Skip to content

Runs - Web 2009

arsc09web

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: arsc09web
  • Participant: ARSC09
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 05b2798c343aadd818c5d710b78a6dd9
  • Run description: The ARSC team used the Google N-Gram Corpus to create a vocabularies of 1-gram, 2-gram, and 3-gram phrase tokens. For the official judged run, a fraction of the Category B collection was indexed using the fixed vocabulary of 1-gram terms--a "go list" rather than a stop list--and searched with a custom index/search application built with the Lucene toolkit. Multiple Sun Fire X4600 nodes, each with 64 GB of RAM, from the ARSC midnight cluster were used for the index and search tasks.

ICTNETADRun3

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: ICTNETADRun3
  • Participant: ICTNET
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: f503aee0854f321a102743de0f28858c
  • Run description: using improved BM25 model with search extension.

ICTNETADRun4

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: ICTNETADRun4
  • Participant: ICTNET
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 28ad749f5f1aa7d3d726094bdc28f253
  • Run description: using improved BM25 model.

ICTNETADRun5

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: ICTNETADRun5
  • Participant: ICTNET
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 5808d4d7d5b31c415dae9580415a14f3
  • Run description: using improved BM25 model and optimizing parameters. please use this run, because we can't replace the former run which PRECEDENCE is 2.

ICTNETDivR1

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: ICTNETDivR1
  • Participant: ICTNET
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: d124d75a8bf95ea5605109f965a5c257
  • Run description: do query expansion by using top k results returned by Google search engine. the weighting model is KL

ICTNETDivR2

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: ICTNETDivR2
  • Participant: ICTNET
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 614393fc25e5f5b3f5a4d0d2f47d4ba6
  • Run description: 1.clustering top N search results form google search engine 2.getting the key words for each subtopic from the clustered texts 3.getting docs by searching document colletion with the inherent key words 4.ranking those docs using a greedy algorithm

ICTNETDivR3

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: ICTNETDivR3
  • Participant: ICTNET
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 88d19930162ad3fd04e8db6af1fd5aed
  • Run description: Using K-MEANS algorithm to make different clusters.

IE09

Results | Participants | Input | Summary | Appendix

  • Run ID: IE09
  • Participant: York
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: cf45ab9065c2ff794fdf655baf10a088
  • Run description: This run is based on using number of inlinks to a page in ranking and also other factors.

irra1a

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: irra1a
  • Participant: IRRA
  • Track: Web
  • Year: 2009
  • Submission: 8/13/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 8f643e2ae813e546bdf2e341554921db
  • Run description: This is the base IRRA run for adhoc task.

irra1d

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: irra1d
  • Participant: IRRA
  • Track: Web
  • Year: 2009
  • Submission: 8/13/2009
  • Type: automatic
  • Task: diversity
  • MD5: 6a6027791df65d4ce844a1766f1a4d34
  • Run description: This run is the base IRRA run for diversity task.

irra2a

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: irra2a
  • Participant: IRRA
  • Track: Web
  • Year: 2009
  • Submission: 8/13/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 25769e901f7e12e25d5370088ad59909
  • Run description: This run is a variation of the base IRRA run for adhoc task.

irra2d

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: irra2d
  • Participant: IRRA
  • Track: Web
  • Year: 2009
  • Submission: 8/13/2009
  • Type: automatic
  • Task: diversity
  • MD5: f53964da7c9fdfb8b1d53de0ef400de7
  • Run description: This run is a variation of the base IRRA run for diversity task.

irra3a

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: irra3a
  • Participant: IRRA
  • Track: Web
  • Year: 2009
  • Submission: 8/13/2009
  • Type: automatic
  • Task: adhoc
  • MD5: e158aec0de79b78eb706f6519bebe9fd
  • Run description: This run is another variation of the base IRRA run for adhoc task.

irra3d

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: irra3d
  • Participant: IRRA
  • Track: Web
  • Year: 2009
  • Submission: 8/13/2009
  • Type: automatic
  • Task: diversity
  • MD5: 1ac189040b7f7f5f9391176a29b78d29
  • Run description: This run is another variation of the base IRRA run for diversity task.

MS1

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: MS1
  • Participant: msrc
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 980d4b0dec45220a6085c5e3957c7e46
  • Run description: We train our system using some training queries. For these queries we use top documents retrieved by bing as relevance judgments.

