Runs - Relevance Feedback 2008¶

Brown.A1¶

Run ID: Brown.A1
Participant: Brown_University
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: 0e5f5cdcad9fedcd62693e1c27daffdb
Run description: Indri system (language model paradigm), same parameterization as the "indri05AdmfS" run at the Terabyte'05 track except minor change of mu set to 1700 instead of 1500 based on additional training data available here. Other parameters (unchanged) dependence model component weights (0.8, 0.1, 0.1), proximity component smoothing parameter (4000), PRF parameters (N=10, k=50, feedback model weight=0.5).

Brown.A2¶

Run ID: Brown.A2
Participant: Brown_University
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: 3d8527ae4f122eb567cba1992a3e27cf
Run description: Indri system (language model paradigm), Dirichlet smoothing mu=1700.

Brown.B1¶

Run ID: Brown.B1
Participant: Brown_University
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 57f55cfe1f4bab80ddd6d7adb6a90422
Run description: ML estimation of relevance distribution from feedback document, most likely 150 terms comprise feedback unigram model. Mixed with ML query model at feedback model weight 0.3. Using Indri, mixed with sequential dependence model, same paramerization as baseline component weights (0.8, 0.1, 0.1), proximity component smoothing parameter 4000. PRF used with parameterization N=10, k=50, feedback model weight=0.25.

Brown.B2¶

Run ID: Brown.B2
Participant: Brown_University
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 3f053acd23ddf6664da7ddb6eb7ac0ea
Run description: ML estimation of relevance distribution from feedback document, most likely 250 terms comprise feedback unigram model. Mixed with ML query model at feedback model weight 0.3. Run via Indri. No sequential dependency model, no PRF.

Brown.C1¶

Run ID: Brown.C1
Participant: Brown_University
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: 94c5b4d89678d8ab74745e7a93b805e6
Run description: ML estimation of relevance distribution from positive feedback documents; negative feedback documents ignored. Most likely 150 terms comprise feedback unigram model. Mixed with ML query model at feedback model weight 0.45. Using Indri, mixed with sequential dependence model using component weights (0.9, 0.05, 0.05); proximity component smoothing parameter 4000 same as baseline. PRF used with parameterization N=10, k=50, feedback model weight=0.15.

Brown.C2¶

Run ID: Brown.C2
Participant: Brown_University
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: a909b6e8b5c928d8b205a07802331033
Run description: ML estimation of relevance distribution from positive feedback documents; negative feedback documents ignored. Most likely 150 terms comprise feedback unigram model. Mixed with ML query model at feedback model weight 0.45. Run via Indri. No sequential dependency model, no PRF.

Brown.D1¶

Run ID: Brown.D1
Participant: Brown_University
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: 62ddd45b7ae81cbeaaf7056a3bd32632
Run description: Same model used for Brown.C1. ML estimation of relevance distribution from positive feedback documents; negative feedback documents ignored. Most likely 150 terms comprise feedback unigram model. Mixed with ML query model at feedback model weight 0.45. Using Indri, mixed with sequential dependence model using component weights (0.9, 0.05, 0.05); proximity component smoothing parameter 4000 same as baseline. PRF used with parameterization N=10, k=50, feedback model weight=0.15.

Brown.D2¶

Run ID: Brown.D2
Participant: Brown_University
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: c4ee6737e271c8bdc9573d6144db9818
Run description: Same strategy used for Brown.C2. ML estimation of relevance distribution from positive feedback documents; negative feedback documents ignored. Most likely 150 terms comprise feedback unigram model. Mixed with ML query model at feedback model weight 0.45. Run via Indri. No sequential dependency model, no PRF.

Brown.E1¶

Run ID: Brown.E1
Participant: Brown_University
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: 839f96603e29b641df2c0138d6230cc0
Run description: ML estimation of relevance distribution from positive feedback documents; negative feedback documents ignored. Most likely 250 terms comprise feedback unigram model. Mixed with ML query model at feedback model weight 0.8. Dependence model and PRF were not used. Run using Indri.

CMURF08.A1¶

Run ID: CMURF08.A1
Participant: CMU-LTI-DIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: c5305d05a2993b3d30cc9a4295df3e5c
Run description: dependency model initial query, stopwords included in queries, Relevance Model PRF w/ top 10 docs and top 50 terms from initial result. Phrase smoothing mu=4000, term smoothing mu=1700, expanded query (weight 0.3) interpolated with original dependency model query (weight 0.7).

CMURF08.B1¶

Run ID: CMURF08.B1
Participant: CMU-LTI-DIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 6fb278aafc2e0018e4a7630067310da3
Run description: dependency model initial query, stopwords excluded in queries, extended Relevance Model for scoring feedback terms, w/ 0.7 on relevant document and 0.3 on the pseudo relevant top 10 docs (excluding the known relevant), top 50 terms expanded. Phrase smoothing mu=4000, term smoothing mu=1700, expanded query (weight 0.3) interpolated with original dependency model query (weight 0.7).

