Firstly, this paper introduces the background and the effect of linked data in recommendation system, summarizes similarities and differences between the recommendation system based on linked data and the traditional recommendation system. This is to help readers understand the cause and application background of the recommendation system based on linked data. Secondly, this paper systematically analyses the main method of recommendation system based on linked data on basis of the general classification of recommendation system and detailed introduction of concrete application examples.

In the ubiquitous environment, personal data is usually distributed in different devices such as PCs, mobile devices and servers, thus achieving its integration management is of urgent need. Data synchronization is the key to complete this process. This paper proposes a method to realize personal data synchronization through exchanging the operation logs, and then illustrates its principle, technical ideas and basic processes. At present, this method has been validated in e-Click system.

This paper introduces the research and practice conditions of the Ontology mapping service provided by United States National Center for Biomedical Ontology (NCBO), analyzes the rationale, classification and method of NCBO Ontology mapping, realizes Ontology mapping between MedDRA and ICD with the NCBO project. Then Java parsing for the semi-structure XML results is used to obtain the semantic computatioin-enabling structured data. This case study showes the application value of NCBO-based Ontology mapping. The research can provide the reference method of data transformation and data basics for semantic processing, specifically semantic computation and data mining in semantic interoperatability, and can provide beneficial reference for Ontology fusion and semantic interconnection research in other fields.

For Chinese text information retrieval, a staged and integrated similarity algorithm of text is proposed, which processes sentences, paragraphs and the whole document stage by stage. The algorithm combines the topic and application ranges of document, and the corresponding weight is given to the feature words via the weighted calculation method with the semantic enhancement. Moreover, these weights are integrated into the calculated factors of the text semantic with the characteristics of each calculation phase, respectively to reach the aim of finding a more accurate similarity calculation results for Chinese text similarity calculation. Finally, a text similarity computing system is built and the improved algorithm of the system achieves better experimental results comparing with the traditional algorithms.

When the count of one class is much more than another class's, the result of IDF in TF-IDF goes the wrong way according to its design idea. This paper solves the problem by using probability to change TF-IDF algorithm. In the end, the experiment proves that the solution mentioned above is good at classifying webpage text through a simple way to cumulative sum the value of characteristic words and the speed is faster and the accuracy rate is promoted.

This article focuses on the main object that users are most concerned in a video frame, proposes a visual feature extraction of main object based on the region segmentation, and validates the mining effect using the proposed feature. The result shows the better mining accuracy using the new feature, and indicates the validity of the method.

Social media generally shows the network of users, which contributes to the rapid formation of the user community. The definition of social media communities and its characteristics are analyzed and summarized firstly in this paper, then the main research contents and progress on community detection are introduced from the five categories of social media networks. Finally, it is concluded that existing studies focus on social media community detection algorithm, statistical analysis of social media community structure and temporal analysis of social media communities, and the existing problems and future research directions are pointed out.

On the basis of the measured frequency distribution of China's inter-regional co-authored papers covered by 10 years' data of 150 economic journals in CNKI database, this paper shows the pattern of China's Inter-Regional Research Collaboration(IRRC) by using statistical softwares, such as SPSS and GeoDa. The results fall into three groups. Firstly, regional scientific productivity affects the collaborative preference. Secondly, geographic distance is an important factor determining the pattern of IRRC. Thirdly, regional collaboration strength increases as the regional productivity increases, and as the distance between two regions decreases.

Based on common interest and purpose, a virtual community in P2P networks is formed easily by a set of peers. Based on the brief summary of the existing research of P2P communities formation, this paper analyzes its formalization definition and structure, describes the representing method of user interests, selects the calculation method of the interest correlation between peer nodes, and then by means of hierarchical clustering method and K-means clustering method to study the P2P community formation process, so as to provide reference for further research of P2P communities.

The common knowledge mining is an effective way for the Internet public opinion analysis. This paper builds the common knowledge base for the common sentimental elements and the common language patterns. The common sentimental knowledge is mined by the semi-supervised method from the training corpus. This knowledge base is also quantified and dynamically expanded. The common language pattern knowledge includes three kinds of fixed models, such as transform model, extreme sentimental verb model and distance dependency model. Finally the common knowledge bases are testified in the domains of religions and hotels, and proved the effectiveness in the system implant performance.

Aiming at the limitations of existing analysis methods of competitive intelligence demands, this paper puts forward a new method for the analysis of competitive intelligence demands for enterprises, the data obtained from questionnaire survey and the method of concept lattice theory are utilized. It builds the concept lattice of competitive intelligence demands and extracts related association rules for enterprises. On this basis, the paper analyzes the competitive intelligence demands of enterprises,and reveals the characteristics of competitive intelligence demands and the correlation among demands, the research conclusions are drawn finally. The research not only provides the domestic related theory and practice of competitive intelligence with references, but also offers some advice on identifying the competitive intelligence for enterprises, government and scientific & technical information institutions.

This paper promotes a method which collects the non-taxonomic relation from the Chinese patents' texts. Firstly, it analyzes the syntax of abstract texts, then constructs the sub-sentences extraction rules by domain sentence,character sentence, module & craft sentence and effect sentence. Secondly, artificially labels the terms of sub-sentences by label symbols such as BIEO, creates a scale of training data set. Thirdly, learns the training data and extracts the new data by CRFs. Finally, analyzes the experiment results and verifies the validity of the method.

Heterogeneous digital information resource integration is a major problem in library currently. Taking Drupal and its expansion module as the basic framework, this paper proposes a novel information sharing approach in multi-sites with low-overhead, high-efficiency and single sign on features, presents the way of gathering information in need using modules-Mashup, and implements a prototype of new generation Mashup platform (Multi-Sites Drupal Mashup) upon Web2.0, which can show data source from heterogeneous system and realize cross-platform sharing among internet cloud, library resources, reader information and interaction with reader information. Library information service ability is improved using Mashup.

One problem SULCMIS OPAC face is displaying only detail information for a bibliographic without some information such as catalog, abstract and so on. Making full use of the advantage of e-reading can solve this problem. This paper proposes an integration solution for SULCMIS OPAC based on e-reading's Open API. Finally SULCMIS OPAC is integrated with catalog, abstract, try-read and full text of e-books service with technologies of HttpWebRequest and XML, which can enrich the service contents of OPAC and improve the service ability of library.

This paper introduces how to design and implement Tsinghua smart seats management system to meet seats management reqiremants of Humanities and Social Sciences Library of Tsinghua University. Based on Tsinghua University Campus Smart Card System and Library Access Control System, this system implements self-service of seats, including reserving, choosing, querying, and exiting. It can tell whether a reader holds a seat unfairly when he gets through library access control system, that helps improve library seats' usage largely.