检索结果-内蒙古大学图书馆

6th International Conference on Fuzzy Systems and Knowledge Discovery, FSKD 2009

作者： Jia, Mei-Ying Zheng, De-Quan Yang, Bing-Ru Chen, Qing-Xuan School of Information Engineering University of Science and Technology Beijing China MOE-MS Key Laboratory of Natural Language Processing and Speech Harbin Institute of Technology China

ISBN: (纸本)9780769537351

Hierarchical Text Categorization refers to assigning of one or more suitable category from a hierarchical category space to a document. In this paper, we used hierarchical feature selection method and multiple classifiers for the Hierarchical text categorization task. Experiments showed that the methods we used was effective, compared with flat classification, top-down level-based approach with the multiple feature selection method, the single classifier obtained better performance;reliability function was introduction to evaluate the determine by single classifier reliability, if the reliability function got a small value, multiple classifiers were used to give the determine which category the unlabeled document belong to, compared to single classifier, Multiple classifiers achieved better performance on flat and hierarchical corpuses, and the time cost increasing is little than using single main classifier. © 2009 IEEE.

关键词： Feature Selection

来源：评论

学校读者我要写书评

暂无评论

Subgraph based multitext grammar for statistical machine translation

引用

Journal of Computational Information Systems 2009年第1期5卷 369-377页

作者： Sun, Jiadong Zhao, Tiejun MOE-MS Key Laboratory of Natural Language Processing and Speech Harbin Institute of Technology Harbin 150001 China

To acquire the alignments and projection of structures at different levels in statistical machine translation (SMT), we define Subgraph and Subgraph pairs in this paper. With Subgraphs of the parse tree, we can decorate any phrase in the phrase-based models. We propose the operations of addition between graphs, which can make the source and target parse trees generated synchronously. Using the subgraph as terminal or nonterminal symbols, we present a Syntax-based model for SMT in the framework of multi-text grammar (MTG). In the experiment, our model outperforms the state-of-the-art Pharaoh, a phrase-based model, by 1.7 Bleu points. Copyright © 2009 Binary Information Press.

关键词： Trees (mathematics)

来源：评论

学校读者我要写书评

暂无评论

A re-ranking approach for categorization information retrieval based on multiple feature selection

引用

Journal of Computational Information Systems 2009年第6期5卷 1609-1616页

作者： Zheng, Bowen Zheng, Dequan Zhao, Tiejun Chen, Qingxuan MOE-MS Key Laboratory of Natural Language Processing and Speech Harbin Institute of Technology Harbin 150001 China

In this paper, a re-ranking approach for categorization information retrieval is proposed to improve precision based on hierarchical feature selection method. This paper discusses the multiple feature selection method and designs an algorithm to implement the method. The multiple feature selection method is to select features from texts that are organized in the form of hierarchy and we can re-rank information retrieval results with these selected multiple features, In addition, this paper discusses how to use multiple features selected to re-rank information retrieval results. At last, an evaluation model is presented to prove this re-ranking approach. Experimental results showed that the method is simple, with common domains, and good results were achieved. 1553-9105/ Copyright © 2009 Binary Information Press.

关键词： Information retrieval

来源：评论

学校读者我要写书评

暂无评论

An unsupervised approach for noun resolution

引用

Journal of Information and Computational Science 2009年第3期6卷 1263-1270页

作者： Yang, Yuhang Zhao, Tiejun Zheng, Dequan Yu, Hao MOE-MS Key Laboratory of Natural Language Processing and Speech Harbin Institute of Technology Harbin 150001 China

Most existing coreference resolution techniques focus on pronoun resolution in the same document. In this paper, an unsupervised approach is presented for noun resolution in different documents. Given two raw corpora, one in the general domain, one in an application domain, domain specific terms are extracted based on the distribution information of strings in different domains. Noun coreference is resolved based on similarity calculation by using both pronunciation and context information. Each extracted term is compared with the given entities to distinguish whether it is coreferential and which entity it is coreferential to. The proposed approach requires no training, no prior domain knowledge, and no need for manually annotated corpora. This method is applicable to any domain corpus and it is especially useful for knowledge-limited and resource-limited domains. Primary experiments conducted on sports domain for noun resolution achieve good performance. 1548-7741/ Copyright © 2009 Binary Information Press.

