Measuring Search Results Shelly Warwick MLS Ph D

Measuring Search Results Shelly Warwick, MLS, Ph. D. 2012 – Permission is granted to reproduce and edit this work for non-commercial educational use as long as attribution is provided and the edited work is also available under the same terms of license.

After This Unit You Should Be Able to State: �The three basic ways to measure search results? �How is each of them calculated? �How can you improve search results if you find too few articles? �How can you improve search results if you find too many articles? �What are some problems with general search engines?

Measures of Search Results used to evaluate search Engines �Relevance – how many of the items retrieved really match what is being sought (subjective judgment of searcher) �Recall – how many of the existing relevant items are found by a search (need to know how many items exist) �Precision – how many of the items found in a search are relevant

Finding the Balance �A search that is very precise – where every or almost item you found meets your needs might mean that you missed some important items �A search that has a high level of recall might retrieve so many items that you can’t read or evaluate them in any meaningful timeframe.

Working with these concepts �When not enough items are retrieved you can improve recall – by BROADENING your search –using fewer terms, or filters �When too many items are retrieved you need to narrow your search to improve precision by adding more terms or filters

Problems with General Purpose Keyword Search Engines (Google, Bing, Yahoo) �Searches based on undisclosed algorithms �What is displayed can be based on your prior searching history �Relevance often based on redundancy or other poor indicator �Payment can insure a higher ranking in search results �Not all resources on the web are indexed and searched by all search engines