Neural Discourse Structure for Text Categorization
This work addresses text categorization for NLP applications, presenting an incremental improvement by integrating existing discourse parsing with new attention mechanisms.
The authors tackled text categorization by incorporating discourse structure from Rhetorical Structure Theory, using a recursive neural network with a novel attention mechanism to focus on salient content, resulting in improved categorization performance with specific experimental gains.
We show that discourse structure, as defined by Rhetorical Structure Theory and provided by an existing discourse parser, benefits text categorization. Our approach uses a recursive neural network and a newly proposed attention mechanism to compute a representation of the text that focuses on salient content, from the perspective of both RST and the task. Experiments consider variants of the approach and illustrate its strengths and weaknesses.