Runs - Enterprise 2007¶

AUTORUN¶

Results | Participants | Input | Summary | Appendix

Run ID: AUTORUN
Participant: iiit-hyderbad
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Automatic query search for experts in specific domain

CSIROdsQfb¶

Run ID: CSIROdsQfb
Participant: csiro-ict.bailey
Track: Enterprise
Year: 2007
Submission: 7/31/2007
Type: feedback
Task: document
Run description: This script takes the query field from the Topic and the pages, and generates the combined result from the clustering and the genre analysis results. The pages are used to improve the ranking. The module used is pfcm_clustering_algo_with_genre_feedback.pm

CSIROdsQnarr¶

Run ID: CSIROdsQnarr
Participant: csiro-ict.bailey
Track: Enterprise
Year: 2007
Submission: 7/31/2007
Type: automatic
Task: document
Run description: This script takes the query field from the Topic and the narrative, and generates the combined result from the clustering and the genre analysis results. The stop words are removed both from the query and the narrative. The module used is pfcm_clustering_algo_with_genre_nrt.pm

CSIROdsQonly¶

Run ID: CSIROdsQonly
Participant: csiro-ict.bailey
Track: Enterprise
Year: 2007
Submission: 7/31/2007
Type: automatic
Task: document
Run description: This script takes the simple query field from the Topic and generates the combined result from the clustering and the genre analysis results. The module used is pfcm_clustering_algo_with_genre.pm, and the results from Algorithm B were chosen.

CSIROdsQsimp¶

Run ID: CSIROdsQsimp
Participant: csiro-ict.bailey
Track: Enterprise
Year: 2007
Submission: 7/31/2007
Type: feedback
Task: document
Run description: This script takes the query field from the Topic and the pages, and generates the combined result from the clustering and the genre analysis results. The pages are used to improve the ranking. The module used is pfcm_clustering_algo_with_genre_feedback.pm. It tries to prioritise similar genres of pages to those already provided as important.

CSIROesQnarr¶

Run ID: CSIROesQnarr
Participant: csiro-ict.bailey
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: expert
Run description: This script takes the query +query + narrative from the Topic and generates the combined result from the 4 algorithms designed for expert search. The four modules used are 1. expert_anchor_term.pm 2. profiles_experts.pm 3. expert_documents_result.pm 4. expert_exactmatch.pm

CSIROesQonly¶

Run ID: CSIROesQonly
Participant: csiro-ict.bailey
Track: Enterprise
Year: 2007
Submission: 7/31/2007
Type: automatic
Task: expert
Run description: This script takes the simple query field from the Topic and generates the combined result from the 4 algorithms designed for expert search. The four modules used are use 1. expert_anchor_term.pm 2. profiles_experts.pm 3. expert_documents_result.pm 4. expert_exactmatch.pm

CSIROesQpage¶

Run ID: CSIROesQpage
Participant: csiro-ict.bailey
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: feedback
Task: expert
Run description: This script takes the query from the topic and the relevant pages given in the topic and finds experts in the relevant pages and tries to map those experts on the employee surrogate pages and tries to find the experts there and ranks the experts. The module used is expert_documents_result_with_pages.pm

CSIROesQprof¶

Run ID: CSIROesQprof
Participant: csiro-ict.bailey
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: expert
Run description: This script takes the query +query expert + narrative and runs it over the profiles index. For example if topic query is genetic modification If topic narrative is what is the current research in CSIRO about genetics The query to the profiles index is genetic modification +genetic +modification expert what current research CSIRO about genetics. The stop words are also removed from the query and the narrative. The module used is profiles_experts_narr.pm

DocRun01¶

Run ID: DocRun01
Participant: cas-liu
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: This result utilizes only the query field. And a simple BM25 model is implemented.

DocRun02¶

Run ID: DocRun02
Participant: cas-liu
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: Both the query field and the narrative field are used. A extended BM25 Model is implemented. URLanchor texttitle are assigned with different weight.

