Evaluation

Summary

Users translate their information needs into a query. Standardized query language (SQL) is often used to communicate with indexing software, but search engines such as the OPACs evaluated here use natural language, full-text and Boolean queries. The user must decide if the results are useful; not whether they match the query. Then, the query is adjusted until the information need is satisfied. User satisfaction can be difficult to measure. This rubric includes recall and precision numbers even though they do not reflect the problem to be solved, the searcher’s knowledge or the cognitive process. Although they can be used to measure relevance, they are used here as a point of interest, or baseline to understand the breadth of the collections in general. Because user happiness is difficult to measure, I included a row for rating whether I would use the IRS again.

Information retrieval systems use structures to index information. Inverted file structures use a vocabulary that contains all the values being indexed along with an inverted list for each distinct value. The inverted list stores the corresponding record tags for the values. Then the IRS evaluates the records for terms that correspond to the query and retrieve the shortest inverted list first, followed by increasingly longer lists. Arrays and hash tables can be used to store vocabularies instead of lists. Conversely, bitslice signature files are sets of bitsliced signature files hashed together with the query to form a bitmap of potential results. They can be blocked into sets to save processing; all of which is controlled by algorithm. This method takes more time and computing power, so inverted file structures are more common drivers of IRSs (Zobel, Moffat, Ramamohanarao, 1998). Query variances were used to reveal as many differences in index structure as possible. The Pueblo IR could not find any two of the original keywords when searched together much less interpret variants.

The query results varied from 3,566 to 0. The University of Wyoming Coe library returned 290 results, but offered to search its interlibrary loan IR on the results page. The Pueblo Public Library also had this feature. The Colorado University IR offered access to their other collections before displaying the interlibrary loan option.

The interface is the point of contact between information and the user. Therefore, it should be free of jargon and clutter. Attributes of interface evaluation include: Text size and font, brightness, menus, response speed, graphics, capabilities, special arrangements, interactive functions like chat, user ability level, satisfaction and error rate (Rubin, 2004). The menu topics and vocabularies varied widely between the OPACs. For example, the Colorado University Norlin Library website required navigation through several pages to arrive at subject guides even though what looked like a direct link was on the home page. Pueblo Public Library used a larger tab bar to consolidate subpages into extensive dropdown menus. This seemed more direct because the user could choose from more destinations at once. The noteworthy aspect of the Coe interface was the three large contact buttons on the home page: Chat, e-mail, text. It’s bright colors and large text were inviting and user friendly. The way the playful interface contrasts with the serious resources was visually refreshing.

References

Rubin R. (2004). Foundations of library and information science. Neal-Schuman Publishers, Inc. New York, NY.

Zobel J., Moffat A., Ramamohanarao K. (1998). Inverted files versus signature files for text indexing. AMC Transactions on Database Systems 23(4). 453-490.