MS2

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: MS2
  • Participant: msrc
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: eabef716a85fa12ba1641931a546068b
  • Run description: This is the same as the first run with some extra features: We train our system using some training queries. For these queries we use top documents retrieved by bing as relevance judgments.

MSDiv1

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: MSDiv1
  • Participant: msrc
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 1dc8301ce6d90939e03f7f5d1ee6e2ba
  • Run description: We train our system using some training queries. For these queries we use top documents retrieved by bing as relevance judgments.

MSDiv2

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: MSDiv2
  • Participant: msrc
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: f9bc6c4cbb46b0f889d7b1c585023c16
  • Run description: This is the same as the first run with some extra features. We train our system using some training queries. For these queries we use top documents retrieved by bing as relevance judgments.

MSDiv3

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: MSDiv3
  • Participant: msrc
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: diversity
  • MD5: 8f78b9cb8d2d0c641425dccd6ef3b7ab
  • Run description: Extremely simple ranker. Host-collapse limits the number of results from each host.

MSRAACSF

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: MSRAACSF
  • Participant: MSRAsia
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: diversity
  • MD5: f558e65107c5a75d746392e77e9e7cb1
  • Run description: diversify results by using clustering (C), sites(S), and anchors(A) only results from main query are reserved(F).

MSRAAF

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: MSRAAF
  • Participant: MSRAsia
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 09772edf9b8cb524672413b47cc62740
  • Run description: baseline + anchor-based rerank

MSRABASE

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: MSRABASE
  • Participant: MSRAsia
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: diversity
  • MD5: c7ecd0f5c5067e8271c5dfd5621df961
  • Run description: baseline: msra2000

MSRAC

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: MSRAC
  • Participant: MSRAsia
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 0049bfcabe7085b3677b7b5f07ef4a45
  • Run description: baseline + clustering-based diversity

MSRACS

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: MSRACS
  • Participant: MSRAsia
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: diversity
  • MD5: 7e26bce094542715b3d503cf42d8dc3e
  • Run description: diversify results by using clustering and sites

MSRANORM

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: MSRANORM
  • Participant: MSRAsia
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: adhoc
  • MD5: e09ed3c89c785f3c58c788c510671cdc
  • Run description: a baseline which uses msra2000 (without term prox in anchor.)

muadanchor

Results | Participants | Input | Summary | Appendix

  • Run ID: muadanchor
  • Participant: unimelb
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: a9cdac46a31e3c09d06c6eebb2f2a8ec
  • Run description: standard impact score on incoming anchor text

muadibm5

Results | Participants | Input | Summary | Appendix

  • Run ID: muadibm5
  • Participant: unimelb
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: bacb296229507ecf0dde1daccb53441c
  • Run description: BM25 on impact score, content only.

muadimp

Results | Participants | Input | Summary | Appendix

  • Run ID: muadimp
  • Participant: unimelb
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 366afc0d6451af761ba21df12230c469
  • Run description: standard impact score, content only.

mudvibm5

Results | Participants | Input | Summary | Appendix

  • Run ID: mudvibm5
  • Participant: unimelb
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: diversity
  • MD5: 3e0528836cf74bdca70b4694c07d68d1
  • Run description: bm25 on content impact

mudvimp

Results | Participants | Input | Summary | Appendix

  • Run ID: mudvimp
  • Participant: unimelb
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 3cdc05cb031401625b57b759ae9c2571
  • Run description: standard impact score on incoming anchor text

NeuDiv1

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: NeuDiv1
  • Participant: NEU
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 46c885114ff3faa108e2122b625e3040
  • Run description: - run Indri to get top 2000 docs - run a simple spam filter based on term freq, statistical match with the expected Zipf distribution on English text, eliminate spam - tag the documents with dictionary-based tags (developed by Neu) - greedily pick documents (out of ones with roughly same Indri score) that maximize a diversity utility against the already retrieved set of documents

NeuDivW75

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: NeuDivW75
  • Participant: NEU
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: diversity
  • MD5: 95555522db738ceeb9d695c434fb1a01
  • Run description: - run Indri to get 2000 docs per query - run a simple statistical spam filter - tag documents using dictionaries - mechanism for diversity based on the (universal tags): repeatedly, out of 75 docs with consecutive Indri scores, select the one that maximizes tag-diversity utility against the set already retrieved.

NeuLMWeb300

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: NeuLMWeb300
  • Participant: NEU
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 93b3c3abf5a23e24c2a60a722a7e66a9
  • Run description: - run Indri to get 2000 docs (for each query) - run a spam filter based on Expected Zipfian distribution entropy for English text after some statistical analysis - Use a threshold of 300 for the spamming value. - Output top remaining 1000 docs.

