to generate query model for opinionated faceted blog distillation. We also use metasearch engine to get personal blogs from blogs.myspace.com to generate query model for personal faceted blog distillation. Results come from permalinks and homepages indexed corpus. We use Global Representation model, which all posts from the same feed are combined into one document when compute the relevance score.

combined¶

Run ID: combined
Participant: USI
Track: Blog
Year: 2009
Submission: 8/28/2009
Type: automatic
Task: feed
MD5: 8067c71b01d25d510d657d8556af3465
Run description: In this run, we combine the two other runs (OWA and Regularization) with a linear combination. It takes into account both similarity of the posts to other retrieved posts (by regularization) and good aggregation over the posts (by OWA) and generate new score for each feed.

FEUPirlab1¶

Run ID: FEUPirlab1
Participant: FEUP
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: f2b1b34fcbca9140be0ecf1b96cdcdf9
Run description: Run combining BM25 scores form a post search.

FEUPirlab2¶

Run ID: FEUPirlab2
Participant: FEUP
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: 2975bc566d4bc6c4e2483212ac82953d
Run description: Run combining BM25 scores form a post search, including a boost for posts with invalid dates.

FEUPirlab3¶

Run ID: FEUPirlab3
Participant: FEUP
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: 71bfa7e3b9b486e4eb25306c3002399a
Run description: Run combining BM25 scores form a post search, including a boost for posts with invalid dates and a boost for newer posts.

FEUPirlab4¶

Run ID: FEUPirlab4
Participant: FEUP
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: 4cbfc8ffcd1c479bb7a933ef794e8f68
Run description: Run combining BM25 scores form a post search, including a boost for posts with invalid dates and a boost for older posts.

ICTNETBDRUN1¶

Run ID: ICTNETBDRUN1
Participant: ICTNET
Track: Blog
Year: 2009
Submission: 8/31/2009
Type: automatic
Task: feed
MD5: 280e25cc3fe00b6380c1b88988a2f074
Run description: For each feed, we only use information of top 10000 ad-hoc relevant posts to decide the facet value.

ICTNETBDRUN2¶

Run ID: ICTNETBDRUN2
Participant: ICTNET
Track: Blog
Year: 2009
Submission: 8/31/2009
Type: automatic
Task: feed
MD5: 1f668943a4ceebf2aa410ca876927529
Run description: For each feed, we only use information of top 10000 ad-hoc relevant posts to decide the facet value.

ICTNETBDRUN3¶

Run ID: ICTNETBDRUN3
Participant: ICTNET
Track: Blog
Year: 2009
Submission: 8/31/2009
Type: automatic
Task: feed
MD5: 7eb2a28013af4a86c62fbec09d845f1a
Run description: for each feed ,we divide corresponding posts according to day they are crawled, and take into account the relevence distribution among days for feed relevece jugement.

ICTNETBDRUN4¶

Run ID: ICTNETBDRUN4
Participant: ICTNET
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: 1982689bf4a5088516d1c84d708447ae
Run description: this run is similar to ICTNETBDRUN3 except some strategy details.

ICTNETTSRun1¶

Run ID: ICTNETTSRun1
Participant: ICTNET
Track: Blog
Year: 2009
Submission: 8/27/2009
Type: automatic
Task: topstories
MD5: f165bee14cc83993f4bea81aea566898
Run description: It's a straightforward method using sum of BM25 scores with posts in given day to measure the improtance of headlines.

ICTNETTSRun2¶

Run ID: ICTNETTSRun2
Participant: ICTNET
Track: Blog
Year: 2009
Submission: 8/27/2009
Type: automatic
Task: topstories
MD5: aaa5f5f0765ec95a9f0efab74ef24c92
Run description: It's a straightforward method using sum of BM25 scores with posts in given day to measure the improtance of headlines.

