1999 Volume 13 Issue 3
Published: 15 June 1999
-
Select all|
-
Review1999, 13(3): 2-9.Word sense disambiguation has been a difficult problem in natural language processing. This paper presents a method of automatically increasing new collocations by the use of the collocations provided by a machine readable dictionary《XianDaiHanYuCiHai》; In order to assuring the learning quality , the size of context was enlarged gradually ; In the procedure of learning and word sense disambiguating , author gives a multi maximal log word sense disambiguation algorithm. At last , the method was tested and proved that it has higner accurancy.
-
Review1999, 13(3): 10-18.This paper analyses the ambiguity of determining boundaries of Chinese phrases in automatic parsing by computer. The type of ambiguity can be classified from three different perspectives. As viewed from component of ambiguous structures , ambiguous phrases can be classified into two kinds : one including terminal symbols , the other not including terminal symbols but only non - terminal symbols. As viewed f rom the influence of ambiguity , ambiguous phrases can also be classified into two kinds : self - confined ambiguous phrases and non - self - confined ambiguous phrases. The influence of the former ambiguity is mainly inside the ambiguous phrases. The influence of the latter ambiguity is outside of the ambiguous phrases. As viewed from differentiated types of relation between type and token , ambiguous phrases can be classified into three kinds : the true - ambiguity , the quasi - ambiguity , and the pseudo - ambiguity. Furthermore , the distribution of these types of ambiguous phrases in Modern Chinese is also surveyed depending on the above analysis and a set of rules used for a Chinese - English Machine Translation system. The authors hope that the analysis on various types of ambiguities mentioned above conduces to solve the problem of phrase structure ambiguities in Chinese.
-
Review1999, 13(3): 19-26.This paper introduces a SVD method in bilingual information filtering. It gives an uniform presentation to bilingual documents. Then any arithmetic used in monolingual information filtering can be easily used in bilingual information filtering. Using this method , we can compress the document vector and filter the noise. This method is used in personal information filtering on the Internet . We provide the WWW Bookmark Service. Through user's Bookmark , we can get user's preference and recommend interesting bilingual documents. According to user's feedback , we can improve the quality of information filtering.
-
Review1999, 13(3): 27-33.Automatic document classification is an important application of text information processing technology. This paper presents the design and implementation of an automatic Chinese documents classification system. Especially , it discusses some key techniques such as text classification model , feature extraction and structure of dictionary.
-
Review1999, 13(3): 34-40.This paper presents a method to recognize handwritten similar Chinese characters based on combining statistics model recognition method with artificial neural networks (ANN) . This method takes advantage of the peculiarities of above two methods efficiently. It not only increases the recognition rate to similar Chinese characters , but also improves the performance of the system . The recognition rate to similar Chinese characters can be improved from 79.02% to 84.32% , about 5 percentage points improving ; the recognition rate to system can be improved about 1.3 percentage points.
-
Review1999, 13(3): 41-48.Speech - to - speech MT (Machine Translation) between Cantonese and Putonghua is a quite demanding job. It makes sense to deal with Cantonese - to - Putonghua text MT first . The text MT program is an important component of a Cantonese - to - Putonghua speech MT system. In addition it can independently play a helpful role in language communication and serve as a CALL ( Computer - Assisted Language Learning) tool for standard Chinese writing as well. The critical task in Cantonese - to - Putonghua text MT is word processing , especially Cantonese dialect words processing. And that is the forcus of the present paper. The discussion covers both dialect words recognition and dialect words translation , which is supported by a brief introduction of a word - based prototype MT program.
-
Review1999, 13(3): 49-54,66.To study the frame and methods of the evaluation system for the learner's pronunciations in a computer aided Chinese teaching program. We build up the feature data base of the standard Putonghua teaching program , and use the Kohonen self - organized neural network in the process of pattern recognition. The quantitative evaluation results of the learner's pronunciations can be got after the treatment of the evaluation system. We propose the frame and the methods of the evaluation system for the learner's pronunciation in a computer aided Chinese teaching program. The experiment results show that the design scheme of the evaluation system is reasonable. And the methods for the realization of the evaluation system can work.
-
Review1999, 13(3): 55-61.In this paper the author analyzed those technical expression of erroneous sentences from the point of view of syntax analysis and extended the concept of word-building rules to meet the need of information processing. The author ,trying to describe semnatic errors clearly and detailedly ,believes that collocation errors should belong to syntax errors.
-
Review1999, 13(3): 62-66.This article introduces the current status of Chinese search engines , analyses the problems involves with Chinese search engines , and makes a comparison with out standing search engines of foreign countries. The article also talks about the techniques that need to enhance the Chinese search engines and the development direction of Chinese search engines.