NeuLMWeb600

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: NeuLMWeb600
  • Participant: NEU
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: e4ecbf1411c911cbe559303e2aa623c6
  • Run description: - run Indri to get 2000 docs (for each query) - run a spam filter based on Expected Zipfian distribution entropy for English text after some statistical analysis - Use a threshold of 600 for the spamming value. - Output top remaining 1000 docs.

NeuLMWebBase

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: NeuLMWebBase
  • Participant: NEU
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 8243ee682215c50ded7f056c72c54b52
  • Run description: - run Indri to get 1000 docs (for each query)

NeuRRWeb300

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: NeuRRWeb300
  • Participant: NEU
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 55397c2afc00eccf94a3851ea9f91e9b
  • Run description: - run Indri to get 2000 docs (for each query) - run a spam filter based on Expected Zipfian distribution entropy for English text after some statistical analysis - Use a threshold of 300 for the spamming value. - Identification of Entity Tags - Run k-means using cosine similarity between document tags - Rank the documents within the cluster by indri - Fuse the clusters by round-robin in a single list.

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: pkuLink
  • Participant: pku2009
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: db86e0db3a391ad46a15f64d75abf797
  • Run description: In this run , we use the html page's struct and link infos. We just remove the html tags and index whole collection on 23 machines with poster stemming and remove stop words. Anchor text and link info we use hadoop to compute. When searching,we use multi-field bm25 formual , term proximity info and link infos to rank results,then we merge the results from 23 machines to generate finally result.

pkuSewmTp

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: pkuSewmTp
  • Participant: pku2009
  • Track: Web
  • Year: 2009
  • Submission: 8/13/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 6847607851242d9d448a35440a089a36
  • Run description: In this run , we only use the html page's content info. We just remove the html tags and index whole collection on 23 machines with poster stemming and remove stop words. When searching,we use bm25 formual and term proximity info to rank results,then we merge the results from 23 machines to generate finally result.

pkuStruct

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: pkuStruct
  • Participant: pku2009
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 0c303db646306a2c551b677c36fac763
  • Run description: In this run , we use the html page's struct info,likes title,anchor, h1. We just remove the html tags and index whole collection on 23 machines with poster stemming and remove stop words. When searching,we use multi-field bm25 formual and term proximity info to rank results,then we merge the results from 23 machines to generate finally result.

RmitDiv

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: RmitDiv
  • Participant: RMIT
  • Track: Web
  • Year: 2009
  • Submission: 8/15/2009
  • Type: automatic
  • Task: diversity
  • MD5: fd2b4f80b42f291a353876af1bd98821
  • Run description: Language Model run allowing a single (top ranked) result from each unique domain.

RmitLm

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: RmitLm
  • Participant: RMIT
  • Track: Web
  • Year: 2009
  • Submission: 8/15/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 8dabce8406ac9242d75a3a46806ffe3d
  • Run description: Baseline Zettair run using Language Modelling

RmitOkapi

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: RmitOkapi
  • Participant: RMIT
  • Track: Web
  • Year: 2009
  • Submission: 8/15/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 6d2965ba3bbf7d5e29839a7b1809d280
  • Run description: Baseline Zettair run using Okapi BM25

Sab9wtBase

Results | Participants | Input | Summary | Appendix

  • Run ID: Sab9wtBase
  • Participant: SABIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: f5975c0a7159ea662eb17cbe07f0231f
  • Run description: Base SMART ltu.Lnu no expansion run.

Sab9wtBDiv1

Results | Participants | Input | Summary | Appendix

  • Run ID: Sab9wtBDiv1
  • Participant: SABIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 3425eda6cba3483e915c90ad4ba641ef
  • Run description: Initial base SMART run ltu.Lnu. rerank top 10-30 docs by starting at rank 2 and choosing the lowest max inner product to higher ranked docs already added from candidate docs. candidate docs here are those with maximum rank MIN (4*i,30) where i is number already added. Stop reranking after 10 docs added.

Sab9wtBDiv2

Results | Participants | Input | Summary | Appendix

  • Run ID: Sab9wtBDiv2
  • Participant: SABIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 4ec028e297e76603197ac70e4b11d5b9
  • Run description: Initial base SMART run ltu.Lnu. rerank top 20-70 docs by starting at rank 2 and choosing the lowest max inner product to higher ranked docs already added from candidate docs. candidate docs here are those with maximum rank MIN (5*i,70) where i is number already added. Stop reranking after 20 docs added. Docs weighted Ltu.