ICTNETTSRun3¶

Run ID: ICTNETTSRun3
Participant: ICTNET
Track: Blog
Year: 2009
Submission: 8/27/2009
Type: automatic
Task: topstories
MD5: 8fb2d1eaafae6ce2e6be0361260fc8c6
Run description: It's a straightforward method using sum of BM25 scores with posts in given day to measure the improtance of headlines.

ICTNETTSRun4¶

Run ID: ICTNETTSRun4
Participant: ICTNET
Track: Blog
Year: 2009
Submission: 8/27/2009
Type: automatic
Task: topstories
MD5: ab76383512b0e9699ccedef83cca9f17
Run description: It's a straightforward method using sum of BM25 scores with posts in given day to measure the improtance of headlines.

IlpsBDm1T¶

Run ID: IlpsBDm1T
Participant: UAms
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: c2448d23613083db59032a22b6cd20ca
Run description: Model1 on title-only index using queries expanded against a news collection and Wikipedia.

IlpsBDm2T¶

Run ID: IlpsBDm2T
Participant: UAms
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: fc99c3bd89822c18cef20ee5b406b898
Run description: Model2 on title-only index using queries expanded against a news collection and Wikipedia.

IlpsBDmxfT¶

Run ID: IlpsBDmxfT
Participant: UAms
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: 8d6d8adeee70fba4203d69e4527bb1e4
Run description: Expanded queries (news and Wikipedia) on title-only index using model1 and model2, externally combined with 0.6 model1, 0.4 model2. Top 300 feeds used in combination: normalize scores, weight them, combine (CombMNZ). Afterwards indicators for facets are applied.

IlpsBDmxT¶

Run ID: IlpsBDmxT
Participant: UAms
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: 8f0f311e33e430060b4be727dbf6ce90
Run description: Expanded queries (news and Wikipedia) on title-only index using model1 and model2, externally combined with 0.6 model1, 0.4 model2. Top 300 feeds used in combination: for normalize scores, weight them, combine (CombMNZ).

IlpsTSExP¶

Run ID: IlpsTSExP
Participant: UAms
Track: Blog
Year: 2009
Submission: 8/29/2009
Type: automatic
Task: topstories
MD5: 90a2f8b2ea251e7d61841f133d81aab2
Run description: extract keyterms from top 5000 blog posts, use as queries on headline index with date restrictions.

IlpsTSExT¶

Run ID: IlpsTSExT
Participant: UAms
Track: Blog
Year: 2009
Submission: 8/29/2009
Type: automatic
Task: topstories
MD5: 6b12291eca5590615602c043de0f0090
Run description: extract keyterms from top 5000 blog post titles, use as queries on headline index with date restrictions.

IlpsTSHlP¶

Run ID: IlpsTSHlP
Participant: UAms
Track: Blog
Year: 2009
Submission: 8/30/2009
Type: automatic
Task: topstories
MD5: b8e232b2d2c36baec917d77cb548d421
Run description: Calculate headline likelihood on full post index of the given period. Use comments as association strength.

IlpsTSHlT¶

Run ID: IlpsTSHlT
Participant: UAms
Track: Blog
Year: 2009
Submission: 8/30/2009
Type: automatic
Task: topstories
MD5: b83e0348434b1cdb6280aaaf4c1a4c9c
Run description: Calculate headline likelihood on title index of the given period. Use comments as association strength.

IowaSBD0901¶

Run ID: IowaSBD0901
Participant: IowaS
Track: Blog
Year: 2009
Submission: 8/28/2009
Type: automatic
Task: feed
MD5: 2c62780cf42fc2f79940b2aebf330242
Run description: The posts are indexed using Lucene engine. The queries are composed using the tag and up to 10 terms extracted from the tag using TFIDF. Two hundred posts are retrieved and are analyzed according to the querys facets. A lingpipe classifier is used for the opinionated/factual facet, the length of the posts us used for the in-depth/shallow facet, and the number of personal pronouns is used for the personal/official facet. External Resource: Lingpipe sentiment polarity training data

