A pressing problem confronted by Information Organization practitioners is subject access of the OPAC. Keyword searches often lead to either failure or retrieving too many references. When searches yield zero hits, whatever the reason, many users will abandon the search. This is not a function of computer literacy. The OPAC is a black box to users and they know very little about what happens inside the system… and they shouldn't need to. This is why post patrons prefer “free-text” searching, because it requires little thought – however it produces less intelligent results. Ranganathan’s law, “Save the time of the reader,” is applicable here. It is the responsibility of those of us organizing information to allow the searcher to find his information accurately in a little time as possible.
Three possible solutions to the problem are increasing instruction in Boolean search techniques, allowing users to browse subjects, and natural language searching.
Boolean Search Techniques
One solution is to increase instruction in Boolean database search techniques at the library. Classes or instructional materials are provided for a patron to “follow” while he is learning how to effectively search the OPAC. Many colleges require these classes for incoming Freshmen to learn these techniques to search the OPAC and library databases. Boolean operators can greatly help users to increase precision. This is a very effective searching strategy when used correctly. One can broaden or narrow the search using the applicable Boolean terms (AND, OR, NOT). It is inexpensive – it does not require the library to invest more scarce resources in the design of the OPAC, other than the costs associate with training.
However, Google does not require instruction – and that is the interface that our patrons are coming to the library with. Boolean logic is not common sense for most people. Libraries are to serve humanity; humanity should not have to serve the library. Most end users search the OPAC only occasionally and do not access the system on a regular basis and they tend to learn only enough to do simple searches reasonably quickly and to regard further instruction as unnecessary and more extensive expertise as a burden.
Browsing Subject Indexes
Another solution is to programmatically permit the patron to browse subject indexes. Subject headings could be clicked through, such as through a tree, until arriving at the subject desired, using Java Script or a similar user interface. This author used a database at the Foundation Center that allowed users to browse the grant database using a similar interface. The user does not have to guess at key terms to search, reducing the introduction of human error and ending with relevant search results. Authority control is easily maintained, since the tree must be traversed and the results are within the domain of the controlled vocabulary. When the patron finds the subject term he wishes, he will find accurate results to his query. This would be a highly usable method for any user to find subjects of interest without having to know ahead of time exactly what term is being searched for.
However, the patron might “get lost” if he traverses the tree in the wrong direction, and become frustrated. This can also be time consuming search method if the desired term is deep within the tree. This writer believes that, while this was an effective method of searching the Foundation Center grant database, the exhaustive scope of the LC Subject Headings would make this a very cumbersome method of searching a library OPAC.
Natural Language Front-End / Subject Headings in the Back-End
We can create an intelligent natural-language front end which uses subject headings and an ontology for searching the OPAC, through enhancements of the database and DBMS layers. The OPAC search engine would understand the concept in the question rather than just matching the text. The database would include an ontological metadata layer. An ontology is more than a taxonomy which places concepts into a hierarchy. An ontology allows concepts to be connected through a variety of multidimensional relationships, which reflects the vocabulary of the domain – in this case the LC Subject Headings. Semantic links are descriptive terms which illustrate the connections between ideas (such as "is the parent of" or "is a subset of") – which are used to create the relationships in an ontology. This provides better routes of topic exploration, and allows for search techniques based on natural language parsing. A programming framework can then be used to interpret the patron’s request and query the database via the ontology. The patron does not have to change his behavior to work with the software – the software will work with what the patron gives it, including suggestions in the case of a typographical error. Authority control is maintained because the software takes the patron’s language and, via the ontology, translates the meaning to the controlled vocabulary of the Subject Headings. The results will be accurate and relevant – finding the intended concept instead of the precise word (please see King, B. E., & Reinold, K. (2008). Finding the concept, not just the word: A librarian's guide to ontologies and semantics. Oxford: Chandos for a much more extensive discussion of this idea).
However, often subject cataloging is superficial and inadequate. This is critical to help guide the patron to the subjects he is seeking. The average bibliographic record contains less than 2 subject headings. Also, ontology creation and programming the user interface can be a very expensive proposition for a budgetary strapped library system.
This writer believes that the most effective solution is clearly creating a natural language front-end; creating a user interface that makes the search process for the user “Google-ish.” This is much like conducting a reference interview. One doesn’t require the patron to ask the question “correctly,” nor does one refuse to serve the patron if the terminology is not within our controlled vocabulary. It is the responsibility of the information professional to lead the patron to the information he desires. The gathering process is often invisible to the patron. The librarian connects what a patron asks for with what else is out there – other topics connected to it, how old the information might be, and where she might look for an answer. The librarian then evaluates results to see if it answers the question.
So should it be with the OPAC. The OPAC should function as the librarian does at a reference interview, and deliver accurate results that answers the patron’s question.