DocRun03¶

Run ID: DocRun03
Participant: cas-liu
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: PageRank is used to rank the document.

DocRun04¶

Run ID: DocRun04
Participant: cas-liu
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: This is the forth one.

DUTDST1¶

Run ID: DUTDST1
Participant: dalianu.yang
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: manual
Task: document
Run description: index indri,porter stemmer,all documents have been cleaned,using BM25 weighting

DUTDST2¶

Run ID: DUTDST2
Participant: dalianu.yang
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: index indri,porter stemmer,all documents have been cleaned,using BM25 weighting

DUTDST3¶

Run ID: DUTDST3
Participant: dalianu.yang
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: feedback
Task: document
Run description: index indri,porter stemmer,all documents have been cleaned,using BM25 weighting

DUTDST4¶

Run ID: DUTDST4
Participant: dalianu.yang
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: index indri,porter stemmer,all documents have been cleaned,using BM25 weighting

DUTEXP1¶

Run ID: DUTEXP1
Participant: dalianu.yang
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: automatic
Task: expert
Run description: Firstly, we build profile for every candidate. Then an index was built by Indri based on the profiles. It's a query-only automatic run, used BM25 to rank the experts and then indri to get support documents.

DUTEXP2¶

Run ID: DUTEXP2
Participant: dalianu.yang
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: manual
Task: expert
Run description: Firstly, we build profile for every candidate. Then an index was built by Indri based on the profiles. use "query" and "narr" fields in topics to form queries. used BM25 to rank the experts and then indri to get support documents.

DUTEXP3¶

Run ID: DUTEXP3
Participant: dalianu.yang
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: manual
Task: expert
Run description: Firstly, we build profile for every candidate. Then an index was built by Indri based on the profiles. use query, narr and page fields in topics to form queries. used BM25 to rank the experts and then indri to get support documents.

DUTEXP4¶

Run ID: DUTEXP4
Participant: dalianu.yang
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: manual
Task: expert
Run description: Firstly, we build profile for every candidate. Then an index was built by Indri based on the profiles. use query and narr fields in topics to form queries. used indri to get support documents and then rank the experts.

ExpertRun01¶

Run ID: ExpertRun01
Participant: cas-liu
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Only query field is used.

ExpertRun02¶

Run ID: ExpertRun02
Participant: cas-liu
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Query field and narrative field are used.

ExpertRun03¶

Run ID: ExpertRun03
Participant: cas-liu
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Based on person profile.

ExpertRun04¶

Run ID: ExpertRun04
Participant: cas-liu
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Based on PageRank.

FDUBase¶

Run ID: FDUBase
Participant: fudanu.niu
Track: Enterprise
Year: 2007
Submission: 7/31/2007
Type: automatic
Task: document
Run description: We use Lemur as the search engine to index and query the topics. We add the score of the document by analyzing the pages which link to this document.

FDUExpan¶

Run ID: FDUExpan
Participant: fudanu.niu
Track: Enterprise
Year: 2007
Submission: 7/31/2007
Type: automatic
Task: document
Run description: By analyzing the narrative field, we automatically choose candidate expansion terms for topics and calculate the frequencies of the terms in the first five documents in the search result by Lemur. We use those high frequency terms as query terms for the second time search. Then we add the score of the document by analyzing the pages which link to this document.

FDUFeedH¶

Run ID: FDUFeedH
Participant: fudanu.niu
Track: Enterprise
Year: 2007
Submission: 7/31/2007
Type: feedback
Task: document
Run description: We use HITS algorithm to judge the quality of the document and regard documents of the same quality as a category. We use FDUBase as the basic result. If the document is in the same category as the documents in the page field, we add the score of the document.

FDUFeedT¶

Run ID: FDUFeedT
Participant: fudanu.niu
Track: Enterprise
Year: 2007
Submission: 7/31/2007
Type: feedback
Task: document
Run description: We use FDUBase as the basic result. Then we analyze the structure of each document and if the document has similar structure as the documents in the page field, we add the score of the document.

