基于语境的语义排歧方法
[作者]郑杰; 茅于杭; 董清富;
[摘要]本文针对英汉机器翻译系统 (ECMT)中的语义排歧问题 ,提出了一种根据单词与语境之间的关系以消除单词语义歧义的模型。该模型利用反映单词之间语义共现关系的知识库词典 ,对有歧义的单词作出排歧。为提高知识库的覆盖率 ,本模型在对大量语料进行分析的基础上建立起单词语义分类之间的相关程度矩阵 ,同时采用动态链表来表示和维护语境 ,给出了寻找歧义单词的最可能的语义的排歧算法。开放测试的实验结果表明本方法使语义排歧的正确率提高约 1 0 %。
[Abstract]This paper proposes a word sense disambiguating model,which is based on the word's semantic context.The model utilizes the knowledge dictionary that reflects the semantic concurrence information of English words.And the correlation of word semantic marks is obtained through statistical analysis of large real corpus.By evaluating each possible semantic context word,the most reasonable sense of the sense ambiguous word is selected.The experiment of this method proves an increase of 10% in precision.
[关键字]自然语言处理; 语义排歧; 机器翻译;
|
“信息时代的文明与古籍数字化系列讲座”在北大举行
[作者]
[摘要]
[Abstract]
[关键字]信息时代; 古籍数字化; 系列讲座; 计算语言学;
|
应用分段辨认序列频度信息的说话人确认方法
[作者]翁武斌; 方棣棠;
[摘要]本文提出了一种基于语音分段辨认序列信息的与文本无关的说话人确认方法 ,并且着重分析了其中关键因素的变化 ,包括聚类数、阈值以及判定准则的变化 ,对确认效果的影响。通过实验证明了分段辨认序列频度信息是一种非常有效的说话人确认信息 ,对于确认结果起到很好的辅助作用。同时也指出了新方法的不足和今后的改进方向。
[Abstract]In this paper,a text independent speaker verification method based on the information of candidate sequences of speech segment identification is described.And the influences caused by the changes of the key factors in the method such as the number of the cluster centers,the threshold and the judge rules are analyzed emphatically.Through experiments it has been proved that the in forma tion of candidate sequences of speech segment identification works very well in speaker verification and it is an effecti...
[关键字]说话人识别; 说话人辨认; 说话人确认; 文本无关; 错误接受率; 错误拒绝率; 分段辨认序列;
|
Java语言的中文处理问题完整解决方案
[作者]余海燕; 郑笑飞;
[摘要]本文从字符编码、编译器及运行环境三方面剖析了Java语言处理中文时出现乱码等现象的原因 ,并给出了一个适应各种平台的完整方案来解决中文的读写、网络传输和输入问题。
[Abstract]This paper discusses the reason that chinese text cannot be processed correctly in Java program from three aspects:character set,Java complier and runtime enviroment.We provide a comprehensive solution for this question which is independent of all platforms,including reading and writing files,transfering chinese by network and inputting chinese.
[关键字]Java; Unicode; GB; UTF-8; JDK; Applet;
|
基于潜在语义索引的文本浏览机制
[作者]林鸿飞; 姚天顺;
[摘要]文本浏览是伴随着因特网上日益增多的在线文本而出现的辅助阅读机制 ,本文给出了基于潜在语义索引的文本浏览机制。它吸取了潜在语义索引和概念标注的优点 ,利用潜在语义索引 ,减少词汇间的“斜交”现象 ,在语义空间上进行项与项、文本与文本、项与文本之间的相似度计算。利用概念词典将文本特征项按语义分类 ,给予层次分类以确定的含义。最后 ,实现以分层概念为基础的信息导航。
[Abstract]Text browsing is the assistant reading mechanism to help users browse the online texts.Text browsing based on Latent Semantic Indexing(LSI)is presented in this paper,and it combines LSI with concept tagging to improve the efficiency of users reading.It applies LSI to reduce the skew intersections and calculates the similarity between terms and texts based on the semantic space,it also divides the terms into several semantic classes and determines the meanings of classes.In additional,it implements the infor...