Sab9wtBf1

Results | Participants | Input | Summary | Appendix

  • Run ID: Sab9wtBf1
  • Participant: SABIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: b55f7ba2ceffabf7cca81df1ebc04d1f
  • Run description: Blind feed back afer SMART ltu.Lnu base run, 25 docs, added 20 terms, weighted Rocchio collwt a,b,c = 32,64,128. Lnu docs, all docs in collection considered nonrel.

Sab9wtBf2

Results | Participants | Input | Summary | Appendix

  • Run ID: Sab9wtBf2
  • Participant: SABIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 317d32011dc361723f8b3dbfe0f66397
  • Run description: Blind feed back after SMART ltu.Lnu base run, 25 docs, added 20 terms, weighted Rocchio a,b,c = 32,8,0. Ltu docs, no nonrel docs.

Sab9wtBfDiv

Results | Participants | Input | Summary | Appendix

  • Run ID: Sab9wtBfDiv
  • Participant: SABIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 369b9f2e377efe9d7e984dacfb1b19e4
  • Run description: Blind feed back afer SMART ltu.Lnu base run, 25 docs, added 20 terms, weighted Rocchio collwt a,b,c = 32,64,128. Lnu docs, all docs in collection considered nonrel. rerank top 20-70 docs by starting at rank 2 and choosing the lowest max inner product to higher ranked docs already added from candidate docs. candidate docs here are those with maximum rank MIN (5*i,70) where i is number already added. Stop reranking after 20 docs added. Docs weighted Ltu.

scutrun1

Results | Participants | Input | Summary | Appendix

  • Run ID: scutrun1
  • Participant: scut_kapok
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 6d93d34c6c0d12de205c1e1a3745096f
  • Run description: the first result of South China University of Technology:scutrun1

scutrun2

Results | Participants | Input | Summary | Appendix

  • Run ID: scutrun2
  • Participant: scut_kapok
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 6bfd7e5bc10d41b699e03a7fc3d09561
  • Run description: the 2nd result of South China University of Technology:scutrun2

scutrun3

Results | Participants | Input | Summary | Appendix

  • Run ID: scutrun3
  • Participant: scut_kapok
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 1e5f12b26335f421fc138b2464848d04
  • Run description: the third result of South China University of Technology:scutrun3

SIEL09

Results | Participants | Input | Summary | Appendix

  • Run ID: SIEL09
  • Participant: SIEL
  • Track: Web
  • Year: 2009
  • Submission: 8/18/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 027ca07d066870f1f90e8edefa567ed9
  • Run description: just simple adhoc retrieval

spc

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: spc
  • Participant: UAms
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: c255887bd5b0f5923d38ebe63b37d899
  • Run description: This run uses single pass clustering to rerank the documents from an initial run with mrf+wikipedia non-article filtering

THUIR09AbClu

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: THUIR09AbClu
  • Participant: THUIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: e6cbaad6398c8bf4ff1a3a8fc12ecece
  • Run description: Baseline + web page abstract extraction + result clustering + diversity document selection with site-based duplicate elimination. Baseline: Improved probabilistic model with PageRank ranking and wordpair model. Retrieve on the combined full text and in-link anchor. Web page spam filtering. multi-field search.

THUIR09An

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: THUIR09An
  • Participant: THUIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: acb7ea691c7436b82733e40704eca9d5
  • Run description: Improved probabilistic model with PageRank ranking and wordpair model. Retrieve only on the in-link anchor text. Web page spam filtering. multi-field search.

THUIR09FuClu

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: THUIR09FuClu
  • Participant: THUIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 1cc0e5c28544b663847089a167a869d5
  • Run description: Baseline + result clustering + diversity document selection with site-based duplicate elimination. Baseline: Improved probabilistic model with PageRank ranking and wordpair model. Retrieve on the combined full text and in-link anchor. Web page spam filtering. multi-field search.

THUIR09LuTA

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: THUIR09LuTA
  • Participant: THUIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: bbe1e6c2c2a33e7c7d28da0bb341acdf
  • Run description: Lucene with BM25 ranking. Retrieve on both full text and in-link anchor. Result re-ranking with PageRank. Web page spam filtering. multi-field search.

