The lacking of studies on speech act analysis in Chinese context can be attributed to the short of speech act taxonomies.This research follows the paradigm of design science by proposing the design requirement for speech act taxonomy, and develops an iterative process that contains five steps to satisfy the design requirement. Further, the authors develop speech act taxonomy for E-learning environment based on the proposed methodology, and evaluate its discernment and generalization ability. The successful application in E-learning environment validates the effectiveness of proposed design framework for speech act taxonomy.

In order to meet the demand of Scientific & Technical Workers(STWs) acquiring personalized, systematized and dynamical Scientific & Technical Literatures(STLs),an approach of STLs collaborative recommend based context aware is proposed.The relation among context, STWs and STLs is analyzed,then the process model of STLs collaborative recommend based on context aware is put forward, and key methods in this process are studied in detail. Finally,corresponding system is built, and an experiment is conducted to verify the effectiveness of the collaborative recommend proposed in this paper.

Traditional collaborative filtering algorithm is usually dependent on single kind of user requirement to generate clustering and this may affect the accuracy of recommendation. In view of the problem, this paper proposes a personalized recommendation method in digital library based on users' polymorphic clustering. This method uses an improved Hamming distance to calculate candidate neighbors, then combines polymorphic similarity to cluster again, finally forecasts user' s requirements degree and generates recommendation. The experiments show that recommendation based on polymorphic clustering is more accurate than the single' s.

This paper selects 15 homepages of digital library for the markup check, CSS check and page speed test. This is to show the current status of digital library with the application of the Web standards and to analyze the cause of the error messages from these results. Based on that, the suggestions and measures of rebuilding digital library websites are put forward.

In this paper, the process of building comparable corpus in special domain for bilingual terminology is designed. Firstly, bilingual sample corpus in a special domain is collected, and keywords are extracted from the sample corpus based on word co-occurrence method. Then, these keywords are used to be a query to a scholar search engine, and the searched result is used to be candidate comparable corpus. Finally, the comparable corpus in the special domain is obtained after filtering noise documents by quantitative evaluation.

This paper presents a new approach to Ontology learning in the domain of information science. A combination of Formal Concept Analysis (FCA) and association rules is used to facilitate Ontology construction from unstructured text. This approach acquires key concepts from documents by using a seeding and expansion mechanism; formulates (key concept by document) context for concept lattice construction, and bootstraps the learning of domain-specific concept hierarchies using FCA; extracts the relationships between the concepts via association rules. To evaluate the quality of the learned Ontology, a comparison with “Golden Standard” is undertaken, and the evaluation results illustrate that it can reach high domain coverage and identify some implicit relations between concepts. It is concluded that the proposed method is practical and useful to support the process of building domain Ontology.

The paper researches automatic new word recognition in specialized field which is represented by phytology. A set of 200 documents on plant description randomly drawn from “Flora of China” is taken as sample set. At first, draw new words candidates are drawn by N-Gram method based on words split by ICTCLAS. Then all the new words candidates are sorted respectively by term frequency (TF), document frequency (D) and average term frequency (TF/D) and the candidates are selected among certain boundary as true new words. The experiments show that new words recognition according to TF is the best and F measurement is 0.65. This method can automatically produce user dictionary of specialized field and is highly portable.

In order to enhance the objectivity and accuracy of reputation value in e-commerce, the algorithm for eliminating differences of users' subjective evaluation is designed by using trust translation. Then a dynamic e-commerce reputation management model is also constructed. Finally, the results of experiments demonstrate the accuracy and validity of this algorithm,and the direction for future research is pointed out.

For most of the patent technology-effect matrixes are now manually constructed,thus, a method for matrix structure construction based on feature degree and lexical model is presented. The feature degree is used for improving the correlation degree of candidate technical and effect words, and the lexical model for optimizing clustering of technical and effect words, generating matrix structure. This method provides technical support and new idea for automatically generating patent technology-effect matrix.

Due to the problems in processing speed, storage space, fault tolerance, access time and others of massive data processing, this paper analyzes principle, implementation process of Google MapReduce programming model, introduces four main MapReduce implementation platforms,including Hadoop, Phoenix, Disco and Mars.Then,it separately compares them in four aspects as the programming language, building platform, functions and features, applications to have a comprehensive understanding of MapReduce programming model and it's application platforms.

Taking the Sina microblog as the object of research, this paper establishes the microblog social network structure through co-link relationship,and then tries to use the social network analysis method and Hierarchical cluster analysis method to carry out centrality analysis, subgroup analysis and the role analysis, in order to find out the relationship between the focus of attention, and give some advice to the present attention recommend system.

Based on the dataset of authoritative journal in the field of library and information science,this paper tries to apply Probabilistic Latent Semantic Analysis (PLSA) algorithm to process documents representing expertise,in order to locate the research areas of experts in this field. The experiment results show that this method is feasible and achieves good results.

Due to the lack of geographic information in national library,the paper introduces the research of bar code and two-dimensional bar code, using Android and other technology, designs and realizes library personalized services system which is based on book navigation service. It focuses on user needs,design thought and implementation of main functions. According to the assessment of the application, the system can greatly enhance the efficiency and quality of library services.

Aiming at how to setup connection between library website and SNS, the paper suggests using SNS accounts logon library website. With the analysis of SNS open platform, library users habit and library website,it uses OAuth and SDK for secondary development, which is proved effective and practical.

In order to ensure the normal operation of existing equipment that used in electronic reading room for the period while IPv6 hosts and routers co-exist with IPv4 systems, this paper introduces a method for building a modular security gateway system based on open source software and designs its main program flow. The application shows the system performs satisfactorily and works stably.