CMURF08.B2¶

Run ID: CMURF08.B2
Participant: CMU-LTI-DIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 4fe57ebf8c0fbc5c385ff8405e37ce6d
Run description: dependency model initial query, stopwords included in queries, extended Relevance Model for scoring feedback terms, w/ 0.8 on relevant document and 0.2 on the pseudo relevant top 10 docs (excluding the known relevant), top 50 terms expanded. Phrase smoothing mu=4000, term smoothing mu=1700, expanded query (weight 0.3) interpolated with original dependency model query (weight 0.7).

CMURF08.C1¶

Run ID: CMURF08.C1
Participant: CMU-LTI-DIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: ffaaab8970670088f4256d411300595c
Run description: dependency model initial query, stopwords excluded in queries, extended Relevance Model for scoring feedback terms, w/ 0.8 on relevant document and 0.2 on the pseudo relevant top 10 docs (excluding the known relevant), top 50 terms expanded. Phrase smoothing mu=4000, term smoothing mu=1700, expanded query (weight 0.3) interpolated with original dependency model query (weight 0.7).

CMURF08.C2¶

Run ID: CMURF08.C2
Participant: CMU-LTI-DIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: 24b3bd7d3533931608c61e2075950d24
Run description: dependency model initial query, stopwords excluded in queries, extended Relevance Model for scoring feedback terms, w/ 0.8 on relevant document and 0.2 on the pseudo relevant top 12 docs (excluding the known relevant), top 50 terms expanded. Phrase smoothing mu=4000, term smoothing mu=1700, expanded query (weight 0.3) interpolated with original dependency model query (weight 0.7).

CMURF08.D1¶

Run ID: CMURF08.D1
Participant: CMU-LTI-DIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: ebccf077298e031e401dcfa8c129ed31
Run description: dependency model initial query, stopwords excluded in queries, extended Relevance Model for scoring feedback terms, w/ 0.8 on relevant document and 0.2 on the pseudo relevant top 10 docs (excluding the known relevant), top 50 terms expanded. Phrase smoothing mu=4000, term smoothing mu=1700, expanded query (weight 0.3) interpolated with original dependency model query (weight 0.7).

CMURF08.D2¶

Run ID: CMURF08.D2
Participant: CMU-LTI-DIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: 92756bbce68c05ce50f7ecf86e8cdfc7
Run description: dependency model initial query, stopwords excluded in queries, extended Relevance Model for scoring feedback terms, w/ 0.8 on relevant document and 0.2 on the pseudo relevant top 12 docs (excluding the known relevant), top 50 terms expanded. Phrase smoothing mu=4000, term smoothing mu=1700, expanded query (weight 0.3) interpolated with original dependency model query (weight 0.7).

CMURF08.E1¶

Run ID: CMURF08.E1
Participant: CMU-LTI-DIR
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: E
MD5: ee5fd11bbdf9da5cdeeec4e5e2b05d6a
Run description: dependency model initial query, stopwords excluded in queries, extended Relevance Model for scoring feedback terms, w/ 0.8 on relevant document and 0.2 on the pseudo relevant top 10 docs (excluding the known relevant), top 50 terms expanded. Phrase smoothing mu=4000, term smoothing mu=1700, expanded query (weight 0.3) interpolated with original dependency model query (weight 0.7).

DUTIRRF08.A1¶

Run ID: DUTIRRF08.A1
Participant: DUTIR
Track: Relevance Feedback
Year: 2008
Submission: 8/26/2008
Task: A
MD5: 5bf4ead8357247f0bb2c30342f2e1915
Run description: Our baseline, we used the Indri retrieval system for our experiments. This run is only a simple query likelihood run. We didnt make any change to the topics, just the original topics.

DUTIRRF08.B1¶

Run ID: DUTIRRF08.B1
Participant: DUTIR
Track: Relevance Feedback
Year: 2008
Submission: 8/26/2008
Task: B
MD5: fe554f249b15c9d234d14e0939c7f891
Run description: We use the local co-occurrence information in the relevance documents. The window of co-occurrence is a sentence. Top 20 terms are selected to extend the original topics. The proportion of original topic terms to extended terms is 0.8 0.2; the other runs are the same proportion.

DUTIRRF08.C1¶

Run ID: DUTIRRF08.C1
Participant: DUTIR
Track: Relevance Feedback
Year: 2008
Submission: 8/26/2008
Task: C
MD5: 399d8cad97874e4c6cf5f0091e389dbb
Run description: First, we use B method to select 30 candidate terms. Because the non-relate documents are available, we use a Rocchio formula to filter out the useless terms.

DUTIRRF08.D1¶

Run ID: DUTIRRF08.D1
Participant: DUTIR
Track: Relevance Feedback
Year: 2008
Submission: 8/26/2008
Task: D
MD5: aba6d057d6d541dde93ffc2cd9d53536
Run description: The same to C methods.