THUIR09QeDiv

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: THUIR09QeDiv
  • Participant: THUIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 8648f32fec1474f2ae38ed604a5acd58
  • Run description: Retrieval with Query expansion based on Google query recommendation + diversity document selection + site and content based duplicate elimination.

THUIR09TxAn

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: THUIR09TxAn
  • Participant: THUIR
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 2d1859b130b046829ca1d28b7e3981e6
  • Run description: Improved probabilistic model with PageRank ranking and wordpair model. Retrieve on the combined full text and in-link anchor. Web page spam filtering. multi-field search.

tm

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: tm
  • Participant: UAms
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 2e5043e7409006dbc52835bade5432b7
  • Run description: topic model + re-ranking with top 2500 docs

twCSodpRBB

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: twCSodpRBB
  • Participant: utwente
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 49545612bc61705e084985833298d0b1
  • Run description: run also relies on ODP categories; no merging of results is performed, simple round robin over the different corpora

twCSodpRNB

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: twCSodpRNB
  • Participant: utwente
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: a81f241d251afe6c48115774b72f58a7
  • Run description: 100 category terms from the ODP directory were used for diversification. Each query was run 100 times together with each query term - the retrieval score of the top document containing all terms (query terms + category term) was recorded and the categories of the top 5 scoring test queries were used for retrieval of the final result list. Spam detection relies on the UK Webspam corpus.

twCSrs9N

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: twCSrs9N
  • Participant: utwente
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 1e2b8dba652d1f52463f8f3c1ebeaf9d
  • Run description: Around 30 retrieval methods were created and run on each subcorpus (ClueWeb_English_1 to ClueWeb_English_10). The performance of each method was estimated/guessed by an automatic method and the best estimated results per subcorpus were used during merging (ZMUV merging). Basic spam detection was implemented, relying on content-only features. For training purposes of the spam decision tree, part of the training data from the UK Webspam collection (barcelona.research.yahoo.net/webspam/) was used.

twCSrsR

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: twCSrsR
  • Participant: utwente
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: c2de41f182a88baa54abe0ebaf3d9805
  • Run description: Very similar to run twCSrs9N. Differences: variation of the automatic estimation method and no normalization in the merging process, instead simple round robin procedure to gather the results of each subcorpus (ClueWeb_English_1 to ClueWeb_English_10).

twJ48rsU

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: twJ48rsU
  • Participant: utwente
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 9350da15080401b28e1994f5ca19f727
  • Run description: less agressive spam detection method used; merging without normalization.

UamsAw7an3

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UamsAw7an3
  • Participant: Amsterdam
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: ba1e99a81be49f50c5089e143219c057
  • Run description: Merge of text run with document length prior and anchor text run with anchor text length prior, weight 0.7 text + 0.3 anchor

UamsAwebQE10

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UamsAwebQE10
  • Participant: Amsterdam
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 6e2cea64c436c350dc15ac5e469e6166
  • Run description: Merge of 10 runs based on 10 distinct expanded queries using the 10 top result of a baseline run

UamsDancTFb1

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UamsDancTFb1
  • Participant: Amsterdam
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 0d442129b07f69d0b4fd5216ca20b078
  • Run description: Run on anchor text index with length prior on anchor text length, top down filtered on the contribution of unseen words.

UamsDwebLFou

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UamsDwebLFou
  • Participant: Amsterdam
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 4a9054bc4893132a24c23e53bed639fd
  • Run description: Run on text index with document length prior, top down filtered on the contribution of unseen outgoing links.

UamsDwQE10TF

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UamsDwQE10TF
  • Participant: Amsterdam
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 542479a8088967e4efd87846a7cad003
  • Run description: Merge of 10 runs on text index based on 10 distinct expanded queries, with a document length prior, top down filtered on the contribution of unseen words.

UCDSIFTdiv

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UCDSIFTdiv
  • Participant: CSIUCD
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 51e7e297b7174ab6311b9f58ec72ac8f
  • Run description: Data fusion using SlideFuse algorithm on input topfiles from the Terrier IR system.

UCDSIFTinter

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UCDSIFTinter
  • Participant: CSIUCD
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 5c4f0a104562ab957dc34532085efccb
  • Run description: Fusion using the InterFuse algorithm and input topfiles from the Terrier IR system.

UCDSIFTprob

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UCDSIFTprob
  • Participant: CSIUCD
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: c187b45d1affe2e2db55cead44f9be9e
  • Run description: Fusion using the ProbFuse algorithm and input topfiles from the Terrier IR system.

