Clarence meets Alcuin
This essay outlines the definition of expert systems, describes how this definition has been applied to reference librarianship, and suggests future directions of study.
Sometimes the functions computers perform seem like magic. They can add faster than any person. They can remember the smallest of details for the longest periods of time. The can consume, propagate, and regurgitate vast amounts of information in an instant. Many times these features of digital computers have been compared to human knowledge and thus has given rise to the field of computer science called artificial intelligence.
Clarence Ried entered the library for the purposes of beginning the art history class essay. A sophomore at State University, Clarence was studying design and hoped to someday be an interior decorator. He, unlike some of his classmates, understood the value of his library's collection of traditional resources. Even though the Network hosted vast amounts of information at a person's finger tips, much of the information was not necessarily academic in nature. The available information that was academic or scholarly in nature was not necessarily accessed for free. Consequently, a person had to have their Personal Debit Card handy in order to fund any purchases. In short, visiting the library was a bit inconvenient but much less expensive, especially for a starving student like himself.
As Clarence entered the library he saw a sign reading, "Try out our 'Portable Library Assistants.' Inquire at the Reference Desk."Feeling a bit overwhelmed about the size of the University's library and having his curiosity piqued by the sign, Clarence made his way to the Reference Desk.
"Hi. I would like to know more about the 'Library Assistants'", said Clarence to one of the reference librarians behind the desk.
"No problem, " said the librarian. As he brought out a small, hand-held computing device out from under the counter the librarian continued, "If you give me your Student Identification Card as collateral, then I will lend you a one of our 'Library Assistants.' The software saved on these devices is designed to help you use the library more effectively. Would you like to try it?"
"Sure", said Clarence giving his card to the librarian.
"Simply turn it on here, " pointing to a switch on the device's side, "and use this special pen to supply your input", said the librarian. "If you have any questions, then just tap the 'Explain' button. Okay?"
Expert sytems (sometimes known as knowledge based systems) are a branch of the study of artificial intelligence. Formally defined, an expert system is "a computer system that uses a representation of human expertise in a specialist domain in order to perform functions similar to those normally performed by a human expert in that domain."  Similarly, an expert system has been defined as "a computer program using expert knowledge to attain high levels of performance in a narrow problem domain."  Put less formally, expert systems are computer programs mimicking the decision based processes of humans in a limited area of expertise. The concept of expert systems have been applied to a wide variety of tasks ranging from the configuration of computers to playing chess to locating natural resource deposits.
Surveys of expert systems work applied to librarianship turns up a range of applications from classification to indexing to database selection. One of the earliest expert systems to be applied to the field of librarianship was a system done by a University of Chicago master's student, Cherie Weil, in 1967 who was studying reference work. The system was designed to answer basic biographical questions.  The method used by Weil seems to be the method employed by many of the implementations following her. Namely, a database of reference materials is created. Each item in the database is described, sometimes using an extended subject/quality classification system. Then, through some sort of question and answer process or a series of menus, the expert system queries its patron, searches its databases for matches, and presents the end-user an answer, usually in the form of a title to a reference work. REFSEARCH by Joseph Meredith, the Hepatitis Knowledge Base by L.M. Bernstien, Plexus by B.C. Vickery, AquaRef by Deborah Hanfman, and POINTER by Karen F. Smith all exemplify technique.
While there seem to have been quite a number of expert systems attempting to model reference services, none of them ever seem to have lasted very long. Why don't we hear about successful, long-standing systems that have been put into place over a long period of time? According to Richardson all of the reference service expert systems to date have fallen short because they have not taken into account all aspects of reference service. By reviewing the Kuhnian paradigms of teaching reference work, Richardson identified three schools of thought: structuralist, proceduralist, and "psychologicalist." The structuralist reference method stresses reference materials. The proceduralist favors standardized methods for accomplishing successful reference services. The psychologicalist believes the analysis of the mental traits of the librarian and patron are paramount for quality services. Based on this analysis, Richardson posits that only when reference is taught emphasising each of these perspectives, then libraries will be providing the best possible service. By extension, if expert systems in reference service do not incorporate these same principles, then any expert system is doomed to be incomplete and lacking.
Clarence went over to a near by chair and tapped the Explain button. The machine responded with a cheerful beep and presented Clarence with the following message.
"Greetings! My name is Alcuin, your personal library assistant. My goal is to help you get the most out of the library by providing you with information about the library's collections and services. More importantly, I am designed to help you find the information you seek through a simple question and answer process. At the end of this process you will be presented with a strategy for fulfilling your information needs. To use me, read the questions carefully and answer them to the best of your ability. If at any time you don't understand something, then tap the 'Explain' button. If worse comes to worse, ask a real live human for assistance. Tab the 'Begin' button to start the process."
