Discourse Structure Extraction from Pre-Trained and Fine-Tuned Language Models in Dialogues
Computation and Language
2023-06-27 v2
Abstract
Discourse processing suffers from data sparsity, especially for dialogues. As a result, we explore approaches to build discourse structures for dialogues, based on attention matrices from Pre-trained Language Models (PLMs). We investigate multiple tasks for fine-tuning and show that the dialogue-tailored Sentence Ordering task performs best. To locate and exploit discourse information in PLMs, we propose an unsupervised and a semi-supervised method. Our proposals achieve encouraging results on the STAC corpus, with F1 scores of 57.2 and 59.3 for unsupervised and semi-supervised methods, respectively. When restricted to projective trees, our scores improved to 63.3 and 68.1.
Cite
@article{arxiv.2302.05895,
title = {Discourse Structure Extraction from Pre-Trained and Fine-Tuned Language Models in Dialogues},
author = {Chuyuan Li and Patrick Huber and Wen Xiao and Maxime Amblard and Chloé Braud and Giuseppe Carenini},
journal= {arXiv preprint arXiv:2302.05895},
year = {2023}
}