This paper discusses and analyzes the current research status from the following three aspects: the construction process and lexical source of the matrix; the definition, differentiation and extraction of the technical and effect words in patents; words clustering and the generation of matrix structure. Then, it points out the problems and challenges of current research.

The paper proposes some methods of definition extraction for concepts in the comprehensive OntoThesaurus. They achieve good experiment effects and are applied to the actual OTCSS. Among them, an integrated algorithm named “two-dimensional relative quantity” based on “high-frequency words vector”and “TF*IDF vector”is presented. This algorithm can much effectively extract good results from that of the first two methods, and the effective information improving ratio can reach 60% generally.

The paper studies on the key technology for realizing the semantic expansion of institutional repository based on linked data. Taking the CASIR for example, the authors extract the entity relationship of the CASIR and add some constraint rules to it, use the D2R to carry out the knowledge representation and semantic annotation as RDF format. Finally, it can expand the CASIR to DC、FOAF、ISWC、VCARD, and so on. In fact, the method lays a good foundation in making CASIR provide not only the basic service but also a real semantic integration service with its rationality and feasibility in the future.

This paper firstly analyzes the relation between search engine evaluation and usability analysis. Based on the characteristics of the three stages of information retrieval by search engine, the demand of usability evaluation is given,and the evaluation hierarchical model on the usability of search engine is constructed based on the four foundations for the choice of evaluation indicators. Then the weights of indicators are accounted by Delphi method,and a valuable search engine usability evaluation system is provided.

Using the methods of association analysis and clustering in the field of data mining, the paper focuses on the theories and methods of discovering user interests and points out the limitations of standard Web log. So it proposes a method of customized Web log in order to enhance the precision of user interests and preferences. The outcome of experiment shows that,by the method,Web log data hidden in the association rules as well as interests and preferences of similar users can be found, the precision of filtering user interest can be improved at the same time.

Based on the query logs and news data provided by Sogou Lab, this paper studies judging queries involving implicit temporal intent and recognizing the implicit temporal attribute of these queries,and it also constructs the retrieval model of implicit temporal intent of queries.The experiment results show that the precision of temporal attributes recognition is 85% and the retrieval model can improve the effectiveness of ranking results.

Aiming at the disadvantages of insufficient mining and analysis of readers' information needs existing in the active book recommendation service of university library, the paper brings forward a construction of personalized book recommender system based on collaborative filter. The system imports the factors of faculty, role, education and the readers' records of visiting the reading rooms to construct the reader's characteristic model. By mining and analyzing the characteristic model which uses optimized collaboration filter algorithm, the system can produce the personalized book recommendation to reader.And the experiment proves that the system is efficient and practical.

Concerning the issues of the more and more title party news in the Web,this paper presents a new algorithm of title party news identification. Firstly, it analyzes the composition of the news page, then puts forward an approach of news title extraction and information extraction based on the features of news page. Secondly, considering the problem of extracting coherent topic sentences from news pages, starting with the relationship matrix of sentences, it puts forward an algorithm of topic sentence extraction. Then, according to the extracted news title and the candidate set of topic sentences, it can compute the similarity value, which is the main basis for judging the title party. Finally, the experiment results show that this method is effective and feasible.

Aiming at the friend recommendation in online social networks, this paper tries to fuse multiple social networks into one mixed graph on which the random walk with restart is implemented to provide personalized friend recomendation for users. The different roles of these networks are adjusted by parameters. Experiment demonstrates that this algorithm can improve the accuracy of friend recommendation in online social networks.

The concepts and algorithms of three types mediation roles including betweenness centrality、brokerage structure and structural hole in assignees citation network are analyzed, and the functions of mediation roles to assignees technology innovation are deducted theoretically based on the mediation theory and knowledge communication theory,then three hypotheses are proposed. Through empirical study,it is found that the influence degrees are different:the influence of betweenness centrality to assignee technology innovation is the strongest,then the five brokerage indexes,and the weakest influence degree is structural hole.

From the view of users, the article designs and implements an image management system model involving the function of acquisition, processing, storage and retrieval of digital images for the public users. It also discusses the effect of the number of marked words for the precision and recall of semantic retrieval. The experiments show that the system model can meet the needs of the public users for digital image management.

In order to make patrons get the latest information of electronic resources timely and grasp the trend of academic research quickly, Tsinghua University Library designs and develops the MMS push service for electronic journals' latest table of contents. It synchronizes the information by calling the Web service interfaces from the electronic journal navigation system. It also implements the functionality of sending the latest table of contents information of electronic journals to subscribed patrons via MMS based on the combined strategy of fix-date and threshold of unsent articles.

This paper investigates the accessibility of library websites based on WCAG2.0 information accessibility and finds that the overall condition of the library Web accessibility needs to be improved. Some solutions are suggested for the outstanding and easily neglected problems in the research in hope to raise the standard and level of the library Web accessibility in China.

Due to the uncommon words of catalogue data, English words and network vocabularies, this paper studies the application status, demand analysis, design ideas, and realization methods of the zoned word translation function of OPAC. By selecting, editing, saving, translating,and updating the difficult words through the technologies of JavaScript, CSS, ASP and database, the function of the zoned word translation is implemented, which can enhance the availability of OPAC, and improve the user experience.

This paper introduces the optimization of the open source virtual machine software KVM in Shenzhen University Library,and proposes problems and solutions during this practice. And this structure could not only improve the performance of virtualized systems,but also have the merit to build a highly-availability server.