UCDSIFTslide

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UCDSIFTslide
  • Participant: CSIUCD
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: e797e5de2dc643c6b9122f948136965b
  • Run description: Fusion using the SlideFuse algorithm and input topfiles from the Terrier IR system.

udelFMRM

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: udelFMRM
  • Participant: UDel
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 9a9eed61f000ad90450cbe3daf3db3b2
  • Run description: Carterette & Chandar facet models constructed from document relevance models.

udelFMWG

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: udelFMWG
  • Participant: UDel
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: a76077d6381ccba18b92b1ea7700b99a
  • Run description: Carterette & Chandar facet models constructed from HITS-style web graph of links among top 200 docs retrieved by basic indri query.

udelIndDMRM

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: udelIndDMRM
  • Participant: UDel
  • Track: Web
  • Year: 2009
  • Submission: 8/18/2009
  • Type: automatic
  • Task: adhoc
  • MD5: fca748cfdb71a6308739288f4a7baa34
  • Run description: indri run with Metzler & Croft dependence models and pseudo-relevance feedback with Lavrenko & Croft relevance models. parameters trained in a semi-supervised fashion using last year's Million Query data.

udelIndDRPR

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: udelIndDRPR
  • Participant: UDel
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 1e1ec7882c6f2b921f858c1c0cb11e9e
  • Run description: indri run with dependence models, relevance models, and pagerank; parameters trained semi-supervisedly using last year's MQ data

udelIndDRSP

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: udelIndDRSP
  • Participant: UDel
  • Track: Web
  • Year: 2009
  • Submission: 8/18/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 3e813551652ac0272a080c79f62d7fca
  • Run description: indri run with Metzler & Croft dependence models and pseudo-relevance feedback with Lavrenko & Croft relevance models, with a "domain trust" document prior. parameters trained in a semi-supervised fashion using last year's Million Query data. "domain trust" calculated by looking at presence of domain on publicly-available URL and sendmail whitelists and blacklists.

udelSimPrune

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: udelSimPrune
  • Participant: UDel
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: ea2ba4ff5c7f4b624b4d278eb049977d
  • Run description: greedy pruning based on document-document similarities among top 200 retrieved by standard indri query.

UDWAxBL

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UDWAxBL
  • Participant: EceUdel
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: a361ca7d721fb0010acac8dab1c46ba8
  • Run description: Axiomatic method

UDWAxQE

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UDWAxQE
  • Participant: EceUdel
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: b96c66bd10bfe8081ccfe0a03a83624b
  • Run description: Axiomatic method and query expansion.

UDWAxQEWeb

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UDWAxQEWeb
  • Participant: EceUdel
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 8736e65ca662834a963569428b5d2b22
  • Run description: Axiomatic method and query expansion with Google snippet.

UMHOObm25B

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UMHOObm25B
  • Participant: UMD
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: bc3de21d3c5c3bc85d6d60a8eb5b6a6d
  • Run description: Ivory run (collaboration between UMD and Yahoo)

UMHOObm25GS

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UMHOObm25GS
  • Participant: UMD
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 4421f1a4b376a954f0606be44d2143f6
  • Run description: Ivory run (collaboration between UMD and Yahoo)

UMHOObm25IF

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UMHOObm25IF
  • Participant: UMD
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: bd2bf8045d045c373c0431b8d5f52c88
  • Run description: Ivory run (collaboration between UMD and Yahoo)

UMHOOqlB

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UMHOOqlB
  • Participant: UMD
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: d56a2c95432239349502e2d0cb09c1f5
  • Run description: Ivory run (collaboration between UMD and Yahoo)

UMHOOqlGS

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UMHOOqlGS
  • Participant: UMD
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 14b50c4de6adeca67e3da39e7630a2e1
  • Run description: Ivory run (collaboration between UMD and Yahoo)

UMHOOqlIF

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UMHOOqlIF
  • Participant: UMD
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 15d460020f697486fac1a19c79de1f91
  • Run description: Ivory run (collaboration between UMD and Yahoo)

UMHOOsd

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UMHOOsd
  • Participant: UMD
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 5530d1ca3c7a5364fcfc5c80737027e5
  • Run description: Ivory run (collaboration between UMD and Yahoo)