FDUGroup¶

Run ID: FDUGroup
Participant: fudanu.niu
Track: Enterprise
Year: 2007
Submission: 8/7/2007
Type: automatic
Task: expert
Run description: We detect email addresses and relevant full names automatically from the corpus. We also filter the candidate list and remain those who are probably contacts on some projects. We divide the corpus to several groups according to the document structure. We give different weights for the different kinds of the documents when calculating each candidate's score on the topic.

FDUn3e7¶

Run ID: FDUn3e7
Participant: fudanu.niu
Track: Enterprise
Year: 2007
Submission: 8/7/2007
Type: automatic
Task: expert
Run description: We detect email addresses and relevant full names automatically from the corpus. We also filter the candidate list and remain those who are probably contacts on some projects. We calculate two scores for each candidate, one for the candidate's name and another for his email. We add the two scores by 30% and 70%.

FDUn5e5¶

Run ID: FDUn5e5
Participant: fudanu.niu
Track: Enterprise
Year: 2007
Submission: 8/7/2007
Type: automatic
Task: expert
Run description: We detect email addresses and relevant full names automatically from the corpus. We also filter the candidate list and remain those who are probably contacts on some projects. We calculate two scores for each candiate, one for the candidate's name and another for his email. We add the two scores by 50% and 50%.

FDUn7e3¶

Run ID: FDUn7e3
Participant: fudanu.niu
Track: Enterprise
Year: 2007
Submission: 8/7/2007
Type: automatic
Task: expert
Run description: We detect email addresses and relevant full names automatically from the corpus. We also filter the candidate list and remain those who are probably contacts on some projects. We calculate two scores for each candidate, one for the candidate's name and another for his email. We add the two scores by 70% and 30%.

feedbackrun¶

Results | Participants | Input | Summary | Appendix

Run ID: feedbackrun
Participant: twenteu.serdyukov
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: feedback
Task: expert
Run description: Modeling expert finding through probabilistic random walk on graph of candidates and documents. The probability of relevance of judged documents is increased and propagated through the graph.

insu1¶

Run ID: insu1
Participant: st.petersburg.nemirovsky
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: document
Run description: Documents are presented as probability distribution over set of words based on tf-profiles and are clustered. Both the query and narrative fields are used. Documents that are considered relevant are sorted according to the distance to the center of the cluster which a document belongs to. No external resources were used.

insu2¶

Run ID: insu2
Participant: st.petersburg.nemirovsky
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: document
Run description: Documents are presented as probability distribution over set of words based on Markov chain theory and are clustered. Both the query and narrative fields are used. Documents that are considered relevant are sorted according to the distance to the center of the cluster which a document belongs to. No external resources were used.

insu3¶

Run ID: insu3
Participant: st.petersburg.nemirovsky
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: document
Run description: Documents are presented as probability distribution over set of words based on tf-profiles and are clustered. The query field is used only. Documents that are considered relevant are sorted according to the distance to the center of the cluster which a document belongs to. No external resources were used.

insu4¶

Run ID: insu4
Participant: st.petersburg.nemirovsky
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: document
Run description: Documents are presented as probability distribution over set of words based on tf-profiles and are clustered. Both the query and narrative fields are used. Documents that are considered relevant are sorted according to PageRank values. No external resources were used.

MANUALRUN¶

Run ID: MANUALRUN
Participant: iiit-hyderbad
Track: Enterprise
Year: 2007
Submission: 8/6/2007
Type: manual
Task: document
Run description: Used the queries from query and description fields for manual run

NARRAUTORUN¶

Run ID: NARRAUTORUN
Participant: iiit-hyderbad
Track: Enterprise
Year: 2007
Submission: 8/6/2007
Type: automatic
Task: document
Run description: Used the query and narrative fields for automatic query and narrative narrative run

ouExNarr¶

Run ID: ouExNarr
Participant: openu.zhu
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: manual
Task: expert
Run description: Multiple windows, query topic, manual tweaking narrative field, anchor texts plus document contents, in-links, and url length.

