A portable natural language interface from Arabic to SQL
[Thesis]
Al-Johar, Badr
University of Sheffield
1999
Thesis (Ph.D.)
1999
In recent years, natural language interface systems have been built based on the Front End and the Back End architecture which gives a guarantee of modularity and portability to the system as a whole. An Arabic Front End has been built that takes an input sentence, producing syntactic and semantic representations, which it maps into First Order Logic. Expressing the meaning of the user's question in terms of high level world concepts makes the natural language interface independent of the database structure. It is then easier to port the interface Front End to a database for a different domain. The syntactic treatments are based on Generalised Phrase Structure Grammar (GPSG) whereas the semantics are expressed in formal semantics theory. The focus is mainly to provide syntactic and semantic analyses for Arabic queries based on correct Arabic linguistic principles. The proposed treatments are proved and tested by building a prototype system. The prototype is implemented using one of the existing systems called Squirrel. An Arabic morphological analyser is also proposed and implemented to distinguish between two types of morphemes: internal morphemes which are a part of the word's pattern, and external morphemes which are independent words attached to the word but which are not part of the word's pattern. So, the system focuses on the extraction of morphemes from the various inflexions or forms of any Arabic word.