UW-BHI at MEDIQA 2019: An Analysis of Representation Methods for Medical Natural Language Inference

William Kearns, Wilson Lau, Jason Thomas


Abstract
Recent advances in distributed language modeling have led to large performance increases on a variety of natural language processing (NLP) tasks. However, it is not well understood how these methods may be augmented by knowledge-based approaches. This paper compares the performance and internal representation of an Enhanced Sequential Inference Model (ESIM) between three experimental conditions based on the representation method: Bidirectional Encoder Representations from Transformers (BERT), Embeddings of Semantic Predications (ESP), or Cui2Vec. The methods were evaluated on the Medical Natural Language Inference (MedNLI) subtask of the MEDIQA 2019 shared task. This task relied heavily on semantic understanding and thus served as a suitable evaluation set for the comparison of these representation methods.
Anthology ID:
W19-5054
Volume:
Proceedings of the 18th BioNLP Workshop and Shared Task
Month:
August
Year:
2019
Address:
Florence, Italy
Venue:
BioNLP
SIG:
SIGBIOMED
Publisher:
Association for Computational Linguistics
Note:
Pages:
500–509
Language:
URL:
https://aclanthology.org/W19-5054
DOI:
10.18653/v1/W19-5054
Bibkey:
Cite (ACL):
William Kearns, Wilson Lau, and Jason Thomas. 2019. UW-BHI at MEDIQA 2019: An Analysis of Representation Methods for Medical Natural Language Inference. In Proceedings of the 18th BioNLP Workshop and Shared Task, pages 500–509, Florence, Italy. Association for Computational Linguistics.
Cite (Informal):
UW-BHI at MEDIQA 2019: An Analysis of Representation Methods for Medical Natural Language Inference (Kearns et al., BioNLP 2019)
Copy Citation:
PDF:
https://preview.aclanthology.org/ingestion-script-update/W19-5054.pdf
Data
MIMIC-III