Design Challenges and Misconceptions in Neural Sequence Labeling

Jie Yang, Shuailong Liang, Yue Zhang
6/12/2018
cs.CL

Official code: jiesutd/NCRFpp (1,901 stars, Apache-2.0, PyTorch, last commit 2022-06-30).

Abstract

We investigate the design challenges of constructing effective and efficient neural sequence labeling systems, by reproducing twelve neural sequence labeling models, which include most of the state-of-the-art structures, and conduct a systematic model comparison on three benchmarks (i.e. NER, Chunking, and POS tagging). Misconceptions and inconsistent conclusions in existing literature are examined and clarified under statistical experiments. In the comparison and analysis process, we reach several practical conclusions which can be useful to practitioners.

View on arXivView PDF

Code Implementations(1)

jiesutd/NCRFppOfficial100%
PyTorchApache-2.0

NCRF++, a Neural Sequence Labeling Toolkit. Easy use to any sequence labeling tasks (e.g. NER, POS, Segmentation). It includes character LSTM/CNN, word LSTM/CNN and softmax/CRF components.

1,901433PythonDec 6, 20174 years agoApache-2.0
artificial-intelligencechar-cnnchar-rnnchunkingcnn+11 more

Similar Papers

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

89% match1 repo1.9k2018

Masked Conditional Random Fields for Sequence Labeling

86% match1 repo442021

Hierarchically-Refined Label Attention Network for Sequence Labeling

85% match1 repo2912019

Empower Sequence Labeling with Task-Aware Neural Language Model

82% match1 repo8442017

Approximate Structured Diffusion for Sequence Labelling

81% match2026

Optimal Hyperparameters for Deep LSTM-Networks for Sequence Labeling Tasks

80% match1 repo8342017

Cite this paper

@article{yang2018design,
  title  = {Design Challenges and Misconceptions in Neural Sequence Labeling},
  author = {Jie Yang and Shuailong Liang and Yue Zhang},
  year   = {2018},
  eprint = {1806.04470},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1806.04470}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1806.04470.svg)](https://codeofpaper.com/paper/1806.04470)

Discussion