ouExNarrAu¶

Run ID: ouExNarrAu
Participant: openu.zhu
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Multiple windows, query topic plus narrative, stop word removing from narrative field, partial matching of topic and narrative fields, anchor texts plus document contents, in-links, and url length.

ouExNarrRF¶

Run ID: ouExNarrRF
Participant: openu.zhu
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: manual
Task: expert
Run description: Multiple windows, query topic, manual tweaking narrative field, relevance feedback, anchor texts plus document contents, in-links, and url length.

ouExTitle¶

Run ID: ouExTitle
Participant: openu.zhu
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: automatic
Task: expert
Run description: Multiple windows, query topic only, anchor texts plus document contents, in-links, and url length.

ouNarr¶

Run ID: ouNarr
Participant: openu.zhu
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: manual
Task: document
Run description: Manual tweaking of query from narrative, anchor texts plus document contents, inlinks, and url length.

ouNarrAuto¶

Run ID: ouNarrAuto
Participant: openu.zhu
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: stop word removing from narrative, narrative plus query topic for automatic search, anchor texts plus document contents, inlinks, and url length.

ouNarrRF¶

Run ID: ouNarrRF
Participant: openu.zhu
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: manual
Task: document
Run description: Manual tweaking of query from narrative, relevance feedback, anchor texts plus document contents, inlinks, and url length.

ouTopicOnly¶

Run ID: ouTopicOnly
Participant: openu.zhu
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: query topic only, anchor texts plus document contents, in-links, and url length.

PRISDF¶

Results | Participants | Input | Summary | Appendix

Run ID: PRISDF
Participant: beijingu-posts-tele.weiran
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Use expert df to improve team leader finding.

PRISEM¶

Results | Participants | Input | Summary | Appendix

Run ID: PRISEM
Participant: beijingu-posts-tele.weiran
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: use email counts for identify team leaders

PRISLM¶

Results | Participants | Input | Summary | Appendix

Run ID: PRISLM
Participant: beijingu-posts-tele.weiran
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Using languague model and window-based expert profile

PRISRR¶

Results | Participants | Input | Summary | Appendix

Run ID: PRISRR
Participant: beijingu-posts-tele.weiran
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: use rerank method

qorw¶

Results | Participants | Input | Summary | Appendix

Run ID: qorw
Participant: twenteu.serdyukov
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Modeling expert finding through probabilistic random walk on graph of candidates and documents

qorwkstep¶

Results | Participants | Input | Summary | Appendix

Run ID: qorwkstep
Participant: twenteu.serdyukov
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Modeling expert finding through probabilistic random walk on graph of candidates and documents. The walking process is restricted to K-step walk. K =5.

qorwnewlinks¶

Results | Participants | Input | Summary | Appendix

Run ID: qorwnewlinks
Participant: twenteu.serdyukov
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Modeling expert finding through probabilistic random walk on graph of candidates and documents. Here, we use extracted links among documents (hyperlinks) and among candidates (through subdomain in their email addresses).

QRYBASICRUN¶

Run ID: QRYBASICRUN
Participant: iiit-hyderbad
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: Basic engine, very restrictive, precision focus, low recall. Uses only query.

QUERYRUN¶

Run ID: QUERYRUN
Participant: iiit-hyderbad
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: Basic engine, somewhat liberal, precision focus with higher recall. Uses only query.

RmitQ¶

Run ID: RmitQ
Participant: rmitu.scholer
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: In this base run, we use words from topic field as a query, index contents from original documents, and rank documents according to KL divergence.

RmitQAnc¶

Run ID: RmitQAnc
Participant: rmitu.scholer
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: In this run, the content of an indexed document is extended to include its anchor texts. As in our base run, we still use words from topic field as a query and rank documents according to KL divergence.