[关键字]文本浏览; 潜在语义索引; 概念标注; 特征抽取;
|
手写体汉字在特征空间的可视化分析
[作者]陈津颖; 金奕江; 马少平;
[摘要]手写体汉字特征一般在几百维以上 ,在这样的高维空间中 ,汉字样本是如何分布的 ?本文从可视化的角度对这一问题进行了探讨。论文首先给出了所选用的汉字特征的定义 ,然后对一些具有代表性的汉字实例 ,从K L变换法、线性投影法和非线性投影法三个方面 ,对汉字在特征空间的分布问题进行了可视化分析 ,结果表明 ,可视化分析可以帮助人们了解汉字在特征空间的分布情况 ,对改进识别器的性能具有指导意义。
[Abstract]The feature vectors of handwritten Chinese characters are often more than several hundred.In such a high dimension space,what's the distribution of Chinese characters? The paper discussed this problem through visualization.At the beginning,we gave a definition of the Chinese characters' features that are used in this paper.Then by using K L Transformation,Linear Projection and Nonlinear Projection,we made a visualization analysis to the distribution of some typical samples of Chinese characters in their fe...
[关键字]汉字识别; 可视化分析; 特征空间;
|
一个联机识别自然手写汉字的多分类器集成系统
[作者]黄襄念; 程萍; 彭健; 杨波;
[摘要]本文提出一种联机识别自然手写体汉字的多分类器集成模型。该模型中 ,我们把依照 0 1、WB和SO特征码设计的不同分类器进行集成 ,综合模式多种全局和局部特征 ,从汉字的多个结构层进行识别。初步实验结果为 ,识别率 98.6 %。
[Abstract]In the paper,a new multiple classifiers integrated model of online recognizing natural handwritten Chinese character is presented.By integrating three independent classification algorithms based on 01 code,WB code and SO code,combining various features into classifiers and recognizing HanZi on pattern different structure layer,our system comes at a correct recognition rate of 98.6 per cent.
[关键字]联机识别; 汉字识别; 综合集成; 集成;
|
语用的通信协调模型
[作者]程显毅;
[摘要]通信不仅仅是为了交换信息 ,它更主要的是 ,当系统出现冲突时 ,起协调作用 ;当系统需要合作时 ,起协商作用。用传统的协议机制实现这个目标是有局限的 ,为此本文基于语用预设讨论了语用在解决协调时的作用 ,并给出了一个语用的通信协调模型。
[Abstract]Communication is not merely for exchange information,but more main action is used for coordination when system occur conflict and negotiation when system need to cooperation.Tradition mechanism of protocol is limited to reach these goals.In this paper,we discuss the action of pragmatics reasoning in solving coordination and give a pragmatics model.
[关键字]协调; 语用预设; 语用模型; Agent;
|
蒙古文多变体附加成分智能化处理研究
[作者]S·苏雅拉图;
[摘要]本文以蒙古文整词复杂特征合一计算和并行处理为科学目的 ,依据算法学可计算性理论 ,基于蒙古文整词输入法编码原理及方法 ,提出了一种多结点连续判断推理求解蒙古文多变体附加成分值的智能化模糊输入生成处理方法与技术。
[Abstract]This article takes the integral computation and processing of the Plax features of Mongolian whole words as the target of research,according to the theory of computability,based on the theory and the methods of Mongolian whole word inputting,puts forward a kind of intelligent processing method and technique of multi nodes continuous judgement of the value of multi variation affixes of Mongolian.
[关键字]蒙古文; 多变体附加成分; 知识表示方法; 模糊输入生成;
|
Mul Tran多语言机器翻译系统实现的语言学基础
[作者]侯敏;
[摘要]MulTran是一个以实用化为目标的多语言机器翻译系统。该系统的特点是以一个统一的翻译处理程序控制运行四个语言翻译系统。该系统的技术核心是多功能逻辑运算分析技术 ,该技术的实现是靠建立一系列的功能函数 ,而这些功能函数是建立在诸如线性、离散性、递归性、层次性等自然语言的特性的基础之上的。
[Abstract]Multran is a multilingual MT system with practicability as its object.The characteristic of the system is that it is able to control and run four different MT systems with one united translating programme.The technical kernel of the system is the multi functional logical operation(MFLO),which is brought about by developing a series of functions,and these functions are based on the properties of natural languages such as linearity,discreteness,recursiveness and hierarchy,etc.
[关键字]多语言机器翻译; 自然语言特性; 功能函数;
|
共95页 当前第44页 