两种词义消歧方法分析与比较  

Analysis and Comparison of Two Word Sense Disambiguation Methods

作　　者：张国清 Zhang Guoqing(The College of Information,Mechanical and Electrical Engineering,Shanghai Normal University,Shanghai 200234,China)

机构地区：[1]上海师范大学信息与机电工程学院,上海200234

出　　处：《信息与电脑》2017年第19期47-48,52,共3页Information & Computer

摘　　要：为了解决中文信息处理中的多义词在特定语境中的语义辨识问题,笔者进行了两种多义词消歧方法的实验并对其适用范围和应用前景进行了分析。首先介绍了有导的利用神经网络进行的词义消歧方法,其次介绍了无导的利用Hownet义原同现频率信息进行的消歧方法。最后给出了两种方法的实验数据,并从消歧知识的获得、消歧知识的表现形式及消歧算法适用范围几个方面指出了不同方法在词义消歧方面的优缺点。In order to solve the problem of semantic identification of polysemous words in Chinese language processing in specific contexts,the author has carried out experiments on two kinds of disambiguation methods,and analyzed their applicable scope and application prospects.It introduced a neuro-network based supervised word sense disambiguation method firstly.And then an unsupervised word sense disambiguation method which uses the co-occurrence frequencies of the sense units in Hownet is introduced.The experimental results were given at the end.The acquisition of the disambiguation knowledge,the expression method of the disambiguation knowledge and the application range of these two methods and their advantages and disadvantages were discussed.

关 键 词：神经网络 词义消歧 HOWNET 义原 同现频率 

分 类 号：TP391.1[自动化与计算机技术—计算机应用技术]