RmitQAncIndg¶

Run ID: RmitQAncIndg
Participant: rmitu.scholer
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: In this run, the content of an indexed document is extended to include its anchor text. The weight of a document is a combination of content weight (KL divergence) and in-degree.

RmitQFir¶

Run ID: RmitQFir
Participant: rmitu.scholer
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: In this run, we divide the collection into sub-collections. All documents in a sub-collection have the same domain name. Each sub-collection is indexed separately and documents are retrieved and ranked in a federated manner.

SJTUEntDS01¶

Run ID: SJTUEntDS01
Participant: sjtu-apex-ent
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: document
Run description: query Phase Query, Proximity Query, Ordinary Query, Query Expansion from Narrative Field, Query of terms with low document frequency BM25 on Title, Anchor, H1, H2, Keywords, Extracted Body Positional Model Added Hostrank used for Re-ranking

SJTUEntDS02¶

Run ID: SJTUEntDS02
Participant: sjtu-apex-ent
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: document
Run description: Basically the same with SJTUEntDS01. Different weights for fields from SJTUEntDS01 in combination. Reshuffled phrase query adopted.

SJTUEntDS03¶

Run ID: SJTUEntDS03
Participant: sjtu-apex-ent
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: Basically the same with SJTUEntDS01. Use document frequency to boost each query word.

SJTUEntDS04¶

Run ID: SJTUEntDS04
Participant: sjtu-apex-ent
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: feedback
Task: document
Run description: Basically the same with SJTUEntDS01. Use words from the given key pages as candidates for query expension.

SJTUEntES01¶

Run ID: SJTUEntES01
Participant: sjtu-apex-ent
Track: Enterprise
Year: 2007
Submission: 8/8/2007
Type: automatic
Task: expert
Run description: Parsing email appearences for Expert Names. Filter out names containing single word or host names. Query Phrase query, Bigram query, Proximity query, Single word query, Stemming query, Expended query. Model Window based model (different size), Section model, Title-author model, Body-author model, Tree model, Reversed Tree model. Add-in Feature Position Refine, Bigram query refine, Dom-structure refine Name-Musk Fullname match, email match, potential name match.

SJTUEntES02¶

Run ID: SJTUEntES02
Participant: sjtu-apex-ent
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: automatic
Task: expert
Run description: Basically the same with SJTUEntES01. Topic sensitive Expert Rank is added.

SJTUEntES03¶

Run ID: SJTUEntES03
Participant: sjtu-apex-ent
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: automatic
Task: expert
Run description: Basically the same with SJTUEntES01. Homepage refine is added.

SJTUEntES04¶

Run ID: SJTUEntES04
Participant: sjtu-apex-ent
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Basically the same with SJTUEntES01. Topic sensitive Expert Rank and Homepage refine are added.

THUDSANCHOR¶

Run ID: THUDSANCHOR
Participant: tsinghuau.zhang
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: document
Run description: Instead of the full text, we just use the inlink anchor text of web pages as collection for retrieval

THUDSFULLSR¶

Run ID: THUDSFULLSR
Participant: tsinghuau.zhang
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: document
Run description: We retrieve documents from the full text collection in the first stage. Then we apply link analysis among the retieved documents.

THUDSSEL¶

Run ID: THUDSSEL
Participant: tsinghuau.zhang
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: document
Run description: Instead of the full text, we used those documents with outlink number more than 100 as collection. In each documents, we also enlarge the weight of the inlink anchor text part.

THUDSSELSR¶

Run ID: THUDSSELSR
Participant: tsinghuau.zhang
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: document
Run description: Instead of the full text, we used those documents with outlink number more than 100 as collection. In each documents, we also enlarge the weight of the inlink anchor text part. Plus, we apply link analysis among the retieved documents.

THUIRMPDD2¶

Run ID: THUIRMPDD2
Participant: tsinghuau.zhang
Track: Enterprise
Year: 2007
Submission: 8/8/2007
Type: automatic
Task: expert
Run description: The result list is a linear combination from two kinds of PDD(Person Description Documents), including original PDD and precisely extracted PDD, .

