Acquisition of triples of knowledge from lecture notes: a natural language processing approach

Abstract

Automated acquisition of knowledge from text has been utilised across several research areas including domain modeling of knowledge-based systems and semantic web. Primarily, knowledge is decomposed as fragments in the form of entities and relations called triples (or triplets). Although empirical studies have already been developed to extract entities (or concepts), relation extraction is still considered as a challenging task and hence, performed semi-automatically or manually in educational applications such as Intelligent Tutoring Systems. This paper presents Natural Language Processing (NLP) techniques to identify subject-verb-object (SVO) in lecture notes, supporting the creation of concept-relation-concept triple for visualisation in concept map activities. Domain experts have already been invested in producing legible slides. However, automated knowledge acquisition is challenging due to potential issues such as the use of sentence fragments, ambiguity and confusing use of idioms. Our work integrates the naturally-structured layout of presentation environments to solve semantically, syntactically missing or ambiguous elements. We evaluate our approach using a corpus of Computer Science lecture notes and discuss further uses of our technique in the educational context.