DADgraph: A Discourse-aware Dialogue Graph Neural Network for Multiparty Dialogue Machine Reading Comprehension
This work addresses the challenge of handling complex dialogue discourse structure in multiparty dialogue MRC, which is incremental as it builds on existing neural network methods by incorporating discourse-aware elements.
The paper tackled the problem of multiparty dialogue machine reading comprehension by proposing DADgraph, a discourse-aware dialogue graph neural network that explicitly constructs dialogue graphs using discourse dependency links and relations, achieving statistically significant improvements over strong neural network baselines on the Molweni corpus.
Multiparty Dialogue Machine Reading Comprehension (MRC) differs from traditional MRC as models must handle the complex dialogue discourse structure, previously unconsidered in traditional MRC. To fully exploit such discourse structure in multiparty dialogue, we present a discourse-aware dialogue graph neural network, DADgraph, which explicitly constructs the dialogue graph using discourse dependency links and discourse relations. To validate our model, we perform experiments on the Molweni corpus, a large-scale MRC dataset built over multiparty dialogue annotated with discourse structure. Experiments on Molweni show that our discourse-aware model achieves statistically significant improvements compared against strong neural network MRC baselines.