IowaSBD0902¶

Run ID: IowaSBD0902
Participant: IowaS
Track: Blog
Year: 2009
Submission: 8/28/2009
Type: automatic
Task: feed
MD5: 653bdb878968ce3af3cf6e1c671f87b1
Run description: The posts are indexed using Lucene engine. The queries are composed using the tag and up to 10 terms extracted from the tag using TFIDF. Fifty posts are retrieved for pseudo-feedback. Using the latent Dirichlet relevance model (Ha-Thuc and Srinivasan), we obtain ten query expansion terms, which are added to the original query. Two hundred posts are then retrieved and analyzed according to the querys facets. A lingpipe classifier is used for the opinionated/factual facet, the length of the posts us used for the in-depth/shallow facet, and the number of personal pronouns is used for the personal/official facet. External resource: Lingpipe sentiment polarity training data

IowaSBT0901¶

Run ID: IowaSBT0901
Participant: IowaS
Track: Blog
Year: 2009
Submission: 8/27/2009
Type: automatic
Task: topstories
MD5: f5a3cc6fc65cc2c4bc11d0db00ea1e5d
Run description: Ranking headlines for a query date: Given a query date (d) we start with the URLs of the corresponding set of headlines. As directed these correspond to headlines published on d-1, d and d+1 dates. We search the blog dataset for posts that contain a headline URL. Headline URLs are then ranked by the number of posts returned. The top 100 are submitted for the query date. Ranking posts for a headline: The posts are indexed using Lucene. When a URL is found in a post, we extract the surrounding window of text. The size of this window is 800 characters before and after the URL (including HTML). Given a headline (h) these texts are accumulated from all posts containing the URL for h. We then extract keywords from these texts using our latent Dirichlet relevance model (Ha-Thuc and Srinivasan, A Latent Dirichlet Framework for Relevance Modeling, 5th Asia Information Retrieval Symposium, 2009). These keywords are used to query the blog dataset. The top 10 are returned as the set of posts to read for headline h.

IowaSBT0902¶

Run ID: IowaSBT0902
Participant: IowaS
Track: Blog
Year: 2009
Submission: 8/27/2009
Type: automatic
Task: topstories
MD5: 4852ace2dd62614f220e861d1fc19b67
Run description: Ranking headlines for a query date: Given a query date (d) we start with the URLs of the corresponding set of headlines. As directed these correspond to headlines published on d-1, d and d+1 dates. We search the blog dataset for posts that contain a headline URL. Headline URLs are then ranked by the number of posts returned. The top 100 are submitted for the query date. Ranking posts for a headline: The posts are indexed using Lucene. When a URL is found in a post, we extract the surrounding window of text. The size of this window is 800 characters before and after the URL (including HTML). Given a headline (h) these texts are accumulated from all posts containing the URL for h. We then extract keywords from these texts using our latent Dirichlet relevance model (Ha-Thuc and Srinivasan). These keywords are used to query the blog dataset. In the set of returned posts the ones containing the headline URL are boosted to the top. The top 10 are returned as the set of posts to read for headline h.

IowaSBT0903¶

Run ID: IowaSBT0903
Participant: IowaS
Track: Blog
Year: 2009
Submission: 8/27/2009
Type: automatic
Task: topstories
MD5: ea0a7e1e1130021785f53c14e971bc53
Run description: Ranking headlines for a query date: Given query date d, we rank headlines by an intensity measure indicating how frequently the queries are discussed in posts in 3-days (d-1), d, and (d+1). The intensity is computed by weighted sum over relevant posts, where the weights are the semantic similarity between keywords of the headline and content of the posts. The keywords for each headline are extracted from windows of text in the corpus surrounding the URL of the headline by our latent Dirichlet relevance model (Ha-Thuc and Srinivasan). Ranking posts for a headline: The posts are indexed using Lucene. When a URL is found in a post, we extract the surrounding window of text. The size of this window is 800 characters before and after the URL (including HTML). Given a headline (h) these texts are accumulated from all posts containing the URL for h. We then extract keywords from these texts using our latent Dirichlet relevance model (Ha-Thuc and Srinivasan). These keywords are used to query the blog dataset. In the set of returned posts the ones containing the headline URL are boosted to the top. The top 10 are returned as the set of posts to read for headline h.

