基于角色的“注意”类心理常识的获取及表示

Acquisition and Role-Based Representation of "Attention" Psychological Commonsense

为了使机算机实现真正的拟人化,使机器具有人类的一些基本的心理常识是一个至关重要的问题。近年来,对于心理常识的研究已成为人工智能领域研究的热点,但已有的工作主要集中在小范围内,对于一些典型的心理状态或情绪的笼统的抽象分析,以及情感的计算方面,并没有对于引起特定心理状态产生的前提事件和原因做出系统的研究和考察。为了使计算机真正“理解”人类的心理,需要建立一个大规模的关于真实世界中人类的心理现象的常识性知识库。本文借鉴了心理学研究的若干方法,通过对一种特殊的心理常识——“注意”类常识的详尽探讨,从Internet上的海量文本资源中将真实世界中“注意”产生情境提取出来,经过加工整理形成一条条关于“注意”产生场景的命识,并在此基础上形成了一个基于角色的描述“注意”类心理状态成因的表示框架,同时提供了一种对常识库进行一致性检查的方法。

One of the greatest goals in AI is to let the computer behave like human, the key of which is to make ma- chine have basic human psychological comonsense. Howerer, what have been done tends to focus on general and ab- stract analysis on some basic and typical mental states or emotions, and hasn't made systematic investigation on the antecedent situation or cause of certain mental state. In order to let intelligent systems 'understand' practically the producing process of human psychology, to construct a large-scaled commonsense knowledge base about human psy- chological phenomena in the real world is necessary and urgent. This article, by discussing a special type of psychologi- cal knowledge-'attention' knowledge in detail, introduces a role-based method to acquire and formalize the contex- tual situations in which 'attention' happens in the real world, so that knowledge base can be constructed.