As Clarence tapped the Begin button Alcuin responded, "Welcome! My name is Alcuin. What is your name?", and naturally, Clarence tapped the Explain button.
The machine responded, "'What is your name?' The purpose of this question it to make your interaction with me a bit more human-friendly. It also gives me the opportunity to remember you the next time you visit the library. Answer the question and tap the 'Next Question' button. Incidentally, my name is Alcuin. I died in 804 after a career as a librarian and advisor to Charlemagne."
Clarence thought the entire concept more than just a bit strange, but he continued anyway by entering his name and tapping the 'Next Question' button.
The question and answer process continued. Alcuin proceeded to query Clarence about his educational background and expertise. Alcuin asked him for the subject of his information need, and in a Socratic sort of way, Alcuin helped Clarence clarify and better define the subject. Alcuin asked Clarence what sort of work he had already done in this area and whether or not Clarence knew of any authors or works of significance. Alcuin queried Clarence for his preference of information media types such as books, journals, slides, Internet resources, etc. Alcuin asked Clarence how much time he had to spend on his work today and asked him how much money he was will to spend if required. Throughout the process, Alcuin presented Clarence with intermediate results. These results were incomplete in that they did not include the full exploration of Alcuin's questions. On the other hand, they gave Clarence the opportunity to verify Alcuin's progress.
Unlike "library assistants" before Alcuin, the Alcuin Library Assistant offers absolutely no answers to any of the end-user's questions. Instead, through the question and answer process (a non-traditional reference interview) Alcuin's primary objective is to create search strategies that can be applied to various databases. The reason for this is simple. There is no way any one library assistant (expert system) can keep track of the myriad individual information resources. But, if the information resources are systematically organized and consistently classified in databases, then searches can be applied to the databases throughout time and draw on new information the databases may contain. In short, Alcuin grounds his usefulness in his ability to ask questions and not necessarily knowing specific answers. Specific answers change as quickly as publishers, but the questions remain the same. While databases may come and go, and while the structure of databases may change over time, databases change at a much slower rate than the data they represent.
The interview was over. It only took about ten minutes. At the end Alcuin knew that Clarence was an undergraduate student studying art history, specifically Rococo. Clarence knew that Versailles was a good example of the Rococo style and the style grew out of Baroque Period. Through the process, Alcuin determined Clarence wanted to locate pictures of Rococo architecture, but Clarence was not willing to pay for anything. Lastly, Clarence had about another hour and a half to spend on the project today before his next class.
Despite the lack of any long term, sustained successes in the area of reference expert systems, I still believe, like Richardson, the profession can develop an expert system for the purposes of enhancing or supplementing reference services. I believe this for two reasons.
First, the majority of reference expert systems are doomed to failure because the try to suggest particular reference works. As the number of reference works increases, so does the difficulty in implementing and maintaining the expert system. As reference librarians interact with their clientele, they assemble qualities describing the client's information need. Then, rather intuitively, the librarian matches the set of qualities with a known reference work. In other words, librarians construct queries in their minds and identify one or more works from their experience embodying one or more of the qualities in question.
The same technique can be employed by computers to supplement reference service. The computer could be used to interact with patrons. This interaction would be in the form of a question and answer session. The questions would address issues surround the patron themselves like:
- education level
- subject expertise
- time and money willing to spend on information gathering
- purpose of the information gathering
Additionally, the questions would address their information need itself querying them about but not limited to the:
- format of materials sought
- preferred location of the materials
- date ranges of acceptable materials
- names of authors or titles of known items on their subject
- the depth of detail in which they want the information
After (and during) this question and answer process, the system constructs search strategies that can be applied to databases of information. These search strategies could be applied to online pubic access catalogs of traditional library materials and bibliographic databases such as the databases hosted by DIALOG or MEDLARS. There is no reason these same search strategies can not be applied to databases of Internet resources as well. Once the strategies are constructed they are given to the end-user for execution. At that time the end-user can evaluate the results for themselves, or, as described below, the system can assist the end-user in evaluating the search results.
The keys to success in this method is first to know the structure of each of the databases intimately and second to construct very finely tuned search strategies individually based on the strengths and weakness of each database. The fortitude of this approach lies in its ability to always identify valid resources without knowing particular items. Its weaknesses, on the other hand, are directly proportional to the completeness and integrity of the databases.
The second reason it will be important to build a successful expert system supplementing reference service is the need to advance and raise the level of services libraries can provide. Traditional library services have surrounded the issues of identifying, collecting, organizing, and disseminating data and information. Rarely have librarians formally asserted their abilities to evaluate information.