关键词： Domain Knowledge

来源：评论

学校读者我要写书评

暂无评论

Research on query translation for clir based on a combination of statistical method and web information

引用

Journal of Computational Information Systems 2009年第3期5卷 1115-1122页

作者： Zhu, Honglei Zheng, Dequan Zhao, Tiejun MOE-MS Key Laboratory of Natural Language Processing and Speech Harbin Institute of Technology Harbin 150001 China

Query translation is an important task for cross-language information retrieval (CLIR), which aims at translating the query described in source language into target language. The approach to query translation based on bilingual dictionary is becoming the mainstream thinking because of its simplicity and the increasing availability of machine readable bilingual dictionary. However, this kind of approach faces two necessary problems that is ambiguity in translation and the incompleteness of the dictionary. This paper presents three statistical models based on HowNet to resolve query translation ambiguity of CLIR: Query translation selection based on semantic relation;Bilingual decaying co-occurrence model and Semantic decaying co-occurrence model. Through test and summarizing this paper gets the best algorithm which integrates the traits of the three models, and it gradually filters and optimizes the translation. This paper also gives an approach to resolve the translation problem of the words which are out of vocabulary (OOV), which uses web feedback information to get translation and expansion of query. 1553-9105/ Copyright © 2009 Binary Information Press.

关键词： Statistical methods

来源：评论

学校读者我要写书评

暂无评论

Opinion Analysis Based on a Fusion of Multiple Classifiers Approach

Opinion Analysis Based on a Fusion of Multiple Classifiers A...

引用

International Conference on Natural language processing and Knowledge Engineering(IEEE自然语言处理与知识工程国际会议 IEEE NLP-KE 2009)

作者： Bing XU Tie-jun ZHAO De-quan ZHENG Qing-xuan CHEN MOE-MS Key Laboratory of Natural Language Processing and Speech Harbin Institute of Technology Harbin 150001

With the rapid expansion of network application, more and more customer reviews are available on-line. In this paper, A method for opinion analysis based on the fusion of multiple classifiers was presented, reliability function was introduced to select the text that is hard to determine by the main classifier, for these texts, multiple classifiers were used to determine which category the unlabeled documents belong to by voting. Experiments showed that the performance of text classification was improved by the proposed method. Compared with single classifier, this method achieved better performance, only increasing a small amount of time than using single main classifier.

关键词： Information analysis Text categorization speech analysis Voting Frequency Laboratories Natural language processing speech processing Data mining Data analysis

来源：评论

学校读者我要写书评

暂无评论

Automatic Domain-Ontology Structure and Example Acquisition from Semi-Structured Texts

Automatic Domain-Ontology Structure and Example Acquisition ...

引用

International Conference on Fuzzy Systems and Knowledge Discovery (FSKD)

作者： Cheng Xiao Dequan Zheng Yuhang Yang MOE-MS Key Laboratory of Natural Language Processing and Speech Harbin Institute of Technology Harbin China

This paper presents a new method to acquire domain-ontology structure and examples from semi-structured data sources. Firstly, extract domain-ontology structure, including candidate attributes extraction using certain patterns and applying a statistic method to filter out the incorrect attributes. Secondly, using domain-ontology structure as a clue, automatically generate example extraction patterns. Finally, acquire ontology examples taking advantage of the special structure feature of the Web pages. Experiments are carried out in the field of film, the precision of the ontology structure extraction is 83.7%, and the highest recall of the examples extraction reaches 90%. Experimental results demonstrate that the method developed in this paper is fairly efficient.