THUIRMPDD4¶

Run ID: THUIRMPDD4
Participant: tsinghuau.zhang
Track: Enterprise
Year: 2007
Submission: 8/8/2007
Type: automatic
Task: expert
Run description: The result list is a linear combination from four kinds of PDD (Person Description Documents), including original PDD, precisely extracted PDD, PDD only based on anchor text and PDD built from those documents with outlinks more than 100.

THUIRPDD2¶

Run ID: THUIRPDD2
Participant: tsinghuau.zhang
Track: Enterprise
Year: 2007
Submission: 8/8/2007
Type: automatic
Task: expert
Run description: The result list is from PDD(Person Description Documents)

THUIRPDD2C40¶

Run ID: THUIRPDD2C40
Participant: tsinghuau.zhang
Track: Enterprise
Year: 2007
Submission: 8/8/2007
Type: automatic
Task: expert
Run description: The result list from PDD with the window of content is 40 character

UALR07Ent1¶

Run ID: UALR07Ent1
Participant: uarkansas-littlerock.bayrak
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: automatic
Task: document
Run description: Our title-only baseline run with pseudo feedback from the top 5 result documents

UALR07Ent2¶

Run ID: UALR07Ent2
Participant: uarkansas-littlerock.bayrak
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: feedback
Task: document
Run description: We used MMRSummApp which is bundled with Lemur toolkit to extract summaries from the authoritative pages of each query. Next, we find search results for these summary keywords (excluding common words) and rerank UALR07Ent1 to include search results of summary keywords with higher weight.

UALR07Ent3¶

Run ID: UALR07Ent3
Participant: uarkansas-littlerock.bayrak
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: feedback
Task: document
Run description: We use a 'word difference' approach. Compare all words from authoritative pages with the words from the top 5 search results of each query. The question is /what is unique in authoritative pages that is not present in the top search results. The net difference in the words will give us unique words (excluding common words) which we use to obtain results for. We rerank UALR07Ent1 using these results.

UALR07Ent4¶

Run ID: UALR07Ent4
Participant: uarkansas-littlerock.bayrak
Track: Enterprise
Year: 2007
Submission: 8/2/2007
Type: manual
Task: document
Run description: We looked at authoritative pages that were extracted using the script. Next, we identified (manually) the keywords from each of these authoritative pages. We rerank UALR07Ent1 according to manual keyword selection method.

UALR07Exp1¶

Run ID: UALR07Exp1
Participant: uarkansas-littlerock.bayrak
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: We used our document search run UALR07Ent3 to identify potential experts for each query

UALR07Exp2¶

Run ID: UALR07Exp2
Participant: uarkansas-littlerock.bayrak
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: We used our document search run UALR07Ent3 to identify potential experts for each query only from the available 1000 results for each query

UALR07Exp3¶

Run ID: UALR07Exp3
Participant: uarkansas-littlerock.bayrak
Track: Enterprise
Year: 2007
Submission: 8/11/2007
Type: manual
Task: expert
Run description: Uses manual verified experts

uams07bfb¶

Run ID: uams07bfb
Participant: uamsterdam.deRijke
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: The MixtureModel constructs the document model based on five components (indices) and a background component. The components used are title, headers, metadata, anchors and body; the background components consists of all 5 components combined. The mixture for this run is based on tests on the TRECweb collection title has a weight of 0.30, headers 0.10, metadata 0.05, anchors 0.40 and body 0.10. The background model has a weight of 0.05. As document prior we used the log of the number of inlinks per document. The query model is constructed using blind feedback on the top 10 documents of the original query; new weights are assigned to the new and original query terms and the query is issued again.