DUTIRRF08.E1¶

Run ID: DUTIRRF08.E1
Participant: DUTIR
Track: Relevance Feedback
Year: 2008
Submission: 8/26/2008
Task: E
MD5: 05f36183aa3b1cc5711b5ff2dafc1cf4
Run description: The same to C methods. But the extended terms are 70. This run examines the amount of improvement with more relevance info.

FubRF08.A1¶

Run ID: FubRF08.A1
Participant: fub
Track: Relevance Feedback
Year: 2008
Submission: 8/27/2008
Task: A
MD5: c09d04b52b94d31f51c6e692dabd96c4
Run description: Baseline obtained using Terrier with the PL2 weighting model and parameter c=1.0

FubRF08.A2¶

Run ID: FubRF08.A2
Participant: fub
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: 603409b2ad01de02ed39197b3f5adcba
Run description: We used Terrier with a modified query expansion model (Pl2 and DFR Bo1 ) For each topic the query is expanded using - not relevant documents (score 0) - highly relevant documents (score 2) The expansion terms are - 100 terms with a "positive weight" from the highly relevant (score 2) documents. The terms selected have the most divergent frequency from the frequency of the collection - 100 terms with a "negative weight" from the not relevant documents (score 0). The terms selected have the most divergent frequency from the frequency of the collection.

FubRF08.B1¶

Run ID: FubRF08.B1
Participant: fub
Track: Relevance Feedback
Year: 2008
Submission: 8/27/2008
Task: B
MD5: 526508e9bc0708cc754c3d92e6fe37a6
Run description: We used Terrier with a modified query expansion model (Pl2 and DFR Bo1 ) For each topic the query is expanded using - not relevant documents (score 0) - highly relevant documents (score 2) The expansion terms are - 100 terms with a "positive weight" from the highly relevant (score 2) documents The terms selected are with the most divergent frequency from the the frequency of the collection - 30 terms with a "negative weight" from the not relevant documents (score 0). The terms selected are with the most divergent frequency from the the frequency of the collection. The terms choosen should not be contained in the list of the positive weight terms.

FubRF08.C1¶

Run ID: FubRF08.C1
Participant: fub
Track: Relevance Feedback
Year: 2008
Submission: 8/27/2008
Task: C
MD5: 363563746a094e3ba1b699b5bf0336a9
Run description: We used Terrier with a modified query expansion model (Pl2 and DFR Bo1 ) For each topic the query is expanded using - not relevant documents (score 0) - highly relevant documents (score 2) The expansion terms are - 100 terms with a "positive weight" from the highly relevant (score 2) documents. The terms selected have the most divergent frequency from the frequency of the collection - 30 terms with a "negative weight" from the not relevant documents (score 0). The terms selected have the most divergent frequency from the frequency of the collection. The negative terms chosen should not be contained in the list of the positive terms.

FubRF08.C2¶

Run ID: FubRF08.C2
Participant: fub
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: 56b6aa96a74d2ccbbc13650a24c3ee21
Run description: We used Terrier with a modified query expansion model (Pl2 and DFR Bo1 ) For each topic the query is expanded using - not relevant documents (score 0) - highly relevant documents (score 2) The expansion terms are - 100 terms with a "positive weight" from the highly relevant (score 2) documents. The terms selected have the most divergent frequency from the frequency of the collection - 100 terms with a "negative weight" from the not relevant documents (score 0). The terms selected have the most divergent frequency from the frequency of the collection.

FubRF08.D1¶

Run ID: FubRF08.D1
Participant: fub
Track: Relevance Feedback
Year: 2008
Submission: 8/27/2008
Task: D
MD5: 5fe88a55af599e7b98393cf6699249ef
Run description: We used Terrier with a modified query expansion model (Pl2 and DFR Bo1 ) For each topic the query is expanded using - not relevant documents (score 0) - highly relevant documents (score 2) The expansion terms are - 100 terms with a "positive weight" from the highly relevant (score 2) documents. The terms selected have the most divergent frequency from the frequency of the collection - 30 terms with a "negative weight" from the not relevant documents (score 0). The terms selected have the most divergent frequency from the frequency of the collection. The negative terms chosen should not be contained in the list of the positive terms.

FubRF08.D2¶

Run ID: FubRF08.D2
Participant: fub
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: D
MD5: 7ea22821ed15aa8eb24b9e0ef6217efc
Run description: We used Terrier with a modified query expansion model (Pl2 and DFR Bo1 ) For each topic the query is expanded using - not relevant documents (score 0) - highly relevant documents (score 2) The expansion terms are - 100 terms with a "positive weight" from the highly relevant (score 2) documents. The terms selected have the most divergent frequency from the frequency of the collection - 100 terms with a "negative weight" from the not relevant documents (score 0). The terms selected have the most divergent frequency from the frequency of the collection.

