Detecting the central units in two different genres and languages: a preliminary study of Brazilian Portuguese and Basque texts

Mikel Iruskieta, Gorka Labaka, Juliano D. Antonio


The aim of this paper is to present the development of a rule-based automatic detector which determines the main idea or the most pertinent discourse unit in two different languages such as Basque and Brazilian Portuguese and in two distinct genres such as scientific abstracts and argumentative answers. The central unit (CU) may be of interest to understand texts regarding relational discourse structure and it can be applied to Natural Language Processing (NLP) tasks such as automatic summarization, question-answer systems or sentiment analysis. In the case of argumentative answer genre, the identification of CU is an essential step for an eventual implementation of an automatic evaluator for this genre. The theoretical background which underlies the paper is Mann and Thompson’s (1988) Rhetorical Structure Theory (RST), following discourse segmentation and CU annotation. Results show that the CUs in different languages and in different genres are detected automatically with similar results, although there is space for improvement.

Texto completo: