FLAT: Chinese NER Using Flat-Lattice Transformer
This work addresses efficiency issues in Chinese NER for NLP practitioners, offering a method that enhances parallel computation while maintaining accuracy, though it is incremental as it builds on existing lattice-based approaches.
The paper tackles the challenge of low inference speed in Chinese named entity recognition (NER) models using lattice structures by proposing FLAT, a Flat-Lattice Transformer that converts the lattice into a flat span-based structure, resulting in improved performance and efficiency on four datasets.
Recently, the character-word lattice structure has been proved to be effective for Chinese named entity recognition (NER) by incorporating the word information. However, since the lattice structure is complex and dynamic, most existing lattice-based models are hard to fully utilize the parallel computation of GPUs and usually have a low inference-speed. In this paper, we propose FLAT: Flat-LAttice Transformer for Chinese NER, which converts the lattice structure into a flat structure consisting of spans. Each span corresponds to a character or latent word and its position in the original lattice. With the power of Transformer and well-designed position encoding, FLAT can fully leverage the lattice information and has an excellent parallelization ability. Experiments on four datasets show FLAT outperforms other lexicon-based models in performance and efficiency.