FubRF08.E1¶

Run ID: FubRF08.E1
Participant: fub
Track: Relevance Feedback
Year: 2008
Submission: 8/27/2008
Task: E
MD5: 41d933ca9a9c0386939d39306e4c3389
Run description: We used Terrier with a modified query expansion model (Pl2 and DFR Bo1 ) For each topic the query is expanded using - not relevant documents (score 0) - highly relevant documents (score 2) The expansion terms are - 100 terms with a "positive weight" from the highly relevant (score 2) documents. The terms selected have the most divergent frequency from the frequency of the collection - 30 terms with a "negative weight" from the not relevant documents (score 0). The terms selected have the most divergent frequency from the frequency of the collection. The negative terms chosen should not be contained in the list of the positive terms.

FubRF08.E2¶

Run ID: FubRF08.E2
Participant: fub
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: f9e6f3b23338bbd97511c0b32303cc9c
Run description: We used Terrier with a modified query expansion model (Pl2 and DFR Bo1 ) For each topic the query is expanded using - not relevant documents (score 0) - highly relevant documents (score 2) The expansion terms are - 100 terms with a "positive weight" from the highly relevant (score 2) documents. The terms selected have the most divergent frequency from the frequency of the collection - 100 terms with a "negative weight" from the not relevant documents (score 0). The terms selected have the most divergent frequency from the frequency of the collection.

HitRF08.A1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: HitRF08.A1
Participant: HIT2
Track: Relevance Feedback
Year: 2008
Submission: 8/27/2008
Task: A
MD5: d2930ebb86afacd0b8277fd43e38165d
Run description: indri is used to generate the baseline run.

HitRF08.B1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: HitRF08.B1
Participant: HIT2
Track: Relevance Feedback
Year: 2008
Submission: 8/27/2008
Task: B
MD5: 66082ffccbe36eeb67d280e64ef6fbdc
Run description: one relevant document is used by indri.

HitRF08.C1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: HitRF08.C1
Participant: HIT2
Track: Relevance Feedback
Year: 2008
Submission: 8/27/2008
Task: C
MD5: 7d0988411bd2a54c73f90da3e610592c
Run description: three relevant document and three nonrelevant documents are used by indri.

HitRF08.D1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: HitRF08.D1
Participant: HIT2
Track: Relevance Feedback
Year: 2008
Submission: 8/27/2008
Task: D
MD5: 5233760a34aee7cc5a38d75838b7f5d0
Run description: ten judged documents are used by indri.

HitRF08.E1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: HitRF08.E1
Participant: HIT2
Track: Relevance Feedback
Year: 2008
Submission: 8/27/2008
Task: E
MD5: 8f361b248eb4e26b9d0678d60936205c
Run description: large number of judged documents are used by indri.

HKPU.A1¶

Run ID: HKPU.A1
Participant: HKPU
Track: Relevance Feedback
Year: 2008
Submission: 8/27/2008
Task: A
MD5: fe08c8207cc0dedd63b6edad027d03ca
Run description: We used the terabyte track 04, 05 and 06 to calibrate the system. So, we guess it is half-trained by the test topics, although not fully. The system combines the passage scores by Fuzzy disjunction.

HKPU.B1¶

Run ID: HKPU.B1
Participant: HKPU
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: a1605f290f7ba9a81cfd1c1f74837329
Run description: We use a document-context dependent term weight with RF. The system was calibrated by 5 even-numbered queries (802, 804, 806, 808, 810).

HKPU.B2¶

Run ID: HKPU.B2
Participant: HKPU
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: B
MD5: c498fe3f44d9551f4b2af1a65ee6f295
Run description: Used the document-context based retrieval model which is calibrated using 5 odd-numbered queries.

HKPU.C1¶

Run ID: HKPU.C1
Participant: HKPU
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: 0e8f864b6b89b736c28a1d80cff9de33
Run description: Used the document-context based term weights. The system was caliberated using 5 even-numbered queries (802, 804, 806, 808 and 810).

HKPU.C2¶

Run ID: HKPU.C2
Participant: HKPU
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: C
MD5: 148989c43937e7954a85c975a6bf7b3b
Run description: Used the document-context based retrieval model that was trained by 5 odd-numbered queries

HKPU.D1¶

Run ID: HKPU.D1
Participant: HKPU
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: 0a6eea128c3b20cd162a990bd2dce8f4
Run description: We used our document-context based retrieval model that was calibrated by the odd number queries.

HKPU.E1¶

Run ID: HKPU.E1
Participant: HKPU
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: c24038414e0ea1cafbe08e6c67b97d03
Run description: We used 5 odd-numbered queries for training our document-context based retrieval model.

IowaSRF08.A1¶

Run ID: IowaSRF08.A1
Participant: UIowa-Srinivasan
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: 280f60e47b06f5cc78f0267fc9d5fd2f
Run description: The queries were expanded using top ten documents retrieved and no relevance feedback documents. The five terms were determined by their TFIDF scores within the group of retrieved documents. The terms were exponentially discounted and query length discounted (the shorter the query, the less emphasis there is on the RF terms). Querying was done using Lucene, and re-ranking of the results using modified Okapi.

