Interacting With Retrieval Systems

A common theme in the primary readings on Evaluation of Retrieval Systems centered on the weaknesses of search engines in general and full-text searching specifically. The supplemental readings followed the same theme. I found some humor in Matthew Koll’s report on Information Retrieval and some insightful suggestions from Daniel Tunkelang in his Reconsidering Relevance and Embracing Interaction. The retrieval systems of the not-too-distant future will undoubtedly require more interaction from information seekers, but will likewise be more powerful and deliver more relevant results in return.

The Evaluation of Retrieval Systems literature makes well the point that information seekers go about the search process with different interests and bring different experiences and knowledge to their search attempts. Koll suggests that too many searches fail because searchers are looking in the wrong place and that casual searchers don’t have time for a lot of interaction and may not even be able to provide all of the information the system requires for a successful search. With some levity, Koll suggests that in future information conferences (like the one he reports on) the classes should be made available by full-text search to see if attendees find the session they truly want. In spite of the noted progress, the basic challenge of retrieval systems providing successful search results continues to be a difficult problem.

Contrasting Koll’s opinion that casual searchers will not take the time to be more interactive, Tunkelang predicts that perhaps the truly ranked search era is coming to a close and we will soon be interacting much more with search engines. He cites the relaunch of Microsoft’s “Bing” web search engine as evidence. Bing offers users an interactive exchange permitting quicker refinement of the search results in categories of applicable data related to the original search terms. Google has likewise made steps in this direction on their advanced and experimental interfaces.

Further evidence for the close of an era is given by Tunkelang by discussion of his company’s (Endeca’s) long ago development of faceted search technology, which is now used by, or simulated in the proprietary systems of web retailers (such as Amazon and eBay). This approach adds to the conventional text search the ability to refine queries in such a way that users can navigate the content in more than one way at once. On shopping sites, for example, product searches are refined by such facets as brands, price or product type.

Tunkelang progresses through a discussion of Human-Computer Information Retrieval (HCIR), a system approach (which includes faceted searching) that calls for computing tools that augment the search process with human interaction to take advantage of human intelligence and attention. The objective of this approach is to give information seekers the chance to clarify or elaborate on their intended search. Ultimately, searches will be more focused and more easily refined, leading to more relevant results. The focus comes from the system asking or prompting the user, so that the information seeker need not yet “know what they don’t know”, as Koll might say it. They just have to be able to answer the questions or prompts from the system.

The concluding statement of this article is that users seem ready for a new interface that utilizes the intelligence and attention of the seeker to guide and refine the in-progress search leading to more successful outcomes.

Retrieval systems will certainly become better as developers invest in new interfaces and technologies. The question remains, will information seekers embrace the responsibility to take control of the search process and become more engaged in their casual searches.

Koll, Matthew. (2000) Annual Meeting Coverage - Track 3: Information Retrieval. Bulletin of The American Society for Information Science and Technology. 26:(2).
http://www.asis.org/Bulletin/Jan-00/track_3.html

Tunkelang, Daniel. (2009) Reconsidering Relevance and Embracing Interaction. Bullletin of the American Society for Information Science and Technology. 36:(1). 20-23http://www.asis.org/Bulletin/Oct-09/OctNov09_Tunkelang.pdf