NAG-NER: a Unifed Non-Autoregressive Generation Framework for Various NER Tasks

Abstract

Recently, the recognition of fat, nested, and discontinuous entities by a unifed generative model framework has received increasing attention both in the research feld and industry. However, the current generative NER methods force the entities to be generated in a predefned order, suffering from error propagation and in-effcient decoding. In this work, we propose a unifed non-autoregressive generation (NAG) framework for general NER tasks, referred to as NAG-NER. First, we propose to generate entities as a set instead of a sequence, avoiding error propagation. Second, we propose incorporating NAG in NER tasks for effcient decoding by treating each entity as a target sequence. Third, to enhance the generation performances of the NAG decoder, we employ the NAG encoder to detect potential entity mentions. Extensive experiments show that our NAG-NER model outperforms the state-of-the-art generative NER models on three benchmark NER datasets of different types and two of our proprietary NER tasks.