UMHOOsdp

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: UMHOOsdp
  • Participant: UMD
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: e3365f611c00b4e4c21ab1e609d58850
  • Run description: Ivory run (collaboration between UMD and Yahoo)

uogTrDPCQcdB

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: uogTrDPCQcdB
  • Participant: uogTr
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: diversity
  • MD5: 50c550cd2331b61d9a3caf2f3516d9dc
  • Run description: Community diversification framework combining subqueries from cluster-based diversification mechanism.

uogTrdphA

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: uogTrdphA
  • Participant: uogTr
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: e72af22ff40dc136ea873d311de68b4d
  • Run description: Parameter free DFR model with anchor text.

uogTrdphCEwP

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: uogTrdphCEwP
  • Participant: uogTr
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: d404aae435c6b528815c03112093adcc
  • Run description: Parameter free DFR model with collection enrichment and some proximity.

uogTrdphP

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: uogTrdphP
  • Participant: uogTr
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 7b79605029ea9d11bd7a1e6bc1e1c694
  • Run description: Parameter free DFR model with some proximity.

uogTrDYCcsB

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: uogTrDYCcsB
  • Participant: uogTr
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: diversity
  • MD5: b22e8ef0e3b13934d58721b920b26f44
  • Run description: Resource diversification framework combining subqueries from external SE.

uogTrDYScdA

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: uogTrDYScdA
  • Participant: uogTr
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 463edba86f8911e96f049d4fc110b24e
  • Run description: Community diversification framework combining subqueries from external SE.

uvaaol

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: uvaaol
  • Participant: UAms
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: d6dabe0fc42da8ade7d2d158db59bbbd
  • Run description: generated multiple query variants using a query log.

uvaee

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: uvaee
  • Participant: UAms
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: e4512552c68547ce0c3cdc195a08e0e8
  • Run description: Pseudo relevance feedback on wikipedia part of Clueweb.

uvamrf

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: uvamrf
  • Participant: UAms
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: a928865901cd2465d7b8fddf1390b55d
  • Run description: markov random field, wikipedia non-article pages removed

uvamrftop

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: uvamrftop
  • Participant: UAms
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 350ae84eb062d6cf78ff90c05cab1d48
  • Run description: markov random field, wikipedia non-article pages removed, Wikipedia pages moved to the top of the rankings

uwgym

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: uwgym
  • Participant: Waterloo
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 6fa401d018480fd66eaa6b348dd31fb0
  • Run description: This is a run based on commercial search engines.

watd1

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: watd1
  • Participant: Waterloo
  • Track: Web
  • Year: 2009
  • Submission: 8/16/2009
  • Type: automatic
  • Task: diversity
  • MD5: b3cf6bcbe1af003e5eb2ebb1883385f6
  • Run description: Discrimination based on the maximum likelihood term in each message.

watd3

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: watd3
  • Participant: Waterloo
  • Track: Web
  • Year: 2009
  • Submission: 8/16/2009
  • Type: automatic
  • Task: diversity
  • MD5: 1f34b6739e41f969fad327bf9ae30906
  • Run description: Discrimination based on the 3 maximum likelihood terms in each message.

watd5

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: watd5
  • Participant: Waterloo
  • Track: Web
  • Year: 2009
  • Submission: 8/16/2009
  • Type: automatic
  • Task: diversity
  • MD5: e1062061b7ff0d8722a33e32cc329436
  • Run description: Discrimination based on the 5 maximum likelihood terms in each message.

watprf

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: watprf
  • Participant: Waterloo
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 257c69a5993685502e78aaa3baaa36e5
  • Run description: Sequential logistic regression used to retrieve 10,000 docs/topic from Wikipedia and another 10,000 docs/topic from the English Clueweb09. Pseudo-relevance feedback based on variable number of documents (by likelihood) used to select the best 1,000 from the 20,000 retrieved.

watrrfw

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: watrrfw
  • Participant: Waterloo
  • Track: Web
  • Year: 2009
  • Submission: 8/16/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 97ce48389aac8cede3e34022368f85e3
  • Run description: Used sequential logistic regression (no indexing). Reciprocal rank fusion of wikipedia-only (watwp) and two relevance feedback runs (watrrf and WAT2.base [relevance feeback submission]).

WatSdmrm3

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: WatSdmrm3
  • Participant: UWaterlooMDS
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 1f42b8baa3f320f4841bb8c0fb96f47a
  • Run description: Dependence models with a relevance models (rm3) feeback component using only the collection.

