%0 Conference Paper %1 zhang2019ernie %A Zhang, Zhengyan %A Han, Xu %A Liu, Zhiyuan %A Jiang, Xin %A Sun, Maosong %A Liu, Qun %B Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics %C Florence, Italy %D 2019 %I Association for Computational Linguistics %K %P 1441--1451 %R 10.18653/v1/P19-1139 %T ERNIE: Enhanced Language Representation with Informative Entities %U https://www.aclweb.org/anthology/P19-1139 %X 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 code and datasets will be available in the future.

@inproceedings{zhang2019ernie, 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 code and datasets will be available in the future.}, added-at = {2020-11-14T16:59:53.000+0100}, address = {Florence, Italy}, author = {Zhang, Zhengyan and Han, Xu and Liu, Zhiyuan and Jiang, Xin and Sun, Maosong and Liu, Qun}, biburl = {https://www.bibsonomy.org/bibtex/22b7a1ca64a0257707544162ec61fe750/marjaw}, booktitle = {Proceedings of the 57th Annual Meeting of the Association for Computational Linguistics}, doi = {10.18653/v1/P19-1139}, interhash = {1462e4040ed1f703424e14745cb0ca8b}, intrahash = {2b7a1ca64a0257707544162ec61fe750}, keywords = {}, month = jul, pages = {1441--1451}, publisher = {Association for Computational Linguistics}, timestamp = {2020-11-14T16:59:53.000+0100}, title = {{ERNIE}: Enhanced Language Representation with Informative Entities}, url = {https://www.aclweb.org/anthology/P19-1139}, year = 2019 }