IowaSBT0904¶

Run ID: IowaSBT0904
Participant: IowaS
Track: Blog
Year: 2009
Submission: 8/27/2009
Type: automatic
Task: topstories
MD5: 04225cba78166df5663ac75c5bb63742
Run description: Ranking headlines for a query date: Given query date d, we rank headlines by posterior probability p(headline| posts), which indicates the importance of the headline in posts in the 3-day duration around d. The probability is computed by p(headline| posts) = p(headline)*p(posts| headline). Prior probability p(headline) is proportional the to number of posts containing URL of the headline. The likelihood is estimated by the similarity between the posts and the headline. Ranking posts for a headline: The posts are indexed using Lucene. When a URL is found in a post, we extract the surrounding window of text. The size of this window is 800 characters before and after the URL (including HTML). Given a headline (h) these texts are accumulated from all posts containing the URL for h. We then extract keywords from these texts using our latent Dirichlet relevance model (Ha-Thuc and Srinivasan). These keywords are used to query the blog dataset. In the set of returned posts the ones containing the headline URL are boosted to the top. The top 10 are returned as the set of posts to read for headline h.

KLEClusPrior¶

Run ID: KLEClusPrior
Participant: POSTECH_KLE
Track: Blog
Year: 2009
Submission: 8/28/2009
Type: automatic
Task: topstories
MD5: 118c78605d8ad8f762212d1467a13617
Run description: The relevance between date query language model and headline language model. Cluster documents to estimate the headline language model. The time information and term weighting of each headline for the prior probability.

KLECluster¶

Run ID: KLECluster
Participant: POSTECH_KLE
Track: Blog
Year: 2009
Submission: 8/28/2009
Type: automatic
Task: topstories
MD5: 8be291210ff30d14b49a4d57cd8cf9a6
Run description: The relevance between date query language model and headline language model. Cluster documents to estimate the headline language model.

KLEFeed¶

Run ID: KLEFeed
Participant: POSTECH_KLE
Track: Blog
Year: 2009
Submission: 8/28/2009
Type: automatic
Task: topstories
MD5: 60605992f71125f9963eb1156dcb25ae
Run description: The relevance between date query language model and headline language model. Using Feed to estimate the headline language model.

KLEFeedPrior¶

Run ID: KLEFeedPrior
Participant: POSTECH_KLE
Track: Blog
Year: 2009
Submission: 8/28/2009
Type: automatic
Task: topstories
MD5: 2599c797fc19822810054c1cc02201c9
Run description: The relevance between date query language model and headline language model. Using Feed to estimate the headline language model. The time information and term weighting of each headline for the prior probability.

nounfull¶

Run ID: nounfull
Participant: knowcenter
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: 54c44fc086ddbd2f914c2bb9d391908c
Run description: We created a Lucene index (search index size 45 GB, approximately 800 000 successfully parsed feeds), and then searched in the Lucene index for the given query terms. We used 2500 blog posts to find the 100 most relevant blogs for the topics. After that, we vectorized the resulting 3800 feeds. For training, we manually labeled 83 blogs into the facet categories. For the faceted distillation, we classified the blogs into the facet categories using a Support Vector Machine based on LibLinear.

OWA¶

Run ID: OWA
Participant: USI
Track: Blog
Year: 2009
Submission: 8/28/2009
Type: automatic
Task: feed
MD5: 6dabc2bdf23c9632786024ab95c19bf9
Run description: In this run, we used OWA (Ordered Weighted Average) operator to combine post relevance scores for a given feed.

pris¶

Run ID: pris
Participant: buptpris___2009
Track: Blog
Year: 2009
Submission: 8/31/2009
Type: automatic
Task: feed
MD5: 76fbed8a63269223690e86e05e527199
Run description: We use indri in the relevance retrieval. Then the Stanford Toolkit is uesd to named entity reconition, ME is used to identify the sentiment orientation.The query expansion words are obtained by machine learning.

