Generation of Natural Language Text using Perspective Descriptor in Frames

Abstract

Generation of natural language text requires the ability to determine what information to include and how to organize it to achieve its communicative goal effectively. Natural language generation is a highly complex task whose automation is performed by accessing huge amount of knowledge sources such as domain knowledge, world knowledge and common sense knowledge. Software documentation is chosen as application for our work because of its closed ended domain i.e. even if the domain is not clearly specified the inputs required have to be properly identified for producing corresponding document. Since different levels of documents (Documenting each one of them is a critical job. The documentation is nothing but the art of accumulation, classification and dissemination of information) have to be produced depending upon the status of the user, the inputs have to be identified accordingly. The main aim of this work is to identify necessary inputs needed for generating a natural language text in the form of software document using knowledge representation technique, frames, by adding a new component called perspective descriptors, by which the epistomological status of the frame is being improved which increases the efficiency of the search engine and tries to reduce the spatio temporal complexities obtained in the existing frame structure. Frames are chosen as knowledge representation scheme because of their efficiency in processing and producing huge amount of information without any deterioration.