This paper firstly constructs the multi-interest feature library from readers’ interest lexicon and index with update algorithms combining gradual forgetting strategy and sliding window, then calculates the similarity measures of readers’ interest lexicon and index with books, and adds the two similarity with linear superposition to propose an operable and extensible hybrid recommendation algorithm. This algorithm synthetically uses the index types of books in Chinese Library Classification, and effectively solves the problem of data sparseness. Finally, the paper achieves a personalized recommendation system of the library books, and correlative experimental results are introduced in details.

This paper proposes a new Web page type relevance judgment strategy based on several statistical characteristics of Web document types to meet the online classification lightweight design requirements of focused crawler. Using the API provided by WEKA, this paper devises appropriate training algorithm and classification algorithm for the relevance judgment strategy. The experiments of classification accuracy, efficiency, and attribute selection demonstrate the validity of the relevance judgment strategy and five Web page statistical characteristics playing a key role in the type identification.

This paper briefly introduces the current research status of automatic MeSH subheading attachment both home and abroad. After systematic research, a “Jigsaw puzzle”–statistical method combined approach is proposed, which is suitable for dealing with Chinese biomedical literature. Moreover, the realization mechanism of this integrated method is analyzed in details and the corresponding results are evaluated. In addition, suggestions expected to improve the practical value of main heading/subheading pair recommendation are further raised.

Using the real data of BibSonomy.org for the study, this paper collects two echelon of the related tags through the “Related Tag” column of website. The formal concept analysis approach is employed to build concept lattice of related tags, and the clustering analysis and mining of association rules are executed based on concept lattice about related tags, then the implicit relationship between the concepts and potential semantic relevance of related tags are revealed. The structural features implicit in the emerging methods of navigation in the related tags are analyzed, and “Semantic Synonym” and “Semantic Hyponym” are proposed, which improve the tag navigation system of folksonomy knowledge organization mode.

In this paper, the model of using rhombus-thinking method to solve retrieval problems is introduced into the Ontology-based concept retrieval model. This paper gives the algorithms of keywords’ matter element divergent expansion and extended words’ quantify convergence, and designs the experiment to prove that the model can improve concept retrieval recall rate and precision rate, and resolves the problem of synonym and polyseme in terms.At the same time, the Ontology-based retrieval recommended function is realized.

This paper elaborates the inherent characteristics of the Chinese-English mixed text in the field of life sciences. On basis of the cybernetic principle, a new integrated indexing control method, which is conducted to improve the accuracy of automatic indexing for Chinese-English mixed text, is proposed in this paper. The method includes three relatively independent and interdependent parts, which are feed-forward control, in-progress control, and feed-back control. Subsequently, the three parts and their integrated application effectiveness for improving the indexing accuracy are introduced in detail. Experimental results show that the proposed new method is successfully applied in indexing of the literature and knowledge database for Hepatitis B subject and the information indexing accuracy is greatly improved.

P2P community is a set of nodes with similar interest, which can help to improve the efficiency of resource sharing and found. According to the characteristics of the different nodes in P2P network, community can not provide high resource sharing rate,a self-organized P2P community model is put forward based on trust. The main contribution of this paper is the trust establishment, trust value storage and trust value calculation. With the trust value changing and updating, P2P network can show different community structures.

Based on PAT array,introducing LCP array to count the length of the common prefixes of text suffixes, a new algorithm without thesaurus is presented for extracting high-frequency words of Chinese text by scanning LCP arrary.The algorithm does not depend on segmentation dictionary and can extract any repeated string,especially the new words and combined words.Experimental results show that high-frequency words extracted by the algorithm achieve a high acceptance rate and this algorithm is more effective in extracting combined words than ICTCLAS.

Aiming at data sparsity and malicious behavior in traditional collaborative filtering algorithm, this paper presents a new algorithm of collaborative filtering based on social network. Depending on social network information, the algorithm integrates user’s trust and preference in order to find the nearest neighbors of the target user, which the algorithm uses to compute weight of neighbors and to form item recommendation. Experimental results show that the algorithm can alleviate the sparsity and malicious behaviors problems and achieve a better prediction accuracy than traditional collaborative filtering algorithms.

The paper proposes users influence index model based on Sina micro-blog platform. Firstly, new users’ attention is used to replace the current false fans value, then the model uses this reasonable value to calculate users’ active index and micro-blog influence, at last it obtains the user influence by combining users’ active index and micro-blog influence reasonably. The model studies many factors of two aspects of users and users’ micro-blogs. The experiment result shows that the model can reduce the false fans’ interference and reflect the users’ real influence ability.

This paper analyzes the characteristics of experienced users’ bookmarks and tags on social bookmarking websites in Web2.0 ages. The research methods used in this study include content analysis, Q method and interviews. Clustering analysis of the test subjects is also conducted to understand the relations between user types and tag selection behaviors. Through the analysis of the diverse factors affecting the test subjects and their tagging orientations, the paper categorizes four social types of users and further discusses the characteristics of sociality and the implications for related system applications.

This paper designs and implements an important part of the infrastructure cloud - online storage system of the cloud storage system. The system follows the REST architectural pattern, and uses cURL, HSQLDB and other open source components to achieve the creation, upload, download and delete of directories and files and other functions.

The inconvenient information access and higher information-push cost are the main problems in the current commerce information push methods in M-commerce. To resolve these problems, this paper introduces a proposal to design and implement a new commerce information push service based on Google C2DM.The requirements analysis and technical ideas, the mobile client implementation, the server-side implementation and the service operating results are elaborated. The new service can help E-commerce businesses accurately push commerce information to customers in low-cost, and can help customers using smart phones to receive and access to commerce information conveniently, and also can save battery power and data traffic.

Due to the language and semantic obstacles when accessing to digital resources in library, this paper uses an open source scripts jQuery to design a browser bookmarklet which is based on cloud translators and dictionaries, and then discusses the feasibilities and technical issues of the proposal in depth. Experiments show that the bookmarklet is practical and full of user experience.

Based on circulation logs of library automation systems, the paper builds a data warehouse using the method of cluster analysis in SPSS data mining tools, and mines the classification modes between readers and books. By the classification result sets of new books personalized recommendations, new books are recommended to readers who are interested in. The paper also takes MELINETS II as an example, designs and achieves the application of the personalized recommendation system.