中国科学院自动化研究所   设为首页   加入收藏  联系我们
 
English
网站首页     实验室概况     研究队伍     组织机构     学术交流     科研成果     人才培养     开放课题     创新文化     资源共享     联系我们
    学术讲座

How grammars help humans and computers avoid surprises

模式识别学术大讲堂

Advanced Lecture Series in Pattern Recognition

   (TITLE)How grammars help humans and computers avoid surprises

讲座人 (SPEAKER)Prof. Stephen Crain, Macquarie University

(CHAIR)Prof. Jianhua Tao

   (TIME):10:00am, March 19 (Tuesday), 2019

   (VENUE):No.1 Conference Room (3rd floor), Intelligence Building

报告摘要(ABSTRACT):

The goal of computational linguistics is to develop cognitively-plausible mechanisms that simulate the human sentence processing mechanisms, or parser. Here are two observations about the human parser. First, so-called ‘agrammatic’ aphasics with memory spans of 2-4 items are able to detect the unacceptability of certain ungrammatical sentences more than a dozen words long. However, these same patients fail to detect the unacceptability of much shorter sentences when they are followed by optional tag-endings (… did he?, … didn’t she?). Second, by age 4, children are nearly the equals of adults in judging the truth or falsity of sentences with multiple logical expressions like Every girl with a dog or a cat received a prize, but children as old as 5-6 fail to follow seemingly simple commands like Put the frog on the napkin into the box. This talk will attempt to reconcile these facts about human behaviour and recent developments in computational linguistics which demonstrate that hierarchically-based language models perform more like humans, in several important respects, as compared to language models that treat sentences as strings of words (n-grams, or LMs without an injection of syntax). The proposal is that, like human parsers, syntax-infused language models are able to comprehend sentences that do not catch them unaware, but are less able to deal with surprises.

告人简介(BIOGRAPHY):

Stephen Crain is Director of the ARC Centre of Excellence in Cognition and its Disorders (CCD). The CCD is funded for 7 years from 2011, and investigates 5 areas of cognition: language, reading, belief formation, memory, and person perception. Stephen Crain is the Program Leader of the Language Program and also Director of the International Center of Child Language Health, in Beijing China. Stephen Crain is Visiting Professor at the Beijing Language and Culture University, China, and at the Kanazawa Institute of Technology, Japan.
His research is in three areas of the psychology of language: child language acquisition, adult language processing, and neurolinguistics. In the study of child language development, he has contributed to the empirical assessment of the theory of Universal Grammar. For the last decade, his research has focused on children’s acquisition of semantic knowledge, in particular young children’s knowledge of logical expressions. Much of this has been cross-linguistic research, with a particular focus on Mandarin Chinese.

友情链接
 
中科院自动化研究所 模式识别国家重点实验室
NLPR, INSTITUTE OF AUTOMATION, CHNESE ACADEMY OF SCIENCES