A system for parsing, transforming, and generating sentences

A sentence can be transformed to other related forms by changing selected features such as voice or tense, by moving auxiliaries, or by inserting or deleting words. The transformations which relate sentences that have the same constituents but a different emphasis are said to be paraphrastic. This property can be exploited in developing user interfaces for applications which make use of the questioning and answering process. An example of such a system is an expert system which may ask a question to guide its conclusion and then restate the information in the form of a sentence when explaining the conclusion. This thesis describes a system that captures syntactic information about English sentences and makes the information available for transformation and generation. A format is proposed for representing syntactic information, and a grammar is presented for parsing an input sentence into the data structure and generating paraphrased forms from it. Procedures are provided for adding word classifications to the lexicon used by the grammar rides and for selecting features for paraphrasing. The system is implemented in Prolog.