/QPP-Overview

Summary of query performance prediction results

QPP-Overview

This page provides an overview of pre/post-retrieval QPP results on common corpora, retrieval models and evaluation metrics. The page is ordered by corpus. The list is by no means complete and will be updated from time to time.

It is limited to correlation-coefficient based results reported on WT10g, GOV2 and ClueWeb09. Predictors in bold are pre-retrieval predictors.

WT10g

Reference Queries Retrieval model Predictor Linear correlation coefficient Spearman's Rho Kendall's Tau
[2] 451-500 LM (Dirichlet) MaxIDF 0.090 0.302 0.227
[2] 451-500 LM (Dirichlet) MaxSCQ 0.447 0.624 0.456
[2] 451-500 LM (Dirichlet) MaxVAR 0.354 0.529 0.358
[2] 501-550 LM (Dirichlet) MaxIDF 0.507 0.408 0.291
[2] 501-550 LM (Dirichlet) MaxSCQ 0.418 0.378 0.274
[2] 501-550 LM (Dirichlet) MaxVAR 0.538 0.502 0.372
[3] 451-550 BM25 NCQ 0.342
[3] 451-550 BM25 Query Clarity 0.358
[3] 451-550 BM25 Sigma-50% 0.447
[4] 451-550 LM (Dirichlet) Clarity 0.348
[4] 451-550 LM (Dirichlet) UEF(Clarity) 0.483
[4] 451-550 LM (Dirichlet) WIG 0.376
[4] 451-550 LM (Dirichlet) UEF(WIG) 0.453
[4] 451-550 LM (Dirichlet) NQC 0.488
[4] 451-550 LM (Dirichlet) UEF(NQC) 0.522
[4] 451-550 LM (Dirichlet) QF 0.426
[4] 451-550 LM (Dirichlet) UEF(QF) 0.526
[5] 451-550 LM (Dirichlet) Clarity 0.432
[5] 451-550 LM (Dirichlet) Clarity+MaxVARTF.IDF 0.450
[5] 451-550 LM (Dirichlet) QF 0.483
[5] 451-550 LM (Dirichlet) QF+MaxVARTF.IDF 0.496
[5] 451-550 LM (Dirichlet) WIG 0.376
[5] 451-550 LM (Dirichlet) WIG+MaxVARTF.IDF 0.431
[7] 451-550 LM (Dirichlet) Delta(QR1)/Delta(QG) 0.260
[9] 451-550 LM (Dirichlet) MaxIDF 0.300 0.280
[9] 451-550 LM (Dirichlet) Simplified Clarity 0.130 0.170
[9] 451-550 LM (Dirichlet) MaxSCQ 0.410 0.340
[9] 451-550 LM (Dirichlet) MaxVAR 0.420 0.330
[10] 451-550 LM (Dirichlet) MaxVAR 0.406
[10] 451-550 LM (Dirichlet) NQC 0.405
[10] 451-550 LM (Dirichlet) UEF(NQC) 0.435
[10] 451-550 LM (Dirichlet) UEF(Improved Clarity) 0.563
[10] 451-550 LM (Dirichlet) LTRoq 0.346

GOV2

Reference Queries Retrieval model Predictor Linear correlation coefficient Spearman's Rho Kendall's Tau
[1] 701-800 Markov random field Clarity 0.333
[1] 701-800 Markov random field Query Feedback 0.480
[1] 701-800 Markov random field Weighted Information Gain 0.574
[1] 701-800 Markov random field Robustness 0.317
[2] 701-850 LM (Dirichlet) MaxIDF 0.297 0.326 0.219
[2] 701-850 LM (Dirichlet) MaxSCQ 0.357 0.347 0.231
[2] 701-850 LM (Dirichlet) MaxVAR 0.359 0.369 0.247
[4] 701-850 LM (Dirichlet) Clarity 0.433
[4] 701-850 LM (Dirichlet) UEF(Clarity) 0.462
[4] 701-850 LM (Dirichlet) WIG 0.479
[4] 701-850 LM (Dirichlet) UEF(WIG) 0.458
[4] 701-850 LM (Dirichlet) NQC 0.360
[4] 701-850 LM (Dirichlet) UEF(NQC) 0.393
[4] 701-850 LM (Dirichlet) QF 0.476
[4] 701-850 LM (Dirichlet) UEF(QF) 0.491
[5] 701-850 LM (Dirichlet) Clarity 0.456
[5] 701-850 LM (Dirichlet) Clarity+MaxVARTF.IDF 0.447
[5] 701-850 LM (Dirichlet) QF 0.566
[5] 701-850 LM (Dirichlet) QF+MaxVARTF.IDF 0.574
[5] 701-850 LM (Dirichlet) WIG 0.486
[5] 701-850 LM (Dirichlet) WIG+MaxVARTF.IDF 0.465
[6] 701-750 Language model Clarity 0.139
[6] 751-800 Language model Clarity 0.171
[6] 701-750 Language model Robustness 0.150
[6] 751-800 Language model Robustness 0.208
[6] 701-750 Language model Autocorrelation 0.454
[6] 751-800 Language model Autocorrelation 0.383
[7] 701-850 LM (Dirichlet) Delta(R2G)/Delta(R1G) 0.204
[8] 701-750 LM (Dirichlet) Clarity 0.305 0.134
[8] 701-750 LM (Dirichlet) Robustness 0.341 0.213
[8] 701-750 LM (Dirichlet) Clarity+Robustness 0.374 0.226
[8] 701-750 LM (Dirichlet) Clarity 0.206 0.171
[8] 701-750 LM (Dirichlet) Robustness 0.301 0.208
[8] 701-750 LM (Dirichlet) Clarity+Robustness 0.362 0.252
[9] 701-850 LM (Dirichlet) MaxIDF 0.350 0.250
[9] 701-850 LM (Dirichlet) Simplified Clarity 0.260 0.200
[9] 701-850 LM (Dirichlet) MaxSCQ 0.420 0.280
[9] 701-850 LM (Dirichlet) MaxVAR 0.430 0.290
[10] 451-550 LM (Dirichlet) MaxVAR 0.384
[10] 451-550 LM (Dirichlet) NQC 0.336
[10] 451-550 LM (Dirichlet) UEF(NQC) 0.417
[10] 451-550 LM (Dirichlet) UEF(Improved Clarity) 0.537
[10] 451-550 LM (Dirichlet) LTRoq 0.570

