LSTM-CRF Models for Named Entity Recognition

Changki LEE

doi:10.1587/transinf.2016EDP7179

LSTM-CRF Models for Named Entity Recognition

Changki LEE

Full Text Views

0

Share
Cite this

Summary :

Recurrent neural networks (RNNs) are a powerful model for sequential data. RNNs that use long short-term memory (LSTM) cells have proven effective in handwriting recognition, language modeling, speech recognition, and language comprehension tasks. In this study, we propose LSTM conditional random fields (LSTM-CRF); it is an LSTM-based RNN model that uses output-label dependencies with transition features and a CRF-like sequence-level objective function. We also propose variations to the LSTM-CRF model using a gate recurrent unit (GRU) and structurally constrained recurrent network (SCRN). Empirical results reveal that our proposed models attain state-of-the-art performance for named entity recognition.

Publication: IEICE TRANSACTIONS on Information Vol.E100-D No.4 pp.882-887

Publication Date: 2017/04/01

Publicized: 2017/01/20

Online ISSN: 1745-1361

DOI: 10.1587/transinf.2016EDP7179

Type of Manuscript: PAPER

Category: Natural Language Processing

Authors

Changki LEE
Kangwon National University

Keyword

LSTM-CRF, LSTM RNN, recurrent neural network, name entity recognition

Cite this

Copy

Changki LEE, "LSTM-CRF Models for Named Entity Recognition" in IEICE TRANSACTIONS on Information, vol. E100-D, no. 4, pp. 882-887, April 2017, doi: 10.1587/transinf.2016EDP7179.
Abstract: Recurrent neural networks (RNNs) are a powerful model for sequential data. RNNs that use long short-term memory (LSTM) cells have proven effective in handwriting recognition, language modeling, speech recognition, and language comprehension tasks. In this study, we propose LSTM conditional random fields (LSTM-CRF); it is an LSTM-based RNN model that uses output-label dependencies with transition features and a CRF-like sequence-level objective function. We also propose variations to the LSTM-CRF model using a gate recurrent unit (GRU) and structurally constrained recurrent network (SCRN). Empirical results reveal that our proposed models attain state-of-the-art performance for named entity recognition.
URL: https://globals.ieice.org/en_transactions/information/10.1587/transinf.2016EDP7179/_p

Copy

@ARTICLE{e100-d_4_882,
author={Changki LEE, },
journal={IEICE TRANSACTIONS on Information},
title={LSTM-CRF Models for Named Entity Recognition},
year={2017},
volume={E100-D},
number={4},
pages={882-887},
abstract={Recurrent neural networks (RNNs) are a powerful model for sequential data. RNNs that use long short-term memory (LSTM) cells have proven effective in handwriting recognition, language modeling, speech recognition, and language comprehension tasks. In this study, we propose LSTM conditional random fields (LSTM-CRF); it is an LSTM-based RNN model that uses output-label dependencies with transition features and a CRF-like sequence-level objective function. We also propose variations to the LSTM-CRF model using a gate recurrent unit (GRU) and structurally constrained recurrent network (SCRN). Empirical results reveal that our proposed models attain state-of-the-art performance for named entity recognition.},
keywords={},
doi={10.1587/transinf.2016EDP7179},
ISSN={1745-1361},
month={April},}

Copy

TY - JOUR
TI - LSTM-CRF Models for Named Entity Recognition
T2 - IEICE TRANSACTIONS on Information
SP - 882
EP - 887
AU - Changki LEE
PY - 2017
DO - 10.1587/transinf.2016EDP7179
JO - IEICE TRANSACTIONS on Information
SN - 1745-1361
VL - E100-D
IS - 4
JA - IEICE TRANSACTIONS on Information
Y1 - April 2017
AB - Recurrent neural networks (RNNs) are a powerful model for sequential data. RNNs that use long short-term memory (LSTM) cells have proven effective in handwriting recognition, language modeling, speech recognition, and language comprehension tasks. In this study, we propose LSTM conditional random fields (LSTM-CRF); it is an LSTM-based RNN model that uses output-label dependencies with transition features and a CRF-like sequence-level objective function. We also propose variations to the LSTM-CRF model using a gate recurrent unit (GRU) and structurally constrained recurrent network (SCRN). Empirical results reveal that our proposed models attain state-of-the-art performance for named entity recognition.
ER -