In today's world technology is empowering individuals to identify, collect, and organize their own sets of data and information more easily. There is less of a need for the tradtional sets of library services. On the other hand, the profession has been harboring its ability to evaluate information. This is something new we can share with our clientele, and since our clientele is increasingly reluctant to visit libraries, we can provide this new service through electronic means such as expert systems. It may be quite possible for these expert systems to determine the importance of particular resources based on things like:
- number of times a resource has been used or checked out of the library
- number of times the resource or resource's author has been cited
- number of footnotes contained in the resource
- length of the resource
- number of times search terms appear in the resource (ranked relevance)
- reviews of the resource
- number of editions of the resource
Librarians could even take the leap and assign value judgments in our databases to the resources like educational level or thoroughness and then expert systems could search for resources based on those judgments.
Based on this information, Alcuin generated a number of search strategies. First, he recommended Clarence browse a number of art-related dictionaries and encyclopedias in order to gather more specific terms like the names of artists. Second, Alcuin generated a search of the library's catalog including the term "plates", "illus." and "pictorial works" in the hopes of generating lists of books containing photographs. This was enough to get Clarence started, but before the last search strategy was presented, Alcuin said, "Clarence, once you have looked over these initial resources, but sure to go through the question and answer process again. I'm sure we can uncover more relevant materials!"
Clarence proceeded to look over the dictionaries and encyclopedias. Sure enough he came across terms he remembered his professor mentioning. After looking at a few of the books located through another of Alcuin's searches, Clarence discovered a number of similar books in the same area. Delighted with his success, Clarence started up Alcuin again to see what could possibly happen next.
"Do you know of any books that are of particular significance about 'Rococo?'", asked Alcuin. Clarence promptly answer "Yes" and proceed to supply the two of the new books he found on the shelf.
"Based on the subjects entries of the books you entered, you may be interested in the books generated from the following strategy", responded Alcuin. Selecting one of the resulting titles, Clarence asked Alcuin for the location of the material. Alcuin responded with a map of the library and blinking cursor pointing to the book's location.
Clarence was getting a bit tired and decided to call it a day. He did check out a few of the books and returned Alcuin to the reference librarian as directed. He was sure he would visit the library and Alcuin again soon.
It has been said that understanding is like a four-rung ladder. The first rung on the ladder represents data and facts. As the data and facts are collected and organized they become information, the second rung on the ladder. The third rung is knowledge where knowledge is information internalized and put to use. The last rung of the ladder is wisdom, knowledge of a timeless nature. Technology has enabled more people to climb between the first and second rungs of the ladder with greater ease. Similarly, technology may enable libraries and librarians to climb higher on the ladder as well and provide knowledge services instead of simply information services. Expert systems in reference may be one manifestation of such a climb.
- 1. This essay can also be found at http://www.lib.ncsu.edu/staff/morgan/clarence-meets-alcuin.html, and you can see the very beginnings of Ask Alcuin at http://www.lib.ncsu.edu/staff/morgan/alcuin/.
- 2. Alex Goodall, The guide to expert systems (Oxford: Learned Information, 1985), 11.
- 3. Donald A. Waterman, A guide to expert systems (Reading, MA: Addison-Wesley, 1985), 11.
- 4. Ralph Alberico and Mary Micco, Expert systems for reference and information retrieval (Westport, CT: Meckler, 1990) 86.
- John V. Richardson, Knowledge-based systems for general reference work: applications, problems, and progress (San Diego: Academic Press, 1995). This relatively new book is the most scholarly of the of readings listed here. It outlines definitions of reference work as well as knowlege-based (expert) systems. It then suggests ways to incorporate the two disciplines into a cohesive whole. Included are many references. If you are going to only read one of the suggested readings, then this is the one to choose.
- Rao Aluri and Donald E. Riggs, eds., Expert systems in libraries (Norwood, NJ: Ablex Publsihing, 1990). This collection of essays describe numerous expert systems for reference work. Contains many references and a large selected bibliography.
- Ralph Alberico and Mary Micco, Expert systems for reference and information retrieval (Westport, CT: Meckler, 1990). This book "attempts to address the development of expert systems for reference and information retrieval." It contains recommended readings and a large bibliography.
- Christine Roysdon and Howard D. White, eds., Expert systems in reference services (New York: Haworth Press, 1989). Another collection of essays describing expert systems, but unfortunately some of the examples are not really "expert systems." This is a good book for understanding the principles of expert systems and methods for constructing them.
Creator: Eric Lease Morgan <firstname.lastname@example.org>
Source: This text was originally published as Morgan, Eric Lease (1997). Clarence meets Alcuin; or, expert systems are still an option in reference work. In P. Ensor (Ed.), The Cybrarian's manual (pp. 127-134). Chicago: American Library Association.
Date created: 1996-03-03
Date updated: 2004-12-10
Subject(s): expert systems; fiction; librarianship; articles;