Do We Really Need That Many Parameters In Transformer For Extractive Summarization? Discourse Can Help !
This work addresses the problem of reducing computational cost and model size for extractive summarization, which is beneficial for researchers and practitioners working with resource-constrained environments or large-scale NLP applications.
This paper explores parameter-lean self-attention mechanisms for extractive summarization, proposing a novel tree self-attention based on document-level discourse information. The approach achieves competitive ROUGE scores compared to the original single-head transformer model with significantly fewer parameters, and outperforms an 8-head transformer on sentence level with an order of magnitude less parameters.
The multi-head self-attention of popular transformer models is widely used within Natural Language Processing (NLP), including for the task of extractive summarization. With the goal of analyzing and pruning the parameter-heavy self-attention mechanism, there are multiple approaches proposing more parameter-light self-attention alternatives. In this paper, we present a novel parameter-lean self-attention mechanism using discourse priors. Our new tree self-attention is based on document-level discourse information, extending the recently proposed "Synthesizer" framework with another lightweight alternative. We show empirical results that our tree self-attention approach achieves competitive ROUGE-scores on the task of extractive summarization. When compared to the original single-head transformer model, the tree attention approach reaches similar performance on both, EDU and sentence level, despite the significant reduction of parameters in the attention component. We further significantly outperform the 8-head transformer model on sentence level when applying a more balanced hyper-parameter setting, requiring an order of magnitude less parameters.