Towards a top-down approach for an automatic discourse analysis for Basque: Segmentation and Central Unit detection...

Area of Science:

• Computational Linguistics
• Natural Language Processing (NLP)

Background:

• Discourse structure analysis is crucial for various NLP tasks like opinion mining and summarization.
• Current discourse parsers excel at intra-sentence relations but struggle with inter-sentence connections.
• Identifying a Central Unit aids in improving rhetorical labeling and discourse analysis accuracy.

Purpose of the Study:

• To develop the initial stages of a discourse parser using a top-down strategy.
• To create a tool for automatic text segmentation and Central Unit detection.
• To lay the groundwork for future advancements in assigning rhetorical relations.

Main Methods:

• Implemented a top-down strategy for discourse parsing.
• Developed a discourse segmenter to identify basic discourse units.
• Created an automatic Central Unit detector to pinpoint the main discourse component.

Main Results:

• Successfully built a tool combining a discourse segmenter and a Central Unit detector.
• The tool addresses the challenge of accurately capturing inter-sentence discourse relations.
• Provides a foundation for more sophisticated discourse analysis and rhetorical labeling.

Conclusions:

• The developed tool represents a significant step towards more accurate discourse parsing.
• Automatic Central Unit detection enhances the capabilities of NLP systems.
• Future work will focus on integrating rhetorical relation assignment into the parser.