IowaSRF08.B1¶

Run ID: IowaSRF08.B1
Participant: UIowa-Srinivasan
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: b58c81f4308c907a7afb86739aeecc81
Run description: The queries were expanded using top ten documents retrieved and one relevance feedback document. The five terms were determined by their TFIDF scores within the group of the RF documents. The terms were exponentially discounted and query length discounted (the shorter the query, the less emphasis there is on the RF terms). Querying was done using Lucene, and re-ranking of the results using modified Okapi.

IowaSRF08.C1¶

Run ID: IowaSRF08.C1
Participant: UIowa-Srinivasan
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: c769866429b7c6d5649c5a201c9cb762
Run description: The queries were expanded using top ten documents retrieved and three relevance feedback documents. The five terms were determined by their TFIDF scores within the group of the RF documents. The terms were exponentially discounted and query length discounted (the shorter the query, the less emphasis there is on the RF terms). Querying was done using Lucene, and re-ranking of the results using modified Okapi.

IowaSRF08.D1¶

Run ID: IowaSRF08.D1
Participant: UIowa-Srinivasan
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: e3a17253eb5d84faf393a17ac364dd03
Run description: The queries were expanded using top ten documents retrieved and ten relevance feedback documents. The five terms were determined by their TFIDF scores within the group of the RF documents. The terms were exponentially discounted and query length discounted (the shorter the query, the less emphasis there is on the RF terms). Querying was done using Lucene, and re-ranking of the results using modified Okapi.

IowaSRF08.E1¶

Run ID: IowaSRF08.E1
Participant: UIowa-Srinivasan
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: cc51a76a52779d36d3a94ec6bd94aac6
Run description: The queries were expanded using top ten documents retrieved and a large number of relevance feedback documents. The five terms were determined by their TFIDF scores within the group of the RF documents. The terms were exponentially discounted and query length discounted (the shorter the query, the less emphasis there is on the RF terms). Querying was done using Lucene, and re-ranking of the results using modified Okapi.

pris.A1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: pris.A1
Participant: BUPT_pris_
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: 447de53bd9b185740ac9ad6f12991080
Run description: This is the baseline retrieval without relevance information .

pris.B1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: pris.B1
Participant: BUPT_pris_
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 6eeeea2138d75e3d8eae9e21e863ac1d
Run description: This run is B with one relevant document.

pris.B2¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: pris.B2
Participant: BUPT_pris_
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 01950f41add0bf0cde7efc1c6a270964
Run description: This run is B with one relevant document.

pris.C1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: pris.C1
Participant: BUPT_pris_
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: fc2d894892b2f81260fea854ccedb143
Run description: This run is C with three relevant documents and three nonrelevant documents.

pris.C2¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: pris.C2
Participant: BUPT_pris_
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: e05848fbe925a8a71a5e9cbfb5e51674
Run description: This run is C with three relevant documents, three nonrelevant documents.

pris.D1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: pris.D1
Participant: BUPT_pris_
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: bf64d16100675905e2aad7aa78c42307
Run description: This run is D with ten judged documents.

pris.D2¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: pris.D2
Participant: BUPT_pris_
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: 43dc95034ff350522b77d67986cc6d49
Run description: This run is D with ten judged documents.

pris.E1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: pris.E1
Participant: BUPT_pris_
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: 8e43f54121616964b126faf4803ca572
Run description: This run is E with large number of judged documents.

pris.E2¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: pris.E2
Participant: BUPT_pris_
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: 62f814509d90db9a748f0b6d91db2fa4
Run description: This run is E with large number of judged documents.

RMIT08.A1¶

Run ID: RMIT08.A1
Participant: rmit
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: A
MD5: 5a65f733499acd725f84aa52e06c88dd
Run description: Baseline run

RMIT08.B1¶

Run ID: RMIT08.B1
Participant: rmit
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: B
MD5: 9937c69ada0d54955e6e992ced2415a6
Run description: Single relevant document run. Terms are selected based on their TFxIDF score. We have selected 25 top scoring terms per document.

RMIT08.C1¶

Run ID: RMIT08.C1
Participant: rmit
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: C
MD5: 451f98e7e677acb0a372132b8b9feeea
Run description: Multi-document run. For each document a set of top scoring terms (using tfxidf) were identified. Up to 25 terms that occur on more than one document's set of top terms were selected (preference given to terms that occur across most)

RMIT08.C2¶

Run ID: RMIT08.C2
Participant: rmit
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: C
MD5: 7cae10b033210ccd4cdeb09539d943d5
Run description: Multi-document run. All relevant documents were joint to a single large document and the tfxidf score of terms computed based on the large document. The top 25 terms were then selected.