WatSdmrm3we

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: WatSdmrm3we
  • Participant: UWaterlooMDS
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: f61e3bad2dca3c48cf6a065e182bd1c9
  • Run description: Dependence models with a relevance models (rm3) feeback component derived from the top results of Microsoft and Yahoo search.

WatSklfb

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: WatSklfb
  • Participant: UWaterlooMDS
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 58eb6ac6075d9e5499a514712a17ae7c
  • Run description: We first obtain the 25 highest scoring point-wise kl-div stems from the 100 most frequent stems in the top 25 documents retrieved by a query likelihood retrieval ranking of the documents containing all of the query terms. We exclude the query stems and 418 stopwords from being part of this set. Using this set, we perform a blind feedback ranking of the documents that contain all query stems.

WatSklfu

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: WatSklfu
  • Participant: UWaterlooMDS
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 161e8336f90579fbb513345403e571f7
  • Run description: We first obtain the 25 highest scoring point-wise kl-div stems from the 100 most frequent stems in the top 25 documents retrieved by a query likelihood retrieval ranking of the documents containing all of the query terms. We exclude the query stems and 418 stopwords from being part of this set. Using this set, we generate 25 additional retrievals. We then use a variant of reciprocal rank fusion to join these lists.

WatSklq

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: WatSklq
  • Participant: UWaterlooMDS
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: 8655e0d3cdab3869a36bbf1f65145964
  • Run description: We first obtain the 25 highest scoring point-wise kl-div stems from the 100 most frequent stems in the top 25 documents retrieved by a query likelihood retrieval ranking of the documents containing all of the query terms. We exclude the query stems and 418 stopwords from being part of this set. Using this set, we generate 25 additional retrievals. We then select between these lists using a priority queue like mechanism.

WatSql

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: WatSql
  • Participant: UWaterlooMDS
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: adhoc
  • MD5: cb1237e12828b95629249c334333d322
  • Run description: Simple query likelihood retrieval given the query with Dirichlet prior smoothing.

watwp

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: watwp
  • Participant: Waterloo
  • Track: Web
  • Year: 2009
  • Submission: 8/16/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 27c7002b6a120a2b34a7c48c803119f5
  • Run description: Wikipedia only run. Used sequential logistic regression (no indexing).

wume1

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: wume1
  • Participant: LU_WUME
  • Track: Web
  • Year: 2009
  • Submission: 8/19/2009
  • Type: automatic
  • Task: diversity
  • MD5: d42d414bcf2f8b495f303f91099c9f37
  • Run description: We get query expansion from google insight and then combine the search result from different queries expansions.

wume2

Results | Participants | Proceedings | Input | Summary | Appendix

  • Run ID: wume2
  • Participant: LU_WUME
  • Track: Web
  • Year: 2009
  • Submission: 8/20/2009
  • Type: automatic
  • Task: diversity
  • MD5: c0c906e32891d576646a5fd9ba655386
  • Run description: A variation of BM25

yhooumd09BFM

Results | Participants | Input | Summary | Appendix

  • Run ID: yhooumd09BFM
  • Participant: yahoo
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 9b65a1b586b8d6d340569debd8534dc6
  • Run description: BM25 run using Ivory (w/ independent fusion). Results are post-filtered using Yahoo!'s "industrial strength" adult and spam classifiers. Retrieval scores are also biased towards high quality pages, which are also uses the output of a proprietary Y! classifier. A moderate amount of filtering and reranking is done.

yhooumd09BGC

Results | Participants | Input | Summary | Appendix

  • Run ID: yhooumd09BGC
  • Participant: yahoo
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: 3f53b4924905c5a7cc23163582638320
  • Run description: BM25 run using Ivory (w/ global statistics). Results are post-filtered using Yahoo!'s "industrial strength" adult and spam classifiers. Retrieval scores are also biased towards high quality pages, which are also uses the output of a proprietary Y! classifier. A conservative amount of filtering and reranking is done.

yhooumd09BGM

Results | Participants | Input | Summary | Appendix

  • Run ID: yhooumd09BGM
  • Participant: yahoo
  • Track: Web
  • Year: 2009
  • Submission: 8/17/2009
  • Type: automatic
  • Task: adhoc
  • MD5: a24963a0d373d8e1f29ea3b5d458a65f
  • Run description: BM25 run using Ivory (w/ global statistics). Results are post-filtered using Yahoo!'s "industrial strength" adult and spam classifiers. Retrieval scores are also biased towards high quality pages, which are also uses the output of a proprietary Y! classifier. A moderate amount of filtering and reranking is done.