机器阅读理解(Machine reading comprehension,MRC)是自然语言处理领域中一项重要研究任务,其目标是通过机器理解给定的阅读材料和问题,最终实现自动答题.目前联合观点类问题解答和答案依据挖掘的多任务联合学习研究在机器阅读理解应用...机器阅读理解(Machine reading comprehension,MRC)是自然语言处理领域中一项重要研究任务,其目标是通过机器理解给定的阅读材料和问题,最终实现自动答题.目前联合观点类问题解答和答案依据挖掘的多任务联合学习研究在机器阅读理解应用中受到广泛关注,它可以同时给出问题答案和支撑答案的相关证据,然而现有观点类问题的答题方法在答案线索识别上表现还不是太好,已有答案依据挖掘方法仍不能较好捕获段落中词语之间的依存关系.基于此,引入多头自注意力(Multi-head self-attention,MHSA)进一步挖掘阅读材料中观点类问题的文字线索,改进了观点类问题的自动解答方法;将句法关系融入到图构建过程中,提出了基于关联要素关系图的多跳推理方法,实现了答案支撑句挖掘;通过联合优化两个子任务,构建了基于多任务联合学习的阅读理解模型.在2020中国“法研杯”司法人工智能挑战赛(China AI Law Challenge 2020,CAIL2020)和HotpotQA数据集上的实验结果表明,本文提出的方法比已有基线模型的效果更好.展开更多
This paper suggests a theoretical model for teaching syntactic relations to foreign students of English Linguistics using insights from Cognitive Grammar as developed in Langacker(e.g.1987,1991,1999) and elaborated on...This paper suggests a theoretical model for teaching syntactic relations to foreign students of English Linguistics using insights from Cognitive Grammar as developed in Langacker(e.g.1987,1991,1999) and elaborated on in Radden and Dirven(2007). The model deals with syntactic relations as participants in a prototypically structured network of events, radiating out from the experiential action chain schema(a skeletal concept of force dynamics involved in the interaction between entities)(see Langacker, 1999). This involves dealing with syntactic relations as meaning-making tools that collaborate with the verb in a regular manner manipulating the prototypical schema. The aim is to draw attention to the need to add a cognitive semiotic dimension to the teaching of sentence structure. Such a dimension exposes learners to the cognitive factors behind grammatical meaning making and may, therefore, result in a better understanding of language structure and use on the part of learners who can become English language teachers.展开更多
文摘机器阅读理解(Machine reading comprehension,MRC)是自然语言处理领域中一项重要研究任务,其目标是通过机器理解给定的阅读材料和问题,最终实现自动答题.目前联合观点类问题解答和答案依据挖掘的多任务联合学习研究在机器阅读理解应用中受到广泛关注,它可以同时给出问题答案和支撑答案的相关证据,然而现有观点类问题的答题方法在答案线索识别上表现还不是太好,已有答案依据挖掘方法仍不能较好捕获段落中词语之间的依存关系.基于此,引入多头自注意力(Multi-head self-attention,MHSA)进一步挖掘阅读材料中观点类问题的文字线索,改进了观点类问题的自动解答方法;将句法关系融入到图构建过程中,提出了基于关联要素关系图的多跳推理方法,实现了答案支撑句挖掘;通过联合优化两个子任务,构建了基于多任务联合学习的阅读理解模型.在2020中国“法研杯”司法人工智能挑战赛(China AI Law Challenge 2020,CAIL2020)和HotpotQA数据集上的实验结果表明,本文提出的方法比已有基线模型的效果更好.
文摘This paper suggests a theoretical model for teaching syntactic relations to foreign students of English Linguistics using insights from Cognitive Grammar as developed in Langacker(e.g.1987,1991,1999) and elaborated on in Radden and Dirven(2007). The model deals with syntactic relations as participants in a prototypically structured network of events, radiating out from the experiential action chain schema(a skeletal concept of force dynamics involved in the interaction between entities)(see Langacker, 1999). This involves dealing with syntactic relations as meaning-making tools that collaborate with the verb in a regular manner manipulating the prototypical schema. The aim is to draw attention to the need to add a cognitive semiotic dimension to the teaching of sentence structure. Such a dimension exposes learners to the cognitive factors behind grammatical meaning making and may, therefore, result in a better understanding of language structure and use on the part of learners who can become English language teachers.
