Discourse marker generation and syntactic aggregation in Bengali text generation

Abstract

In discourse, the elementary text spans are semantically connected by coherence relations. Discourse markers linguistically realize the coherence relations in the surface form. On the other hand, by text aggregation, redundant entities are eliminated, resulting in more fluent, coherent, and concise text. For any but the most application of text generation, appropriate discourse marker selection and text aggregation are two important aspects for coherent text generation. In this paper, we explore the prevalent syntactic aggregation constructs in Bengali and present a rule based approach towards generating Bengali compound sentences using the identified constructs. We present a user based evaluation to validate our approach. At the end, we have also given an outline of a corpus based approach for generating suitable discourse marker in Bengali.