Citation Context

...ser query sessions as a source of information to improve retrieval include Furnas [10], who explicitly asked users whether their reformulations should be used for indexing, and Radlinski and Joachims =-=[15]-=- who use query chains followed by clicks as a source of relevance ordering information. Cucerzan and Brill [4] user query logs for spelling correction, but do not take advantage of user query reformu...

Citation Context

...94, Boyan et al. 1996, Freund et al. 1998, Cohen et al. 1999, Herbrich et al. 2000, Crammer and Singer 2001, Kemp and Ramamohanarao 2002, Joachims 2002, Holland et al. 2003, Almeida and Almeida 2004, =-=Radlinski and Joachims 2005-=-, Burges et al. 2005]). However, a great bottleneck in the application of machine learning techniques is the availability of training data. In this paper we explore and evaluate strategies for how to ...

Citation Context

...mined the relationship between the inter-query time interval and these classes in order to better use timeouts to predict topic transitions in a probabilistic Bayesian network. Radlinski and Joachims =-=[15]-=- identified sequences of queries on the same topic. The authors manually grouped sequences of queries into topics that had no hierarchical or interleaved structure. They built a classifier using featu...

Citation Context

...studied different timeouts to segment user sessions, and later extended their work [13] to consider other features such as the overlap between terms in two consecutive queries. Radlinski and Joachims =-=[19]-=- observe that users often perform a sequence, or chain, of queries with a similar information need; they refer to this sequence of reformulated queries as query chains. Their paper presents a simple m...

Citation Context

...ch has become an essential component of the Internet infrastructure, and has hence attracted significant interest from the machine learning community (e.g. Herbrich et al., 2000; Burges et al., 2005; =-=Radlinski & Joachims, 2005-=-; Chu & Ghahramani, 2005; Metzler & Croft, 2005; Yue et al., 2007; Taylor et al., 2008). The conventional approach to this learningto-rank problem has been to assume the availability of manually label...

Citation Context

...it could still fall into the problem of data sparsity. Interestingly, one can notice that many probabilistic personalized search algorithms proposed nowadays are essentially computing ˆp(Url|Q, User) =-=[20, 17, 16]-=-. This means that we can easily adopt any such personalized search algorithm, and compute p(QT |Url, User) = ˆp(URL|QT,User)p(QT |User) . (2) p(URL|User) This suggests that without twisting the algori...

Citation Context

... Keywords Clickthrough data, Web search, Active exploration, Learning to rank 1. INTRODUCTION There has recently been an interest in training search engines automatically using machine learning (e.g. =-=[20, 4, 24]-=-). The ideal training data would be rankings of documents ordered by relevance for some set of queries. In some cases it is practical to hire experts to manually provide relevance information for part...

Citation Context

...owever, this approach only considers page visitation statistics independently of each other, not taking into account the pages’ relative positions on post-query browsing paths. Radlinski and Joachims =-=[13]-=- have utilized such collective user intelligence to improve retrieval accuracy by using sequences of consecutive query reformulations, yet their approach does not consider users’ interactions beyond t...

Citation Context

...s idea is not new, and has underpinned many of the ideas in the work presented here. A learning to rank community (Joachims 2002; Burges, Shaked, Renshaw, Lazier, Deeds, Hamilton, and Hullender 2005; =-=Radlinski and Joachims 2005-=-; Roussinov and Fan 2005; Agarwal 2006; Cao, Xu, Liu, Li, Huang, and Hon 2006; Richardson, Prakash, and Brill 2006) is making advances for both static and query-specific ranking. There may be unexplor...