PromptNER: Prompt Locating and Typing for Named Entity Recognition

Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics Volume 1: Long Papers, pages 12492–12507 July 9-14, 2023 ©2023 Association for Computational Linguistics PromptNER: Prompt Locating and Typing for Named Entity Recognition Yongliang Shen1, Zeqi Tan1, Shuhui Wu1, Wenqi Zhang1, Rongsheng Zhang2, Yadong Xi2, Weiming Lu1†, Yueting Zhuang1 1College of Computer Science and Technology, Zhejiang University 2Fuxi AI Lab, NetEase Inc. {syl, luwm}@zju.edu.cn Abstract Prompt learning is a new paradigm for utilizing pre-trained language models and has achieved great success in many tasks. To adopt prompt learning in the NER task, two kinds of methods have been explored from a pair of symmetric perspectives, populating the template by enu- merating spans to predict their entity types or constructing type-specific prompts to locate en- tities. However, these methods not only require a multi-round prompting manner with a high time overhead and computational cost, but also require elaborate prompt templates, that are dif- ficult to apply in practical scenarios. In this paper, we unify entity locating and entity typ- ing into prompt learning, and design a dual-slot multi-prompt template with the position slot and type slot to prompt locating and typing re- spectively. Multiple prompts can be input to the model simultaneously, and then the model extracts all entities by parallel predictions on the slots. To assign labels for the slots dur- ing training, we design a dynamic template filling mechanism that uses the extended bipar- tite graph matching between prompts and the ground-truth entities. We conduct experiments in various settings, including resource-rich flat and nested NER datasets and low-resource in- domain and cross-domain datasets. Experi- mental results show that the proposed model achieves a significant performance improve- ment, especially in the cross-domain few-shot setting, which outperforms the state-of-the-art model by +7.7% on average1. 1 Introduction Named entity recognition (NER) is a fundamen- tal task in natural language processing that aims to identify specific types of entities in free text, such as person, location, and organization. Tradi- tional sequence labeling methods (Ma and Hovy, † Corresponding author. 1 Our code will be available at https://github.com/ tricktreat/PromptNER. ⑦ (c) PromptNER { [ Pi ] is a [ Ti ] entity }i=1,2,…, only 1 round ① & Typing Locating (a) Type-oriented rounds Find person entities Find organization entities Find location entities ······ Locating ① ② (b) Span-oriented San Francisco is a [MASK] entity rounds Jobs is a [MASK] entity Jobs was is a [MASK] entity ······ Typing ① ② ⑩ ··· ··· : Figure 1: A comparison of the type-oriented (a) and span-oriented (b) prompt learning with the proposed PromptNER (c). C, N and M denote the number of entity types, words and prompts, respectively. 2016) have difficulty coping with nested entities, and recent works have transformed NER into other paradigms such as reading comprehension (Li et al., 2020; Shen et al., 2022), set prediction (Tan et al., 2021; Wu et al., 2022a) and sequence generation (Paolini et al., 2021; Yan et al., 2021; Lu et al., 2022). However, low-resource and …

Content truncated to 500 words