RMIT08.D1¶

Run ID: RMIT08.D1
Participant: rmit
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: D
MD5: 4520b083e2d736ce6cbe982a2d374ddc
Run description: Multi-document run. For each document a set of top scoring terms (using tfxidf) were identified. Up to 25 terms that occur on more than one document's set of top terms were selected (preference given to terms that occur across most). Same as C1 just with more relevant docs to choose from

RMIT08.D2¶

Run ID: RMIT08.D2
Participant: rmit
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: D
MD5: bf4303844d8463dd7f891da3e85823e6
Run description: Multi-document run. All relevant documents were joint to a single large document and the tfxidf score of terms computed based on the large document. The top 25 terms were then selected. Same as C2 just with more relevant docs to choose from.

RMIT08.E1¶

Run ID: RMIT08.E1
Participant: rmit
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: E
MD5: ceac3b631b6be707510f9ee747d3a127
Run description: Multi-document run. For each document a set of top scoring terms (using tfxidf) were identified. Up to 25 terms that occur on more than one document's set of top terms were selected (preference given to terms that occur across most documents). Same as C1 and D1, just with more relevant documents.

RMIT08.E2¶

Run ID: RMIT08.E2
Participant: rmit
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: E
MD5: 5800a1d894b9e9871e928a5c34e29ce0
Run description: Multi-document run. All relevant documents were joint to a single large document and the tfxidf score of terms computed based on the large document. The top 25 terms were then selected. Same as C2 and D2, just with more relevant documents.

SabRF08.A1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: SabRF08.A1
Participant: sabir.buckley
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: 4e994d8b88cb2d30e5e5744090865dc9
Run description: Base SMART Lnu.ltu run. No feedback.

SabRF08.B1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: SabRF08.B1
Participant: sabir.buckley
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 57906fab58d02583a120e2366f6ff5b0
Run description: Rocchio Feedback 16,16,8 one rel doc. Expand by top 50 terms

SabRF08.C1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: SabRF08.C1
Participant: sabir.buckley
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: 495c3dcd4dc3bc0791a4355b3e934a22
Run description: Rocchio Feedback 16,16,8 three rel/nonrel docs. Expand by top 50 terms

SabRF08.D1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: SabRF08.D1
Participant: sabir.buckley
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: ac9984ed02a18fae5c8f3b8555e93800
Run description: Rocchio Feedback a,b,c = 16,16,8 10 judged docs. Expand by top 50 terms

SabRF08.E1¶

Participants | Summary (mtc) | Summary (statAP) | Summary (top10) | Appendix

Run ID: SabRF08.E1
Participant: sabir.buckley
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: 89a8e0d48b54d50ae4eb0c0124451a1a
Run description: Rocchio Feedback a,b,c = 16,16,8 full judgments. Expand by top 50 terms

THUFB.A1¶

Run ID: THUFB.A1
Participant: THUIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: 07c4fcd1fa7a4d9d982129ae49686aa9
Run description: The baseline of all runs, the TMiner retrieval system of THU with BM2500 model.

THUFB.B1¶

Run ID: THUFB.B1
Participant: THUIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 0bd088a3946ea0554d55fa17c8b1f18b
Run description: Query Expansion, result combination of gerenal queries and expansional query.

THUFB.B2¶

Run ID: THUFB.B2
Participant: THUIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 4b8435c82c3f8330390fa17ad5a6eb93
Run description: Result reranking with the nearest neighbor distance reweighting with feedback samples.

THUFB.C1¶

Run ID: THUFB.C1
Participant: THUIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: cf46293c9841939c80f0cece2ba37bcf
Run description: Query Expansion, result combination of gerenal queries and expansional query.

THUFB.C2¶

Run ID: THUFB.C2
Participant: THUIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: 8a1ee63b60ef6968556203ff8311a16a
Run description: Result reranking with the nearest neighbor distance reweighting with feedback samples.

THUFB.D1¶

Run ID: THUFB.D1
Participant: THUIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: b0169ba1d12ddecb4607f7596f0cc5b4
Run description: Query Expansion, result combination of gerenal queries and expansional query,.

THUFB.D2¶

Run ID: THUFB.D2
Participant: THUIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: b65d3777c88355fc3880ffc323e8b3a3
Run description: Result reranking with the nearest neighbor distance reweighting with feedback samples.

THUFB.E1¶

Run ID: THUFB.E1
Participant: THUIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: 21f8e29044996f688b881f6508b444c7
Run description: Query Expansion, result combination of gerenal queries and expansional query.

