Few-shot Named Entity Recognition with Entity-level Prototypical Network Enhanced by Dispersedly Distributed Prototypes

Proceedings of the 29th International Conference on Computational Linguistics, pages 163–177 October 12–17, 2022. 163 COMMA: Modeling Relationship among Motivations, Emotions and Actions in Language-based Human Activities Yuqiang Xie Yue Hu† Wei Peng Guanqun Bi Luxi Xing Institute of Information Engineering, Chinese Academy of Sciences, Beijing, China School of Cyber Security, University of Chinese Academy of Sciences, Beijing, China {xieyuqiang,huyue,pengwei,biguanqun,xingluxi}@iie.ac.cn Abstract Motivations, emotions, and actions are inter- related essential factors in human activities. While motivations and emotions have long been considered at the core of exploring how people take actions in human activities, there has been relatively little research supporting analyzing the relationship between human mental states and actions. We present the first study that investigates the viability of modeling motiva- tions, emotions, and actions in language-based human activities, named COMMA (Cognitive Framework of Human Activities). Guided by COMMA, we define three natural language pro- cessing tasks (emotion understanding, motiva- tion understanding and conditioned action gen- eration), and build a challenging dataset HAIL‡ through automatically extracting samples from Story Commonsense. Experimental results on NLP applications prove the effectiveness of modeling the relationship. Furthermore, our models inspired by COMMA can better reveal the essential relationship among motivations, emotions and actions than existing methods. 1 Introduction Human activities are continuous interactions be- tween external environment (physical world and social events, etc.) and internal mind (motivations, emotions, etc.). For example, Figure 1 demon- strates human activities of character ‘I’ about ‘eat- ing bread’ in external environment, as well as the mental states of ‘I’. The motivation of character ‘I’ is a physiological need. Conditioned on this moti- vation and history actions, ‘I’ enjoyed the part of bread on the direction of joy emotion. While hu- man mental states have long been considered at the core of exploring how people take actions between the lines in language-based human activities, there has been relatively little research supporting analyz- ing the relationship between human mental states †Corresponding author. ‡We will make our dataset and code publicly available at https://github.com/IndexFziQ/COMMA. Motivation I have a physiological need. However, I enjoyed the part that I took out. Current Action I would feel joy. Emotion History Actions I threw the whole wheat bread into the toaster. It heated the bread for several minutes in order to toast it. I had difficulty taking out the bread since it was stuck. When I took one loaf out, part of it was left in the toaster. Character: I (myself) ! " # Figure 1: An example of human activity. History ac- tions and motivation cause current action. And emotion is the effect of motivation and actions. Here, motiva- tion/action/emotion is colored by yellow/cyan/green. and actions. It is challenging to comprehensively model the relationship of motivations, emotions and actions in language-based human activities, which can allow researchers to reason the essential causes of human activities from the cognitive per- spective and supply reasonable explanations. This technology will have a profound impact on various natural language processing (NLP) downstream ap- plications, such as intelligent dialogue, …

Content truncated to 500 words