Query Weighting for Ranking Model Adaptation

Proceedings of the 49th Annual Meeting of the Association for Computational Linguistics, pages 112–122, Portland, Oregon, June 19-24, 2011. c⃝2011 Association for Computational Linguistics Query Weighting for Ranking Model Adaptation Peng Cai1, Wei Gao2, Aoying Zhou1, and Kam-Fai Wong2,3 1East China Normal University, Shanghai, China pengcai2010@gmail.com, ayzhou@sei.ecnu.edu.cn 2The Chinese University of Hong Kong, Shatin, N.T., Hong Kong {wgao, kfwong}@se.cuhk.edu.hk 3Key Laboratory of High Confidence Software Technologies, Ministry of Education, China Abstract We propose to directly measure the impor- tance of queries in the source domain to the target domain where no rank labels of doc- uments are available, which is referred to as query weighting. Query weighting is a key step in ranking model adaptation. As the learning object of ranking algorithms is divided by query instances, we argue that it’s more reasonable to conduct importance weighting at query level than document level. We present two query weighting schemes. The first compresses the query into a query feature vector, which aggregates all document instances in the same query, and then con- ducts query weighting based on the query fea- ture vector. This method can efficiently esti- mate query importance by compressing query data, but the potential risk is information loss resulted from the compression. The second measures the similarity between the source query and each target query, and then com- bines these fine-grained similarity values for its importance estimation. Adaptation exper- iments on LETOR3.0 data set demonstrate that query weighting significantly outperforms document instance weighting methods. 1 Introduction Learning to rank, which aims at ranking documents in terms of their relevance to user’s query, has been widely studied in machine learning and information retrieval communities (Herbrich et al., 2000; Fre- und et al., 2004; Burges et al., 2005; Yue et al., 2007; Cao et al., 2007; Liu, 2009). In general, large amount of training data need to be annotated by domain experts for achieving better ranking per- formance. In real applications, however, it is time consuming and expensive to annotate training data for each search domain. To alleviate the lack of training data in the target domain, many researchers have proposed to transfer ranking knowledge from the source domain with plenty of labeled data to the target domain where only a few or no labeled data is available, which is known as ranking model adapta- tion (Chen et al., 2008a; Chen et al., 2010; Chen et al., 2008b; Geng et al., 2009; Gao et al., 2009). Intuitively, the more similar an source instance is to the target instances, it is expected to be more useful for cross-domain knowledge transfer. This motivated the popular domain adaptation solution based on instance weighting, which assigns larger weights to those transferable instances so that the model trained on the source domain can adapt more effectively to the target domain (Jiang and Zhai, 2007). Existing instance weighting schemes mainly focus on the adaptation problem for classification (Zadrozny, 2004; Huang et al., 2007; Jiang and Zhai, 2007; Sugiyama et al., 2008). Although instance weighting scheme may …

Content truncated to 500 words