THUFB.E2¶

Run ID: THUFB.E2
Participant: THUIR
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: c3b5fb9274f5b7782096f58490a4fc33
Run description: Result reranking with the nearest neighbor distance reweighting with feedback samples.

uams08bl.A1¶

Run ID: uams08bl.A1
Participant: UAms_De_Rijke
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: 71b94ff451c3945aef6e42aded670375
Run description: Baseline run out-of-the-box Indri.

uams08bl.A2¶

Run ID: uams08bl.A2
Participant: UAms_De_Rijke
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: 6a54e0c61341ff9695731a5c5b04e243
Run description: Baseline run out-of-the-box Indri

uams08m6.B1¶

Run ID: uams08m6.B1
Participant: UAms_De_Rijke
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 6911532ea3f19595c01795479aeb0211
Run description: Terms for query models are selected based on information from relevant and non relevant documents (if available) and the background collection. Indri is used for final retrieval run.

uams08m6.C1¶

Run ID: uams08m6.C1
Participant: UAms_De_Rijke
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: 40ce5f72bbff3e586fa0e545a2904aa9
Run description: Terms for query models are selected based on information from relevant and non relevant documents (if available) and the background collection. Indri is used for final retrieval run.

uams08m6.D1¶

Run ID: uams08m6.D1
Participant: UAms_De_Rijke
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: 666ce21a5059369982c398f91cad6b51
Run description: Terms for query models are selected based on information from relevant and non relevant documents (if available) and the background collection. Indri is used for final retrieval run.

uams08m6.E1¶

Run ID: uams08m6.E1
Participant: UAms_De_Rijke
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: d36f9c06c6f4fe06335f9fe3fc2aea15
Run description: Terms for query models are selected based on information from relevant and non relevant documents (if available) and the background collection. Indri is used for final retrieval run.

uams08m9.B2¶

Run ID: uams08m9.B2
Participant: UAms_De_Rijke
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 37617e7e2184b15c84759085e129f857
Run description: Terms for query models are selected based on information from relevant documents and the background collection. Information from non relevant documents is not used. Indri is used for final retrieval run.

uams08m9.C2¶

Run ID: uams08m9.C2
Participant: UAms_De_Rijke
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: 527b00270342bd95e55b842e8c93fd2a
Run description: Terms for query models are selected based on information from relevant documents and the background collection. Information from non relevant documents is not used. Indri is used for final retrieval run.

uams08m9.D2¶

Run ID: uams08m9.D2
Participant: UAms_De_Rijke
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: e616a1d33ae97f38a9fb88f01a22d223
Run description: Terms for query models are selected based on information from relevant documents and the background collection. Information from non relevant documents is not used. Indri is used for final retrieval run.

uams08m9.E2¶

Run ID: uams08m9.E2
Participant: UAms_De_Rijke
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: eb91e31afe49eea03323d77cdd4857e1
Run description: Terms for query models are selected based on information from relevant documents and the background collection. Information from non relevant documents is not used. Indri is used for final retrieval run.

UAmsR08CJ.B2¶

Run ID: UAmsR08CJ.B2
Participant: UAmsterdam
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: bc8f29f9f3871a813fcc17ad486e86dc
Run description: Indri Language Model with Parsimonious Relevance Feedback using relevant and non-relevant documents and Pseudo Relevance Feedback (10 documents, 50 terms) and Jelinek-Mercer Smoothing

UAmsR08CJ.C2¶

Run ID: UAmsR08CJ.C2
Participant: UAmsterdam
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: 820a4e1b868b9b2a5832f9ce43763493
Run description: Indri Language Model with Parsimonious Relevance Feedback using relevant and non-relevant documents and Pseudo Relevance Feedback (10 documents, 50 terms) and Jelinek-Mercer Smoothing

UAmsR08CJ.D2¶

Run ID: UAmsR08CJ.D2
Participant: UAmsterdam
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: f063a98085981fdda0ed00925cc85169
Run description: Indri Language Model with Parsimonious Relevance Feedback using relevant and non-relevant documents and Pseudo Relevance Feedback (10 documents, 50 terms) and Jelinek-Mercer Smoothing

UAmsR08CJ.E2¶

Run ID: UAmsR08CJ.E2
Participant: UAmsterdam
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: 5ec38fdead60d6d6d37534ec8205e3ed
Run description: Indri Language Model with Parsimonious Relevance Feedback using relevant and non-relevant documents and Pseudo Relevance Feedback (10 documents, 50 terms) and Jelinek-Mercer Smoothing

UAmsR08PD.A1¶

Run ID: UAmsR08PD.A1
Participant: UAmsterdam
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: ae475d72bdf1a83e919271822a0da4f1
Run description: Indri Language Model with Pseudo Relevance Feedback (10 documents, 50 terms) and Dirichlet Smoothing

UAmsR08PD.B1¶

Run ID: UAmsR08PD.B1
Participant: UAmsterdam
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: d1f668438277c16fa340ac84062969d6
Run description: Indri Language Model with Parsimonious Relevance Feedback and Pseudo Relevance Feedback (10 documents, 50 terms) and Dirichlet Smoothing

UAmsR08PD.C1¶

Run ID: UAmsR08PD.C1
Participant: UAmsterdam
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: 68ed8733a0725edaac8d94a9b7bb13bc
Run description: Indri Language Model with Parsimonious Relevance Feedback and Pseudo Relevance Feedback (10 documents, 50 terms) and Dirichlet Smoothing

UAmsR08PD.D1¶

Run ID: UAmsR08PD.D1
Participant: UAmsterdam
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: D
MD5: 756ca5931a2ef69f19ed2765a89f008e
Run description: Indri Language Model with Parsimonious Relevance Feedback and Pseudo Relevance Feedback (10 documents, 50 terms) and Dirichlet Smoothing

