Web Search: Public Searching of the Web

"Web Search: Public Searching of the Web", co-authored by Drs. Amanda Spink and Bernard J. Jansen, is one of the first manuscripts that address the human - system interaction of Web searching in a thorough and complete manner. The authors provide an examination of Web searching from multiple levels of analysis, from theoretical overview to detailed study of term usage, and integrate these different levels of analysis into a coherent picture of how people locate information on the Web using search engines.

Drawing primarily on their own research and work in the field, the authors present the temporal changes in, the growth of, and the stability of how people interact with Web search engines. Drs. Spink and Jansen present results from an analysis of multiple search engine data sets over a six year period, giving a firsthand account of the emergence of Web searching. They also compare and contrast their findings to the results of other researchers in the field, providing a valuable bibliographic resource. This research is directly relevant to those interested in providing information or services on the Web, along with those who research and study the Web as an information resource. Graduate students, academic and corporate researchers, search engine designers, information architects, and search engine optimizers will find the book of particular benefit. 


Leseprobe:

Chapter 4 (p. 55-56)

SEARCH TERMS


1. INTRODUCTION

This chapter reports results from an analysis of the search terms submitted to Web search engines – AlltheWeb.com, AltaVista and Excite. Terms are the basic building blocks through which a Web searcher expresses their information problem when searching on a Web search engine. Single or multiple term and operators form a Web query. What are the subjects of Web users’ search terms? Where do the search terms come from? Why does a user select one term instead of another? What influences a searcher’s decisions?

Major findings suggest: (1) the topic interests of Web search engine users has shifted to commercial and informational from the sexual and technology domains, (2) the information problems of Web search engine users are becoming increasingly more diverse, (3) there is a notable increase in non- English terms, numbers, and acronyms used as Web search terms, (4) a set of approximately 20% of search terms are used with great regularity while approximately 10% of the terms are used only once, and (5) major news events and holidays influence search term usage.

Many researchers view Web search as a communication process in which there is a dialog or discourse occurring between the searcher and the Web search engine (Jansen, 2003; Spink, 1997). A dialog is a communication exchange about a certain topic between a user and a Web search engine that includes thinking on the part of the user. Iivonen and Sonnenwald (1998) note that when selecting search terms, searchers appear to navigate a variety of dialogs. Searchers evaluate and synthesize information among these dialogs in order to select search terms.

Hsieh-Yee (1993) reports that the level of a user’s search experience and domain knowledge affects the searchers' selection of search terms. Along with domain knowledge and searching experience, Spink and Saracevic (1997) identified three other sources of search terms pertinent to Web searching, namely (1) the users' level of domain knowledge of their search topic, (2) the Web systems output, and (3) a thesaurus or related terms. They noted that search terms from the user’s domain and the system’s output were the terms that helped the most in retrieving relevant documents.

Researchers have also investigated reformulation (Dennis, Bruza and McArthur, 2002) and search term weighting in order to improve performance. The underlying assumption is that not all terms in a query are of equal importance. The most well known case being that of stop words (Fox, 1990), which are query terms that occur so frequently that they are deemed of little content value (e.g. and, or, the). Some Web search engines automatically remove stop words from queries unless the user specifically tells the search engine (via query operators such as PHRASE or MUST APPEAR) to keep them in the query. Members of some communities refer to stop words as filter words (WebMasterWorld.com, 2004), in which case stop words refer to terms in Web documents that cause a Web search engine spider to stop indexing.

The idea behind term weighting is that the terms with the most importance should have more effect on the retrieval process. Budzik, Hammond, and Birnbaum (2001) use a version of term weighting in an application to automatically formulate queries. Some Web search engines have attempted to implement term weighting automatically using clickthrough data from query transaction logs (Schaale, Wulf-Mathies and Lieberam-Schmidt, 2003).