prisb¶

Run ID: prisb
Participant: buptpris___2009
Track: Blog
Year: 2009
Submission: 8/31/2009
Type: automatic
Task: feed
MD5: e3048479f0b4232de5a602a44a13f938
Run description: We use indri in the relevance retrieval. Then the Stanford Toolkit is uesd to named entity reconition, ME is used to identify the sentiment orientation.

punctfull¶

Run ID: punctfull
Participant: knowcenter
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: c5b2c55e776ae47162251121c35e8dd2
Run description: We created a Lucene index (search index size 45 GB, approximately 800 000 successfully parsed feeds), and then searched in the Lucene index for the given query terms. We used 2500 blog posts to find the 100 most relevant blogs for the topics. After that, we vectorized the resulting 3800 feeds. For training, we manually labeled 83 blogs into the facet categories. For the faceted distillation, we classified the blogs into the facet categories using a Support Vector Machine based on LibLinear.

RegLDM¶

Run ID: RegLDM
Participant: USI
Track: Blog
Year: 2009
Submission: 8/31/2009
Type: automatic
Task: feed
MD5: 8a39203cc99ddfff1e5fd8c488793aed
Run description: We implemented a version of Large Document Model by concatenating top relevant posts in each blog to make a large document. Then we regularized blog relevance score based in it's similarity to other retrieved blogs.

regularized¶

Run ID: regularized
Participant: USI
Track: Blog
Year: 2009
Submission: 8/28/2009
Type: automatic
Task: feed
MD5: 4f2882443bd0fcb232ef47527ad15bce
Run description: In this run, we do a score regularization step on post relevance scores and generate new scores for posts and then calculate feed score with simple averaging over posts score.

ri1025rw2b¶

Run ID: ri1025rw2b
Participant: shakwat
Track: Blog
Year: 2009
Submission: 8/30/2009
Type: automatic
Task: topstories
MD5: c8949784df3fe753658613259b324376
Run description: For each topic, Random indexing (Sahlgren 2006) was used to built a semantic space containing the blog posts and the headlines in a window around the date of the topic. This geometric representation of meaning of the episodes (posts and headlines) was then crawled using a Random-walk-like algorithm to find the closest posts for each headline. The ranking of the headlines take into account the number of steps needed to find n relevant posts for a headline, together with the density of posts around the headline, and also the average similarity between each headline and the associated posts. For each headline, the posts are ranked w.r.t their similarity with the headline.

ri1025rw5432¶

Run ID: ri1025rw5432
Participant: shakwat
Track: Blog
Year: 2009
Submission: 8/31/2009
Type: automatic
Task: topstories
MD5: 3a2a5b99841fbb2cd0da864269c1751c
Run description: For each topic, Random indexing (Sahlgren 2006) was used to built a semantic space containing the blog posts and the headlines in a window around the date of the topic. This geometric representation of meaning of the episodes (posts and headlines) was then crawled using a Random-walk-like algorithm to find the closest posts for each headline. The ranking of the headlines take into account the number of steps needed to find n relevant posts for a headline, together with the density of posts around the headline, and also the average similarity between each headline and the associated posts. For each headline, the posts are ranked w.r.t their similarity with the headline.

ri1025rw5h2b¶

Run ID: ri1025rw5h2b
Participant: shakwat
Track: Blog
Year: 2009
Submission: 8/31/2009
Type: automatic
Task: topstories
MD5: 6a626b421e9de25134ef1b21478eeea0
Run description: For each topic, Random indexing (Sahlgren 2006) was used to built a semantic space containing the blog posts and the headlines in a window around the date of the topic. This geometric representation of meaning of the episodes (posts and headlines) was then crawled using a Random-walk-like algorithm to find the closest posts for each headline. The ranking of the headlines take into account the number of steps needed to find n relevant posts for a headline, together with the density of posts around the headline, and also the average similarity between each headline and the associated posts. For each headline, the posts are ranked w.r.t their similarity with the headline.