UAmsR08PD.E1¶

Run ID: UAmsR08PD.E1
Participant: UAmsterdam
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: E
MD5: 39f1309d181434037a4482c4b28ac39e
Run description: Indri Language Model with Parsimonious Relevance Feedback and Pseudo Relevance Feedback (10 documents, 50 terms) and Dirichlet Smoothing

UAmsR08PD.F1¶

Run ID: UAmsR08PD.F1
Participant: UAmsterdam
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: F
MD5: 1db9c6cbc4f26107740cded2bf67cbde
Run description: Indri Language Model with Parsimonious Topical Feedback and Pseudo Relevance Feedback (10 documents, 50 terms) and Dirichlet Smoothing

UIUC.A1¶

Run ID: UIUC.A1
Participant: UIUC
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: A
MD5: 2c5011936327018e5654f4557a64f559
Run description: Language Model approach with Dirichlet smoothing method

UIUC.B1¶

Run ID: UIUC.B1
Participant: UIUC
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: B
MD5: 140fc84926633d539210c2bdcc473b5a
Run description: Adaptive Relevance Feedback with Regularization

UIUC.B2¶

Run ID: UIUC.B2
Participant: UIUC
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: B
MD5: 73669ce465355bde25a422ae7ef8ba84
Run description: Adaptive Relevance Feedback

UIUC.C1¶

Run ID: UIUC.C1
Participant: UIUC
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: C
MD5: d62b263e22d351232f28ea818bd4e162
Run description: Adaptive Relevance Feedback with Regularization

UIUC.C2¶

Run ID: UIUC.C2
Participant: UIUC
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: C
MD5: ebbd536325c68ed8f08aeb6c768a596c
Run description: Adaptive Relevance Feedback with Interpolation

UIUC.D1¶

Run ID: UIUC.D1
Participant: UIUC
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: D
MD5: c083993889f6038bfc2987f672888626
Run description: Adaptive Relevance Feedback with Regularization

UIUC.D2¶

Run ID: UIUC.D2
Participant: UIUC
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: D
MD5: 596bada51a1ba37575e16d7201f4cb60
Run description: Pure Adaptive Relevance Feedback without prior knowledge

UIUC.E1¶

Run ID: UIUC.E1
Participant: UIUC
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: E
MD5: 1ee313ec14fc1aac462d208f01fbf067
Run description: Adaptive Relevance Feedback with Regularization

UIUC.E2¶

Run ID: UIUC.E2
Participant: UIUC
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: E
MD5: bbc6ec8a6f2f5438167e510d734cce43
Run description: Adaptive Relevance Feedback plus Pseudo Feedback

uogRF08.A1¶

Run ID: uogRF08.A1
Participant: UoGtr
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: A
MD5: 4135f3e9ff3c0dadf238f35d2e7885b7
Run description: pseudo relevance feedback using surrogates

uogRF08.A2¶

Run ID: uogRF08.A2
Participant: UoGtr
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: A
MD5: c3e88718365abdebd0f18bdbb835ce2e
Run description: pseudo relevance feedback

uogRF08.B1¶

Run ID: uogRF08.B1
Participant: UoGtr
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 04c52a68093fd6fe30ea8e6905de6816
Run description: positive relevance feedback using surrogates

uogRF08.B2¶

Run ID: uogRF08.B2
Participant: UoGtr
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: B
MD5: 982706ae9fc28d1b33884f01c69d626d
Run description: positive relevance feedback

uogRF08.C1¶

Run ID: uogRF08.C1
Participant: UoGtr
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: e2fd0fe5c82d9aeac643bbe1c798224c
Run description: positive relevance feedback using surrogates

uogRF08.C2¶

Run ID: uogRF08.C2
Participant: UoGtr
Track: Relevance Feedback
Year: 2008
Submission: 8/28/2008
Task: C
MD5: 7fcb54ea54a996fa6ed3b6f3e7cfe03c
Run description: positive relevance feedback

uogRF08.D1¶

Run ID: uogRF08.D1
Participant: UoGtr
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: D
MD5: 0691a26df01b9b5c826cb378cf549458
Run description: positive relevance feedback using surrogates

uogRF08.D2¶

Run ID: uogRF08.D2
Participant: UoGtr
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: D
MD5: b9e0532f9905773cb3d9bb38048160b3
Run description: positive relevance feedback

uogRF08.E1¶

Run ID: uogRF08.E1
Participant: UoGtr
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: E
MD5: 9a475222366b96270c88e0d59fb24e8d
Run description: positive relevance feedback using surrogates

uogRF08.E2¶

Run ID: uogRF08.E2
Participant: UoGtr
Track: Relevance Feedback
Year: 2008
Submission: 8/29/2008
Task: E
MD5: 2d86e588012f14bea936314373519b24
Run description: positive relevance feedback