SOTAVerified

On Extractive and Abstractive Neural Document Summarization with Transformer Language Models

2019-09-07EMNLP 2020Code Available0· sign in to hype

Sandeep Subramanian, Raymond Li, Jonathan Pilault, Christopher Pal

Code Available — Be the first to reproduce this paper.

Reproduce

Code

Abstract

We present a method to produce abstractive summaries of long documents that exceed several thousand words via neural abstractive summarization. We perform a simple extractive step before generating a summary, which is then used to condition the transformer language model on relevant information before being tasked with generating a summary. We show that this extractive step significantly improves summarization results. We also show that this approach produces more abstractive summaries compared to prior work that employs a copy mechanism while still achieving higher rouge scores. Note: The abstract above was not written by the authors, it was generated by one of the models presented in this paper.

Tasks

Benchmark Results

DatasetModelMetricClaimedVerifiedStatus
Arxiv HEP-TH citation graphTLM-I+EROUGE-142.43Unverified
Arxiv HEP-TH citation graphSent-PTRROUGE-142.32Unverified
Arxiv HEP-TH citation graphSent-CLFROUGE-134.01Unverified
PubmedSent-CLFROUGE-145.01Unverified
PubmedSent-PTRROUGE-143.3Unverified
PubmedTLM-I+EROUGE-141.43Unverified

Reproductions