ri2049rw3¶

Run ID: ri2049rw3
Participant: shakwat
Track: Blog
Year: 2009
Submission: 8/30/2009
Type: automatic
Task: topstories
MD5: 7e1ad0b2ccb809b791c86f2ee44661dd
Run description: For each topic, Random indexing (Sahlgren 2006) was used to built a semantic space containing the blog posts and the headlines in a window around the date of the topic. This geometric representation of meaning of the episodes (posts and headlines) was then crawled using a Random-walk-like algorithm to find the closest posts for each headline. The ranking of the headlines take into account the number of steps needed to find n relevant posts for a headline, together with the density of posts around the headline, and also the average similarity between each headline and the associated posts. For each headline, the posts are ranked w.r.t their similarity with the headline.

runtag¶

Run ID: runtag
Participant: USI
Track: Blog
Year: 2009
Submission: 8/31/2009
Type: automatic
Task: topstories
MD5: 326ce9079776b6b01d3eaf55922f9b2b
Run description: Time-aware clustering

sentence¶

Run ID: sentence
Participant: knowcenter
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: ea285f71b79e3c0513be23c87e8b2641
Run description: We created a Lucene index (search index size 45 GB, approximately 800 000 successfully parsed feeds), and then searched in the Lucene index for the given query terms. We used 2500 blog posts to find the 100 most relevant blogs for the topics. After that, we vectorized the resulting 3800 feeds. For training, we manually labeled 83 blogs into the facet categories. For the faceted distillation, we classified the blogs into the facet categories using a Support Vector Machine based on LibLinear.

uogTrFBAlr¶

Run ID: uogTrFBAlr
Participant: uogTr
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: fc51046a8e16041905020bb2bb82f6c9
Run description: Parameter free DFR model, combined with the voting model and a learning to rank approach.

uogTrFBHlr¶

Run ID: uogTrFBHlr
Participant: uogTr
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: a9cf9dd527d32c569205ec51589a07a8
Run description: DFR parameter free model combining voting model, and some features for facets.

uogTrFBMclas¶

Run ID: uogTrFBMclas
Participant: uogTr
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: b99870ef2279c3faea09049a9cdf1fd6
Run description: Parameter free DFR model, combined with the voting model and several classifiers.

uogTrFBNclas¶

Run ID: uogTrFBNclas
Participant: uogTr
Track: Blog
Year: 2009
Submission: 9/1/2009
Type: automatic
Task: feed
MD5: f11f27aac97e1337c0c9b11df278847e
Run description: Parameter free DFR model, combined with the voting model and a classifier.

uogTrTSbmmr¶

Run ID: uogTrTSbmmr
Participant: uogTr
Track: Blog
Year: 2009
Submission: 8/30/2009
Type: automatic
Task: topstories
MD5: 986f5e3fb7bd58e6ad88e9cfd861b65d
Run description: Voting Model, Baseline, MMR.

uogTrTSemmrs¶

Run ID: uogTrTSemmrs
Participant: uogTr
Track: Blog
Year: 2009
Submission: 8/30/2009
Type: automatic
Task: topstories
MD5: 2cdef5b78d866de4bfcbf20f7a2638d3
Run description: Voting Model, Enriched Headlines, MMR, Supporting Documents

uogTrTStimes¶

Run ID: uogTrTStimes
Participant: uogTr
Track: Blog
Year: 2009
Submission: 8/31/2009
Type: automatic
Task: topstories
MD5: d00e52b4b54c20177f70f9a77a4dde0c
Run description: Voting Model, Enriched Headlines, Time Reranking and Supporting Document Ranking

uogTrTSwtime¶

Run ID: uogTrTSwtime
Participant: uogTr
Track: Blog
Year: 2009
Submission: 8/31/2009
Type: automatic
Task: topstories
MD5: 42c08f72d8236ab534abc0c054cd33ec
Run description: Voting Model, 4Day Boosting, Time Diversity