Abstractive Summarization of English Legal Documents

Anonymous

Keywords: legal text summarization, text generation, transformers

Abstract: Automatic text summarization has been found more and more useful nowadays because it can help to find relevant information quickly. In the legal domain, documents are usually long and filled with many technical terms. Some recent approaches focused on extractive summarization methods to generate summaries for English legal documents. Most of the existing works using abstractive summarization, however, are for non-English legal documents. This study presents the first attempt to utilize a distillation version of the BART model (distilBART) for abstractive summarization of English legal documents. The results on benchmark legal corpora show that distilBART outperforms the state-of-the-art summarization models on this task.

Paper Type: long

Research Area: Summarization