人工智能支持的英语图文推理及教学应用  

English inference of image-text and its application in teaching supported by artificial intelligence

作　　者：曾立英[1] 金心怡 陈艺宣 Zeng Liying;Jin Xinyi;Chen Yixuan

机构地区：[1]中央民族大学国际教育学院

出　　处：《西安外国语大学学报》2023年第1期59-65,共7页Journal of Xi’an International Studies University

基　　金：教育部产学合作协同育人项目“人工智能背景下专业师资培育与提升研究”(项目编号:220660228290403)的阶段性研究成果。

摘　　要：推理任务在人工智能技术发展中逐渐受到重视,联合图像与英语文本前提的推理,是一项比较重要的多模态推理任务。文章在情绪ABC理论支持下,依托人工智能标注平台,针对欧美电影2951幅图片及其备选前提,人工选择符合图片的前提,撰写了5902条英文正反向推理句,构建了图文推理库。在此基础上结合推理前提的选择情况,分析推理句中关键词的语义分类、句子的形式标记等特征,归纳出图文推理的类型与特点。其中的推理类型主要涵盖推理行为动作、认知心理活动、事件因果关系等几类。最后根据人工智能支持的标注任务,以促进信息技术与英语语言教育的融合发展。Inference process plays an important role in artificial intelligence. This paper completes a multimodal inference task which combined image with English text. It tries to build an English database of image-text inference from the perspective of ABC Theory of Emotion. The task is conducted on the annotation platform designed by artificial intelligence, and it generates 5902 true and false hypothetical actions for 2951 image-premise pairs. Images are taken from European and American movie screenshots, and alternative premises are automatically generated by computer. This paper sums up the types and features of image-text inferences by analyzing the selection of premises, the semantic classification of keywords and the formal marker of sentences. The types of inferences can be divided into four categories: the action of human behavior, the mental activity of cognition, the causality of events and others. It concludes by designing the application of inference tasks in English teaching,to promote the integration of information technology and English language education.

关 键 词：人工智能 推理 图片 前提 英语教学 

分 类 号：H030[语言文字—语言学]