uams07bfbex¶

Run ID: uams07bfbex
Participant: uamsterdam.deRijke
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: feedback
Task: document
Run description: The MixtureModel constructs the document model based on five components (indices) and a background component. The components used are title, headers, metadata, anchors and body; the background components consists of all 5 components combined. The mixture for this run is based on tests on the TRECweb collection title has a weight of 0.30, headers 0.10, metadata 0.05, anchors 0.40 and body 0.10. The background model has a weight of 0.05. As document prior we used the log of the number of inlinks per document. The query model is constructed using a combination of blind feedback on the top 10 documents and using the example documents. New weights are assigned to the new and original query terms and the query is issued again.

uams07bl¶

Run ID: uams07bl
Participant: uamsterdam.deRijke
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: The MixtureModel constructs the document model based on five components (indices) and a background component. The components used are title, headers, metadata, anchors and body; the background components consists of all 5 components combined. This run assigns equal weights to all components (0.18) and 0.10 to the background model. No document priors are used. The query model is constructed using the number of occurences of a term in the query (e.g. a two-word query assign 0.5 weight to each word).

uams07exbl¶

Run ID: uams07exbl
Participant: uamsterdam.deRijke
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: baseline run document-centric language model, naive document-candidate associations

uams07exfr¶

Run ID: uams07exfr
Participant: uamsterdam.deRijke
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: frequency-based document-candidate associations. mixture model for document retrieval.

uams07exmm¶

Run ID: uams07exmm
Participant: uamsterdam.deRijke
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: document-centric, using mixture model for document retrieval

uams07expp¶

Run ID: uams07expp
Participant: uamsterdam.deRijke
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: combination of document-centric and candidate-centric lm approach. model 1 builds a lm from personal pages of candidates, model 2 is the same as in run uams07exmm. The two models are combined using equal weights.

uams07pr¶

Run ID: uams07pr
Participant: uamsterdam.deRijke
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: The MixtureModel constructs the document model based on five components (indices) and a background component. The components used are title, headers, metadata, anchors and body; the background components consists of all 5 components combined. The mixture for this run is based on tests on the TRECweb collection title has a weight of 0.30, headers 0.10, metadata 0.05, anchors 0.40 and body 0.10. The background model has a weight of 0.05. As document prior we used the log of the number of inlinks per document. The query model is constructed using the number of occurences of a term in the query (e.g. a two-word query assign 0.5 weight to each word).

uiowa07entD1¶

Run ID: uiowa07entD1
Participant: uiowa.srinivasan
Track: Enterprise
Year: 2007
Submission: 8/4/2007
Type: automatic
Task: document
Run description: + High recall search boolean search on meta field (title, subject, and keywords) and document content. + Retrieval feedback

uiowa07entD2¶

Run ID: uiowa07entD2
Participant: uiowa.srinivasan
Track: Enterprise
Year: 2007
Submission: 8/4/2007
Type: automatic
Task: document
Run description: High recall search boolean search on meta fields (title, subject, and keywords) and document content

uiowa07entD3¶

Run ID: uiowa07entD3
Participant: uiowa.srinivasan
Track: Enterprise
Year: 2007
Submission: 8/4/2007
Type: feedback
Task: document
Run description: High precision search (phrase search), and relevance feedback

uiowa07entD4¶

Run ID: uiowa07entD4
Participant: uiowa.srinivasan
Track: Enterprise
Year: 2007
Submission: 8/4/2007
Type: feedback
Task: document
Run description: High recall search (boolean search), and relevance feedback

uiowa07entE1¶

Run ID: uiowa07entE1
Participant: uiowa.srinivasan
Track: Enterprise
Year: 2007
Submission: 8/13/2007
Type: feedback
Task: expert
Run description: First, created profiles for topics using top 25 documents (top 75 for topic CE-25) retrieved for expanded queries created from original query terms and high frequency metadata terms from relevant documents (within field). A profile consists of weighted stemmed terms. Terms with higher weight better characterize the topic. Second, identified all PERSON named-entities from these documents that map to a csiro.au email address and created profiles for these experts. Third, ranked experts for each topic based on cosine similarity of expert profile with topic profile. Fourth, ranked expert documents using original topic query + expansion terms from relevant documents as query.

