ERNIE: Enhanced Language Representation with Informative Entities

TL;DR

ERNIE is a language model that integrates knowledge graphs with large-scale text to improve understanding and performance on knowledge-driven NLP tasks, outperforming previous models like BERT.

Contribution

This paper introduces ERNIE, a novel language representation model that incorporates structured knowledge from knowledge graphs alongside textual data.

Findings

ERNIE significantly improves performance on knowledge-driven NLP tasks.

ERNIE is comparable to BERT on general NLP tasks.

The model effectively combines lexical, syntactic, and knowledge information.

Abstract

Neural language representation models such as BERT pre-trained on large-scale corpora can well capture rich semantic patterns from plain text, and be fine-tuned to consistently improve the performance of various NLP tasks. However, the existing pre-trained language models rarely consider incorporating knowledge graphs (KGs), which can provide rich structured knowledge facts for better language understanding. We argue that informative entities in KGs can enhance language representation with external knowledge. In this paper, we utilize both large-scale textual corpora and KGs to train an enhanced language representation model (ERNIE), which can take full advantage of lexical, syntactic, and knowledge information simultaneously. The experimental results have demonstrated that ERNIE achieves significant improvements on various knowledge-driven tasks, and meanwhile is comparable with the state-of-the-art model BERT on other common NLP tasks. The source code of this paper can be obtained from https://github.com/thunlp/ERNIE.