8. ASK and it will be given

I use different search methods for different purposes. When looking for resources for this blog, my first port of call is recommended course reading before embarking on Google for quick answers; Zdnet for technical definitions; W3Schools and Webmonkey for tutorials and tips and ISI Web of Knowledge and City University London library for academic books and journals.

Belkin, Oddy and Brooks refer to Information Retrieval (IR) as "resolving a user's anomalous state of knowledge", or ASK. They define the process as "ASK, query and evaluate" (Journal of Documentation, 38(2) p61-71, June 1982). When searching for this article, I used the university library only to find it is subscription only, so I used Google and found it on Nicholas Belkin's website.

Search engines are fast ways to find information, operating by web crawling, sometimes referred to as spider or robot (see Search Engine Watch). As a test, I try to find out about information or library science using Boolean operators:

information OR library AND science

The results were (click to enlarge):

As the results were mostly irrelevant, I changed the search to:

information AND science OR library AND science

This improved the relevance but unfortunately a band called Library Science was included in the top results:

This shows how framing questions in different ways can affect the results. Different search engines will also reveal different results (see Google results). Search operators can be used to make the results more meaningful.

IR systems, like search engines, are subjective and differ from data retrieval which is objective (see SQLs and Pearls). To improve search results, data needs to be indexed by collecting, parsing and storing. This involves removing stop words, suffixes and constructing a thesaurus. IR systems use inverted files which enable fast information retrieval. This link shows how an inverted file works.

Evaluating the success of the IR process can be qualitative such as interviews and questionnaires, or quantitative by testing in lab conditions (see Brin and Page's evaluation of Google search results). Another method is to assess the financial value of the information.

The ASK model works with text but may not be appropriate for media such as sound - Zdnet gives examples of music search methods, including Midomi, a website that can identify songs using your voice by singing or humming.