Designing Natural Language Objects

Designing a database has been a tough problem since the early days of business data processing. As domains of database applications extend to textual and multimedia information processing, the design problem has grown even harder although the recent object-oriented approach introduced a richer set of building blocks into the database arena[2]. Natural language (NL, for short) databases are particularly important among non-traditional database applications because NL appears so frequently in our everyday documents, mails, dialogs, etc. It seems quite reasonable that we keep such information as NL databases and everybody can ask/update information through an NL interface. That is, all three forms, data[l5], query[6], and answerll7] can be expressed in NL. Since the expressive power of a natural language is far more than that of a relational language, an NL query has to be limited only to retrievals without deep understanding of either NL databases or the query in order to guarantee tractability of query processing. Meaning of sentences, however, can be symbolically stored and handled by a database system if we define a set of semantic concepts and mappings between NL syntax and semantic expressions. Thus, a variety of NL applications, machine translation, NL interface, text retrieval, etc. can be built upon such semantic NL concepts. In this paper we propose an architecture to handle NL databases. We define NL classes and NL objects as a set of semantic concepts to be stored in an object-oriented database system. Mapping rules between .NL syntax and NL objects are then introduced. We will also discuss an application interface provided by the architecture. The