Thematic Relations Based QA Generator for Sinhala

Abstract

Abstract: Question-Answer Generation is one of the research areas in the field of Natural Language Processing. This paper presents how the Sinhala question-answer generator operates through the thematic relation on the given input sentence. This Sinhala question generator operates on an input of a well-defined Sinhala sentence with the subject, object, and verb along with some specifically identified POS tags for the system like Nouns, Pronouns, Adjectives, Verbs and Adverbs which can make noun phrases and verb phrases from all the possible combinations. The system uses a Sinhala tokenizer, Sinhala POS tagger, Chunker, and the semantic relationship identifier with the support of the NLTK libraries. Through the syntax analysis, it can identify thematic relations for the Sinhala input sentence. Through these semantics relations, it has the ability to generate Sinhala questions and related answers through the rule-based approach. The present system has been tested with 56 sample sentences. According to the evaluation, the system shows a 93% correctness rate.