摘要
自然语言处理发展过程中一直存在着基于规则和基于统计2种研究方向,2种研究方法在很长一段时间里都停滞不前,直到人工智能的进步带来了数据、计算能力和算法的巨大变化,从而推动了自然语言处理的发展.近年来,自然语言处理在语义分析、知识库建设和文本处理方面都有很大程度的进展,大部分进展源于深度学习带来的性能提升.本文梳理了深度学习驱动下自然语言处理的进步,系统地分析了进步的原因,指出深度学习本身的局限性.最后指出自然语言处理面临的挑战,并对该领域未来的研究方向提出了展望.
Natural language processing has always existed both rule-based and statistical-based research directions in the development,both approaches had stagnated for a long time until the advancement of artificial intelligence brought about tremendous changes in data,computational power and algorithms,which has driven the development of natural language processing.In recent years,natural language processing has made great progress in semantic analysis,knowledge base construction and text processing,with most of the progress stemming from the performance improvement brought by deep learning.This paper reviewes the progress of natural language processing driven by deep learning,systematically analyzes the reasons for the progress,and points out the limitations of deep learning.Finally,it is pointed out the challenges facing natural language processing and put forward the prospect of future research in this field.
作者
王飞
陈立
易绵竹
谭新
张兴华
WANG Fei;CHEN Li;YI Mianzhu;TAN Xin;ZHANG Xinghua(Information Engineering University of Luoyang Campus,Luoyang 471003,China;91709 Units,Hunehun 133300,China;68002 Units,Lanzhou 730058,China)
出处
《武汉大学学报（工学版）》
CAS
CSCD
北大核心
2018年第8期669-678,共10页
Engineering Journal of Wuhan University
基金
国家自然科学基金项目(编号:11590771)
关键词
自然语言处理
深度学习
神经网络
机器学习
语义特征
natural language processing （NLP）
deep learning
neural network
machine learning
semanticsfeature
