互联网话语理解的心理机制与计算建模

Web discourse communicates via internet. It not only shades increasing impact on social life, language systems, and cognitive mode, but also brings severe challenges for the cognitive theories of language comprehension and the techniques of natural language processing. This project focuses on web discourse, teams up researchers from different disciplines, and takes aim at studying the cognitive mechanisms of web discourse comprehension and developing corresponding computational models. First, we will use cognitive neural experiments and computational models to explore the way humans learn new concepts and update their knowledge systems in web context, the cognitive mechanisms for analyzing discourse structure as well as comprehending semantic and sentimental information. These researches will be guided by theories from psychology and cognitive science and stick to the features of web discourse and web context. Second, under the guidance of cognitive mechanisms and taking structural analysis as the break through point, we will develop computational models for comprehending themes and sentiments of web discourse. This project will hopefully make innovative achievements in the following aspects: the cognitive mechanisms of web discourse comprehension, the use of cognitive mechanisms in analyzing web discourse structure, and the development of computational models for web discourse comprehension. This project will promote the development of language comprehension, social intelligence, and social sentiment analysis. Furthermore, the project will also provide some theoretical basis for intelligent web applications and web public opinion monitoring.

互联网话语依托互联网进行信息交流和传播。它深刻影响了社会生活、语言系统和认知模式，对语言理解的认知理论和自然语言处理技术提出了崭新的课题。本项目以互联网话语为研究对象，针对其互动性、复杂性、自然性和群体性等特点，组成多学科融合的研究团队，深入研究互联网话语理解的认知机理和计算模型，力争取得突破性进展。项目将：1）以心理学和认知科学的理论为基础，以互联网情境和话语为特色，以认知神经科学实验和计算建模相结合的方法，探索人在互联网环境中进行概念学习和知识更新、话语结构分析和语义情感理解的认知机制；2）以认知机制为指导，以互联网话语的结构分析为突破点，建构互联网话语主旨和情感理解的计算模型。项目有望在互联网话语理解的认知机理、融合认知机理的互联网话语结构分析、互联网话语理解建模等方面取得创新成果，推动语言理解、社会智能和社会感知计算的发展，对网络智能服务、网络舆情监控等具有指导和借签作用。

解码互联网话语是当前国家发展的迫切任务之一。本项目围绕互联网话语理解的认知机理和计算模型展开深入研究，取得了突破性的进展。在认知机制方面，我们通过多个系列实验深入研究了互联网话语结构如何调控注意和工作记忆，并影响话语整合和更新的过程，同时，我们以语境发挥作用的认知机制为核心，我们通过系列实验考查了语境如何调控概念学习过程和以及语境的性质如何调控主动预期和整合的过程。基于上述认知实验的发现，我们在互联网话语的结构分析和计算建模中融入了互联网话语结构和语境调控下的注意和工作记忆等认知成分，充分考虑话语整合和更新中不同认知成分的作用机制，提出了多种融合认知机制的话语理解算法。项目执行期间团队成员产出丰富研究成果：发表文章101篇，完成专著2部，申请/获批专利6项，提交技术报告3项，软件著作权1项。这些成果将大大推动互联网话语理解语言理解、人工智能和社会感知计算的发展。