Discourse Structure Extraction from Pre-Trained and Fine-Tuned Language Models in Dialogues
This addresses data sparsity in discourse processing for dialogues, offering incremental improvements in a domain-specific task.
The paper tackled discourse structure extraction in dialogues by using attention matrices from pre-trained language models, proposing unsupervised and semi-supervised methods that achieved F1 scores up to 68.1 on the STAC corpus.
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.