Multi-stage Pretraining for Abstractive Summarization
This work addresses the need for better abstractive summarization models for natural language processing applications, but it is incremental as it builds on existing pretraining and modeling techniques.
The paper tackles the problem of improving abstractive summarization by showing that multi-stage pretraining, which includes generic language tasks and specialized summarization tasks, can enhance performance. It achieves improvements of 1.05 ROUGE-L points on Gigaword and 1.78 ROUGE-L points on CNN/DailyMail compared to a baseline.
Neural models for abstractive summarization tend to achieve the best performance in the presence of highly specialized, summarization specific modeling add-ons such as pointer-generator, coverage-modeling, and inferencetime heuristics. We show here that pretraining can complement such modeling advancements to yield improved results in both short-form and long-form abstractive summarization using two key concepts: full-network initialization and multi-stage pretraining. Our method allows the model to transitively benefit from multiple pretraining tasks, from generic language tasks to a specialized summarization task to an even more specialized one such as bullet-based summarization. Using this approach, we demonstrate improvements of 1.05 ROUGE-L points on the Gigaword benchmark and 1.78 ROUGE-L points on the CNN/DailyMail benchmark, compared to a randomly-initialized baseline.