Abstract

Natural language (NL) queries formed by novice, inexperienced and occasional users tend to be incomplete, mainly because this class of users are not likely to be familiar with the functional or organizational specifications of the underlying database. A subclass of incomplete NL queries is identified, namely queries with missing infonnation. The focus of the paper is on data semantics issues involved in handling the NL queries with missing infurmation. In particular, the following issues are addressed: What kinds of semantics about the data are necessary for the system to determine what information is missing in a query? What techniques can the system employ to carry out the decision process? If the user fails to provide the answer to a supplementary information request, how can the system calculate an alternative way of requesting the supplementary information? An approach to solving these problems is also provided.

Share

COinS