Query formulation in web search

Authors


Abstract

Over 2 billion searches are conducted daily on Web search engines. Some Web searches are re-finding what was seen before, but many searches are query reformulation or refinements as attempts to retrieve better results. Previous studies have examined query formulation and reformulation from the mid-seventies to today, including online intermediary and end-user searching, CD-ROM and OPAC environments, and Web search studies. Query research is important theoretically and practically for the Web industry. This panel presents three recent query formulation and reformulation studies in diverse environments and user groups, including Web search logs and Web searching by 4–5 year olds. The short panel presentations will highlight the theoretical and practical implications of the results and allow sufficient time to engage the audience in discussion.

Overview of Query Formulation & Refining

This brief presentation will provide an introduction and overview of research in query formulation and refining. (presented by Dr. Efthimis N. Efthimiadis)

Query Reformulation: Analyzing and Evaluating Search Strategies in Web Search Engine Logs

This presentation discusses how frequently users modify a previous search query by query reformulations or query refinements. Previous studies have examined how Web search engines propose reformulations, but less attention has focused on how people perform query reformulations across different user groups. We examine how Web searchers refine queries and form a theoretical foundation for query reformulation by studying users' reformulation strategies in the context of AOL Web search logs. We provide taxonomy of query refinement strategies and build a high precision rule-based classifier to detect each type of reformulation. Effectiveness of reformulations is measured by user click behavior. Most reformulation strategies positively affect these metrics. Certain strategies like add/remove words, word relation, acronym expansion, and spelling correction result in higher ranked unique clicks. However, users often click the same result as their previous query or select no results when forming acronyms and reordering words. From these findings, applications are possible to assist Web searchers, such as improved interfaces supporting query reformulation assistance and personalized interaction. (presented by Jeff Huang and Dr. Efthimis N. Efthimiadis)

Exploring Young Childrens' Web Query Formulation and Reformulation

This presentation reports findings from an exploratory study of the Web interactions by young children 4 or 5 years of age. Few studies have included children younger than six years of age. The study audio and videotaped the Google Web searches by 12 young children enrolled in a ‘preparatory classroom’ or kindergarten in Queensland Australia. The data was qualitatively analysed to understand the young childrens' Web search behaviour. Findings show that despite spelling and literacy issues, study participants engaged in complex Web search behaviors, including query formulation and reformulation. Examples of the young childrens' query formulation and reformulation processes are discussed. Web search interactions by young children are an important research area with implications for educators and Web technology design. (Presented by Dr. Amanda Spink)

Classifying and Predicting Query Reformulation in Web Searching

Query reformulation is a key user behavior in Web searching. This presentation will highlight ongoing efforts to develop predictive models of query reformulation during Web searching. We report results from lab and Web search logs studies using n-gram approaches to show the probability of users transitioning from one query reformulation state to another in order to predict their next state. Results show that first and second order models are generally the best for next state prediction. The n-gram approach can be used to improve searching systems and assistance. (Presented by Dr. Jim Jansen)

Ancillary