Masked Conditional Random Fields for Sequence Labeling

Tianwen Wei, Jianwei Qi, Shenghuan He, Songtao Sun

3/19/2021

cs.LGcs.CL

Official code: DandyQi/MaskedCRF (44 stars, last commit 2021-09-26).

Abstract

Conditional Random Field (CRF) based neural models are among the most performant methods for solving sequence labeling problems. Despite its great success, CRF has the shortcoming of occasionally generating illegal sequences of tags, e.g. sequences containing an "I-" tag immediately after an "O" tag, which is forbidden by the underlying BIO tagging scheme. In this work, we propose Masked Conditional Random Field (MCRF), an easy to implement variant of CRF that impose restrictions on candidate paths during both training and decoding phases. We show that the proposed method thoroughly resolves this issue and brings consistent improvement over existing CRF-based models with near zero additional cost.

View on arXiv View PDF

Code Implementations(1)

DandyQi/MaskedCRFOfficial100%

447Shell, PythonMar 18, 20215 years ago

Similar Papers

Design Challenges and Misconceptions in Neural Sequence Labeling

86% match1 repo1.9k2018

NCRF++: An Open-source Neural Sequence Labeling Toolkit

84% match1 repo1.9k2018

Empower Sequence Labeling with Task-Aware Neural Language Model

83% match1 repo8442017

Hierarchically-Refined Label Attention Network for Sequence Labeling

82% match1 repo2912019

Levenshtein Transformer

80% match1 repo32.2k2019

Towards Neural Phrase-based Machine Translation

80% match1 repo1772017

Cite this paper

@article{wei2021masked,
  title  = {Masked Conditional Random Fields for Sequence Labeling},
  author = {Tianwen Wei and Jianwei Qi and Shenghuan He and Songtao Sun},
  year   = {2021},
  eprint = {2103.10682},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/2103.10682}
}

Add a badge to your README

Shows whether official code is linked for this paper.

[![Code of Paper](https://codeofpaper.com/badge/2103.10682.svg)](https://codeofpaper.com/paper/2103.10682)