Do Transformer Attention Heads Provide Transparency in Abstractive Summarization?

Abstract

Learning algorithms become more powerful, often at the cost of increased complexity. In response, the demand for algorithms to be transparent is growing. In NLP tasks, attention distributions learned by attention-based deep learning models are used to gain insights in the models' behavior. To which extent is this perspective valid for all NLP tasks? We investigate whether distributions calculated by different attention heads in a transformer architecture can be used to improve transparency in the task of abstractive summarization. To this end, we present both a qualitative and quantitative analysis to investigate the behavior of the attention heads. We show that some attention heads indeed specialize towards syntactically and semantically distinct input. We propose an approach to evaluate to which extent the Transformer model relies on specifically learned attention distributions. We also discuss what this implies for using attention distributions as a means of transparency.

To appear at FACTS-IR 2019, SIGIR

References in corpus (6)

• Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
• Towards A Rigorous Science of Interpretable Machine Learning
• Teaching Machines to Read and Comprehend
• Explaining Explanations in AI
• Abstractive Text Summarization Using Sequence-to-Sequence RNNs and Beyond
• Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned

Cited by in corpus (3)

• On Identifiability in Transformers
• Understanding Multi-Head Attention in Abstractive Summarization
• Attention Approximates Sparse Distributed Memory