ClueWeb09

Reference Queries Retrieval model Predictor Linear correlation coefficient Spearman's Rho Kendall's Tau
[5] 1-50 (cat B) LM (Dirichlet) Clarity 0.105
[5] 1-50 (cat B) LM (Dirichlet) Clarity+MaxVARTF.IDF 0.555
[5] 1-50 (cat B) LM (Dirichlet) Clarity+e-dispersion 0.353
[5] 1-50 (cat B) LM (Dirichlet) QF 0.516
[5] 1-50 (cat B) LM (Dirichlet) QF+MaxVARTF.IDF 0.651
[5] 1-50 (cat B) LM (Dirichlet) QF+e-dispersion 0.569
[5] 1-50 (cat B) LM (Dirichlet) WIG 0.507
[5] 1-50 (cat B) LM (Dirichlet) WIG+MaxVARTF.IDF 0.587
[5] 1-50 (cat B) LM (Dirichlet) WIG+e-dispersion 0.525
[5] 51-100 (cat B) LM (Dirichlet) Clarity -0.147
[5] 51-100 (cat B) LM (Dirichlet) Clarity+MaxVARTF.IDF 0.206
[5] 1-50 (cat B) LM (Dirichlet) Clarity+e-dispersion 0.095
[5] 51-100 (cat B) LM (Dirichlet) QF 0.393
[5] 51-100 (cat B) LM (Dirichlet) QF+MaxVARTF.IDF 0.557
[5] 1-50 (cat B) LM (Dirichlet) QF+e-dispersion 0.495
[5] 51-100 (cat B) LM (Dirichlet) WIG 0.415
[5] 51-100 (cat B) LM (Dirichlet) WIG+MaxVARTF.IDF 0.333
[5] 1-50 (cat B) LM (Dirichlet) WIG+e-dispersion 0.471
[10] 1-200 (cat B) LM (Dirichlet) MaxVAR 0.358
[10] 1-200 (cat B) LM (Dirichlet) NQC 0.234
[10] 1-200 (cat B) LM (Dirichlet) UEF(NQC) 0.272
[10] 1-200 (cat B) LM (Dirichlet) UEF(Improved Clarity) 0.265
[10] 1-200 (cat B) LM (Dirichlet) LTRoq 0.512

Bibliography

[1] Zhou, Yun, and W. Bruce Croft. "Query performance prediction in web search environments." Proceedings of the 30th annual international ACM SIGIR conference on Research and development in information retrieval. ACM, 2007.

[2] Zhao, Ying, Falk Scholer, and Yohannes Tsegay. "Effective pre-retrieval query performance prediction using similarity and variability evidence." Advances in Information Retrieval. Springer Berlin Heidelberg, 2008. 52-64.

[3] Cummins, Ronan, Joemon Jose, and Colm O'Riordan. "Improved query performance prediction using standard deviation." Proceedings of the 34th international ACM SIGIR conference on Research and development in Information Retrieval. ACM, 2011.

[4] Shtok, Anna, Oren Kurland, and David Carmel. "Using statistical decision theory and relevance models for query-performance prediction." Proceedings of the 33rd international ACM SIGIR conference on Research and development in information retrieval. ACM, 2010.

[5] Kurland, Oren, et al. "Back to the roots: A probabilistic framework for query-performance prediction." Proceedings of the 21st ACM international conference on Information and knowledge management. ACM, 2012.

[6] Diaz, Fernando. "Performance prediction using spatial autocorrelation." Proceedings of the 30th annual international ACM SIGIR conference on Research and development in information retrieval. ACM, 2007.

[7] Collins-Thompson, Kevyn, and Paul N. Bennett. "Predicting query performance via classification." Advances in Information Retrieval. Springer Berlin Heidelberg, 2010. 140-152.

[8] Zhou, Yun, and W. Bruce Croft. "Measuring ranked list robustness for query performance prediction." Knowledge and Information Systems 16.2 (2008): 155-171.

[9] Hauff, Claudia, Djoerd Hiemstra, and Franciska de Jong. "A survey of pre-retrieval query performance predictors." Proceedings of the 17th ACM conference on Information and knowledge management. ACM, 2008.

[10] Raiber, Fiana, and Oren Kurland. "Query-performance prediction: Setting the expectations straight." Proceedings of the 37th international ACM SIGIR conference on Research & development in information retrieval. ACM, 2014.