Chinese NER Using Lattice LSTM

ACL 2018  ·  Yue Zhang, Jie Yang ·

We investigate a lattice-structured LSTM model for Chinese NER, which encodes a sequence of input characters as well as all potential words that match a lexicon. Compared with character-based methods, our model explicitly leverages word and word sequence information. Compared with word-based methods, lattice LSTM does not suffer from segmentation errors. Gated recurrent cells allow our model to choose the most relevant characters and words from a sentence for better NER results. Experiments on various datasets show that lattice LSTM outperforms both word-based and character-based LSTM baselines, achieving the best results.

PDF Abstract ACL 2018 PDF ACL 2018 Abstract

Datasets


Introduced in the Paper:

Resume NER

Used in the Paper:

Weibo NER MSRA CN NER OntoNotes 4.0
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Chinese Named Entity Recognition MSRA Lattice F1 93.18 # 19
Chinese Named Entity Recognition OntoNotes 4 Lattice F1 73.88 # 14
Chinese Named Entity Recognition Resume NER Lattice F1 94.46 # 13
Chinese Named Entity Recognition Weibo NER Lattice F1 58.79 # 15

Methods