On Text Tiling for Documents: A Neural-Network Approach
| Autor: | Siang-Yun Yoong, 翁湘雲 |
|---|---|
| Rok vydání: | 2019 |
| Druh dokumentu: | 學位論文 ; thesis |
| Popis: | 107 Segmenting documents or conversation threads into semantically coherent segments have been one of the challenging tasks in Natural Language Processing (NLP). BERT (Bidirectional Encoder Representation Transformer) is a language representation model that shows outstanding results on many natural language processing task. In this work, we introduce three new text segmentation models that employ BERT for post-training. Extensive experiments are conducted using benchmark datasets, and the experiment results demonstrate that our BERT-based models show significant improvements over the state-of-the-art text segmentation algorithms. |
| Databáze: | Networked Digital Library of Theses & Dissertations |
| Externí odkaz: |
|