On a novice-user-focused approach to flexible querying: The case of initially unavailable explicit user preferences

摘要

A novel solution is proposed to an important problem of learning real querying preferences and intentions from users who need to retrieve interesting information from a database but are not in a position to specify their information needs and/or intentions using a query language due to lack of knowledge and/or experience. A solution is proposed that is based on the presentation to the user of consecutive examples of data items, requesting his/her evaluations of those data items, and then using a classification method to learn the user preferences to be converted into their corresponding query. A novel element is that the system proposed allows for both positive and negative responses from the user which makes possible to represent the bipolarity of preferences. The considerations are illustrated on an example of a querying system for a real estate agency that makes possible to retrieve houses from a database that more adequately reflects the users' intentions and preferences. A Web based implementation is briefly presented. Results obtained are promising.