uiowa07entE2¶

Run ID: uiowa07entE2
Participant: uiowa.srinivasan
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: feedback
Task: expert
Run description: Probabilistic Topic Modelling

uogEDSCLCDIS¶

Run ID: uogEDSCLCDIS
Participant: glasgow.ounis
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: feedback
Task: document
Run description: linking structure evidence of feedback documents is used in this run

uogEDSComPri¶

Run ID: uogEDSComPri
Participant: glasgow.ounis
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: feedback
Task: document
Run description: mixed linking structure evidence is used in this run

uogEDSF¶

Run ID: uogEDSF
Participant: glasgow.ounis
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: A simple automatic run uses document structure information

uogEDSINLPRI¶

Run ID: uogEDSINLPRI
Participant: glasgow.ounis
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: feedback
Task: document
Run description: linking structure evidence is used in this run

uogEXFeMNZcP¶

Run ID: uogEXFeMNZcP
Participant: glasgow.ounis
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Document structure, voting technique, proximity (2).

uogEXFeMNZdQ¶

Run ID: uogEXFeMNZdQ
Participant: glasgow.ounis
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Document structure, voting technique, QE (2)

uogEXFeMNZP¶

Run ID: uogEXFeMNZP
Participant: glasgow.ounis
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Document structure, voting technique, proximity.

uogEXFeMNZQE¶

Run ID: uogEXFeMNZQE
Participant: glasgow.ounis
Track: Enterprise
Year: 2007
Submission: 8/10/2007
Type: automatic
Task: expert
Run description: Document structure, voting technique, QE

uwKLD¶

Run ID: uwKLD
Participant: uwaterloo-olga
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: feedback run (automatic) using kld

uwRF¶

Run ID: uwRF
Participant: uwaterloo-olga
Track: Enterprise
Year: 2007
Submission: 8/4/2007
Type: feedback
Task: document
Run description: Using pages to extract expansion terms.

uwtbase¶

Run ID: uwtbase
Participant: uwaterloo-olga
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: query only run

WHU10¶

Run ID: WHU10
Participant: wuhanu.lu
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: automatic
Task: expert
Run description: Using cocument combined method for this run, and the document cutoff is set to 10

WHU15¶

Run ID: WHU15
Participant: wuhanu.lu
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: automatic
Task: expert
Run description: Using cocument combined method for this run, and the document cutoff is set to 15

WHUC5¶

Run ID: WHUC5
Participant: wuhanu.lu
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: automatic
Task: expert
Run description: Using common method for this run, and the document cutoff is set to 5

WHUE10¶

Run ID: WHUE10
Participant: wuhanu.lu
Track: Enterprise
Year: 2007
Submission: 8/9/2007
Type: automatic
Task: expert
Run description: Using common method for this run, and the expert relevant document cutoff is set to 10

york07ed1¶

Run ID: york07ed1
Participant: yorku.huang
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: 1. Use db1 jtxt. 2. No query expansion. 3. Use only terms extracted from the raw topic for retrieval.

york07ed2¶

Run ID: york07ed2
Participant: yorku.huang
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: 1. Use db1 ltxt. 2. No query expansion. 3. Use only terms extracted from the raw topic for retrieval.

york07ed3¶

Run ID: york07ed3
Participant: yorku.huang
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: 1. Use db2 ltxt. 2. Expand query terms with "narrative" from raw topics.

york07ed4¶

Run ID: york07ed4
Participant: yorku.huang
Track: Enterprise
Year: 2007
Submission: 8/3/2007
Type: automatic
Task: document
Run description: 1. Use db1 jtxt. 2. Expand query terms with "narrative" from raw topics.

zslrun¶

Results | Participants | Input | Summary | Appendix

Run ID: zslrun
Participant: pekingu.zhou
Track: Enterprise
Year: 2007
Submission: 8/13/2007
Type: automatic
Task: expert
Run description: THIS RUN IS MISSING A DESCRIPTION, PLEASE PROVIDE ONE.