2021/09/10 by Gengyu Wang, Xiaochen Hou, Wang, Gengyu +7
Computer Science · #Computation and Language (cs.CL) #FOS: Computer and information sciences #Multimodal Machine Learning Applications #Natural Language Processing Techniques #Topic Modeling
paper · pdf · doi:10.48550/arxiv.2109.05168
openalex publication_date 2021/09/10 · openalex created_date 2021/11/22 · openalex updated_date 2026/07/28
Large pre-trained language models (PLMs) have led to great success on various\ncommonsense question answering (QA) tasks in an end-to-end fashion. However,\nlittle attention has been paid to what commonsense knowledge is needed to\ndeeply characterize these QA tasks. In this work, we proposed to categorize the\nsemantics needed for these tasks using the SocialIQA as an example. Building\nupon our labeled social knowledge categories dataset on top of SocialIQA, we\nfurther train neural QA models to incorporate such social knowledge categories\nand relation information from a knowledge base. Unlike previous work, we\nobserve our models with semantic categorizations of social knowledge can\nachieve comparable performance with a relatively simple model and smaller size\ncompared to other complex approaches.\n