My work into intelligent question answering is a response to a real problem in librarianship. The problem is that more than 125 million reference questions are asked in U.S. libraries every year. However, research suggests that the accuracy of the libra rian's response is only about 50%--one out of every two questions is answered correctly. Therefore, an intelligent decision-support system could serve librarians well. Such a system could free up the valuable time of reference librarians so that they ca n spend it on answering the more demanding research-type questions. As an aside, another response would be pedagogical--better training is needed. In that respect, my system will aid library and information science instructors in the following way.Copyright © 1996 by John V. Richardson Jr., OCLC Visiting Distinguished ScholarLibrarians make a big deal about the "real" question--the one determined by question negotiation during the reference transaction. Question negotiation, as taught in most places, involves the phrasing of questions--most notably in asking open-ended qu estions. Unfortunately, this approach only addresses the structure of a question, but does not address the issue of what is the content of this negotiation. My work answers that question in the following way.
This project assumes that the reference problem-solution boundaries (i.e., the search space of all reference questions and reference sources) are finite. Furthermore, the implementation is predicated upon the so-called Mudge Method or the Hutchins Heu ristic--that each and every reference source can be classified into a finite number of categories (in this case, there are fourteen) for cognitive efficiency. Cognitive effectiveness is achieved by classifying reference questions by format and then by sp ecific source. In order to determine their utility in the reference environment, I used a kind of distinctive feature analysis to categorize the sources. Hence, the interface poses questions much like the reference librarian should be doing in order to r each the correct conclusion.
As alluded to above, the actual implementation of my intelligent question answering system is based on decision-rules using a multiple-choice classification process. Without a doubt, the theory is a reductive transformation of the reference libraria n's complex decision-making task; nonetheless, the advantage is that it converts this complex task into a much, more manageable one for a computer-mediated environment.
To read about this process in greater detail, please consult: Knowledge-based Systems for General Reference Work: Applications, Problems, and Progress (San Diego, CA: Academic Pr ess, 1995) or "Question Master: An Evaluation of a Web-based Decision-Support System for Use in Reference Environments," College and Research Libraries 59 (January 1998): 29-37.