Query reformulation using anchor text

Abstract

Query reformulation techniques based on query logs have been studied as a method of capturing user intent and improving retrieval effectiveness. The evaluation of these techniques has primarily, however, focused on proprietary query logs and selected samples of queries. In this paper, we suggest that anchor text, which is readily available, can be an effective substitute for a query log and study the effectiveness of a range of query reformulation techniques (including log-based stemming, substitution, and expansion) using standard TREC collections. Our results show that logbased query reformulation techniques are indeed effective with standard collections, but expansion is a much safer form of query modification than word substitution. We also show that using anchor text as a simulated query log is as least as effective as a real log for these techniques.

Contextual Search: A Computational Framework

“…When query expansion selects the number of expansion terms depending on the user and on the the user's query, it outperforms both the original ranking and the personalization in the case of a fixed number of expansion terms as Dang and Croft [50] and Luxenburger et al [122] report. A formalization of the combination of click-through data, content and user profiles has been described by Sontag et al [158].…”

“…The overall problem and approaches to tagging and linking large photocollections is, for example, described by Crandall and Snavely [46]. A well-written paper with some interesting ideas on surrogating the availability proprietary query logs is by Dang and Croft [50]. Diaz [53] clearly explains the methods and gives a background on probability that is rather helpful for understanding the content of many contextual search papers, although the topic is not easy to newcomers.…”

“…To address the issues of noise in click-through data, Dang and Croft [50] suggest using anchor text to simulate the queries and, as a proof of the concept, they construct a simulated query log from the anchor text in the TREC .gov2 test collection. A paper that explores whether interaction can be helpful in realistic environments has been written by Agichtein et al [2] who exploit statistical models for WWW search and IR.…”

“…Dai et al [49] use "live" WWW. Dang and Croft [50] use TREC collections. Dang and Croft [50]'s work is an example of careful design and detailed implementation allow to collect many useful data about user interaction at no cost and preserving the user's privacy.…”

Simulation of Within-Session Query Variations Using a Text Segmentation Approach

“…The reformulations are qualitatively evaluated by manual judgments. Related work on automatic query reformulation for web-search to better answer the initial original information need itself, includes that of Dang and Croft [4] which uses anchor text to reformulate a query by substituting some of the original terms. Wang and Zhai [5] exploited co-occurrence of terms in search-engine query logs to add terms to correct the mis-specification and under-specification of a query.…”