关键词： Ontologies Data mining Natural language processing Terminology Web pages Statistical distributions Fuzzy systems Laboratories speech processing Statistics

来源：评论

学校读者我要写书评

暂无评论

Automatic domain ontology construction based on thesauri

Automatic domain ontology construction based on thesauri

引用

6th International Conference on Fuzzy Systems and Knowledge Discovery, FSKD 2009

作者： Sun, Weicong Jia, Meiying Zheng, Dequan Cao, Hongqiang Yang, Bingru Yu, Hao MOE-MS Key Laboratory of Natural Language Processing and Speech Harbin Institute of Technology Harbin 150001 China School of Information Engineering University of Science and Technology Beijing 100083 China Beijing Graphic Institution Beijing 100029 China

ISBN: (纸本)9780769537351

The research on the automatic ontology construction has become very popular. It is very useful for the ontology construction to reengineer the existing knowledge resource, such as the thesauri. But many relationships in the thesauri are incorrect or are defined too broadly. Accordingly, extracting ontological relations from the thesauri becomes very important. This paper proposes the method to reengineer the thesauri to ontology, and especially the method to how to obtain the correct semantic relations. The test result shows the accuracy of the semantic relations is 86.23%, and one is the hierarchical relations with 89.02%, and the other is non-hierarchical relations with 83.44%. © 2009 IEEE.

关键词： Ontology

来源：评论

学校读者我要写书评

暂无评论

Hierarchical Text Categorization Based on Multiple Feature Selection and Fusion of Multiple Classifiers Approaches

Hierarchical Text Categorization Based on Multiple Feature S...

引用

International Conference on Fuzzy Systems and Knowledge Discovery (FSKD)

作者： Mei-ying Jia De-quan Zheng Bing-ru Yang Qing-xuan Chen School of Information Engineering University of Science and Technology Beijing China MOE-MS Key Laboratory of Natural Language Processing and Speech Harbin Institute of Technology China

Hierarchical text categorization refers to assigning of one or more suitable category from a hierarchical category space to a document. In this paper, we used hierarchical feature selection method and multiple classifiers for the Hierarchical text categorization task. Experiments showed that the methods we used was effective, compared with flat classification, top-down level-based approach with the multiple feature selection method, the single classifier obtained better performance; reliability function was introduction to evaluate the determine by single classifier reliability, if the reliability function got a small value, multiple classifiers were used to give the determine which category the unlabeled document belong to, compared to single classifier, Multiple classifiers achieved better performance on flat and hierarchical corpuses, and the time cost increasing is little than using single main classifier.

关键词： Text categorization Space technology Tree graphs Classification tree analysis Knowledge engineering Laboratories Natural language processing speech processing Support vector machines Support vector machine classification

来源：评论

学校读者我要写书评

暂无评论

Automatic Domain-Ontology Relation Extraction from Semi-structured Texts

Automatic Domain-Ontology Relation Extraction from Semi-stru...

引用

International Conference on Asian language processing (IALP)

作者： Cheng Xiao Dequan Zheng Yuhang Yang Guojun Shao MOE-MS Key Laboratory of Natural Language Processing and Speech Harbin Institute of Technology Harbin China Automatic Information Technology Company Limited Beijing Shougang Qinhuangdao China

This paper presents a new method to acquire domain-ontology relations from semi-structured data sources. First, obtain Web documents according to the co-occurrence of concept instance and attribute value. Further, define formats of relation patterns, and extract pattern instances from Web documents, including pattern clustering and pattern combining in each cluster. Finally, relation pattern instances are applied to gain attribute values of new concept instances in domain-ontology. Experiments are carried out in the field of film, the rate of pattern incorrect-division and pattern leakage are respectively 0.19% and 1.31%, the highest precision of combined relation patterns reaches 85%. Experimental results demonstrate that the method developed in this paper is fairly efficient.

关键词： Data mining Terminology Natural languages Natural language processing Search engines Clustering algorithms Information analysis Lattices Data analysis Laboratories

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：