Clustering of Deep Contextualized Representations for Summarization of Biomedical Texts

TL;DR

This paper presents a BERT-based summarization method for biomedical texts that effectively measures sentence similarity and content informativeness without relying on domain knowledge bases, improving summarization performance.

Contribution

The study demonstrates that pre-trained BERT representations can replace domain knowledge bases in biomedical summarization, capturing sentence context more accurately.

Findings

BERT-based summarizer outperforms traditional methods

Effective sentence similarity measurement without domain knowledge

Source code and data are publicly available

Abstract

In recent years, summarizers that incorporate domain knowledge into the process of text summarization have outperformed generic methods, especially for summarization of biomedical texts. However, construction and maintenance of domain knowledge bases are resource-intense tasks requiring significant manual annotation. In this paper, we demonstrate that contextualized representations extracted from the pre-trained deep language model BERT, can be effectively used to measure the similarity between sentences and to quantify the informative content. The results show that our BERT-based summarizer can improve the performance of biomedical summarization. Although the summarizer does not use any sources of domain knowledge, it can capture the context of sentences more accurately than the comparison methods. The source code and data are available at https://github.com/BioTextSumm/BERT-based-Summ.