Pre-trained text encoders have rapidly advanced the state of the art on many NLP tasks. We focus on one such model, BERT, and aim to quantify where linguistic information is captured within the network. We find that the model represents the steps of the traditional NLP pipeline in an interpretable and localizable way, and that the regions responsible for each step appear in the expected sequence: POS tagging, parsing, NER, semantic roles, then coreference. Qualitative analysis reveals that the model can and often does adjust this pipeline dynamically, revising lower-level decisions on the basis of disambiguating information from higher-level representations.
BERT Rediscovers the Classical NLP Pipeline
BERT captures linguistic information in a structured manner similar to the traditional NLP pipeline, with representations for POS tagging, parsing, NER, semantic roles, and coreference appearing in sequence, and often dynamically revising lower-level decisions.
- Year
- 2019
- Venue
- bert-rediscovers-the-classical-nlp-pipeline-1
- Authors
- 3
- Hosting
- Abstract onlyARXIV-DEFAULT
Cite
Notes
Only stored in your browser.
Attribution
- Abstract & full text
- arxiv.org/abs/1905.05950v2ARXIV-DEFAULT
- TL;DR
- Semantic Scholar