Abstract | ||
---|---|---|
Accurate estimates of examination bias are crucial for unbiased learning-to-rank from implicit feedback in search engines and recommender systems, since they enable the use of Inverse Propensity Score (IPS) weighting techniques to address selection biases and missing data citep{Joachims/etal/17a}. Unfortunately, existing examination-bias estimators citep{Agarwal/etal/18c, wang2018position} are limited to the Position-Based Model (PBM) citep{chuklin2015click}, where the examination bias may only depend on the rank of the document. To overcome this limitation, we propose a Contextual Position-Based Model (CPBM) where the examination bias may also depend on a context vector describing the query and the user. Furthermore, we propose an effective estimator for the CPBM based on intervention harvesting citep{Agarwal/etal/18c}. A key feature of the estimator is that it does not require disruptive interventions but merely exploits natural variation resulting from the use of multiple historic ranking functions. Semi-synthetic experiments on the Yahoo Learning-To-Rank dataset demonstrate the superior effectiveness of the new approach. |
Year | DOI | Venue |
---|---|---|
2018 | 10.1145/3331184.3331238 | Proceedings of the 42nd International ACM SIGIR Conference on Research and Development in Information Retrieval |
Keywords | Field | DocType |
examination bias, propensity estimation, unbiased learning-to-rank | Recommender system,Data mining,Weighting,Search engine,Ranking,Propensity score matching,Computer science,Robustness (computer science),Missing data,Estimator | Journal |
Volume | Citations | PageRank |
abs/1811.01802 | 6 | 0.40 |
References | Authors | |
0 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Zhichong Fang | 1 | 6 | 0.40 |
Aman Agarwal | 2 | 60 | 3.91 |
Thorsten Joachims | 3 | 17387 | 1254.06 |