Towards a Relation-Based Argument Extraction Model for Argumentation Mining

• Tipo: Artigo em Livro de Atas de Conferência Internacional
• Data: 2017

Autores

Gil Rocha (Autor) FEUP; Henrique Lopes Cardoso (Autor) FEUP

Ata de Conferência Internacional

Título: Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

Páginas: 94-105

5th International Conference on Statistical Language and Speech Processing, SLSP 2017

23 October 2017 through 25 October 2017

Indexação

Scopus - 3 Citações

Outras Informações

ID Authenticus: P-00N-6VW

DOI: 10.1007/978-3-319-68456-7_8

Abstract (EN): Argumentation mining aims to detect and identify the argumentative content expressed in text. In this paper we present a relation-based approach that aims to capture the relation of inference between the premise and conclusion. We follow a supervised machine learning approach and explore features at different levels of abstraction. Then, we apply this system for the task of argumentative sentence detection and compare the performance of the system with a competitive baseline approach. The corpus used in our experiments was annotated with arguments from textual resources written in Portuguese, namely opinion articles. The proposed system outperforms the baseline system, achieving 0.75 of f1-score on the test set. © Springer International Publishing AG 2017.

Idioma: Inglês

Tipo (Avaliação Docente): Científica