Aims to fully understand the recent research focuses in domestic and foreign academic blogs, further expanding the theoretical connotation and denotation of user behavior, knowledge dissemination, knowledge services. [Coverage] More than 60 papers at home and abroad are searched via "academic blogs" as keyword from CNKI, Wanfang Data and Web of Science. [Methods] Based on the previous study, this paper investigates the current researches of user communication and knowledge content itself at two levels, points out the main problems and deeply analyzes them from three aspects including user research, knowledge dissemination and empirical research. [Results] This paper proves the necessity and feasibility of the construction of subject communication model based on academic blogs from the dimensionality of users and knowledge, drawing more attention to the tacit knowledge dissemination mode in academic blog community. [Limitations] Only analyzes and summarizes the researches of academic blogs from the two perspectives, user communication and knowledge content, while the way of knowledge dissemination is constantly getting improved with the development of the research of academic blogs. [Conclusions] It's necessary to focus on the mining and expressing of tacit knowledge in academic blogs (to transform the necessary tacit knowledge to explicit knowledge), to study on the dissemination channels of tacit knowledge, and to explore the dissemination mode of tacit knowledge.
[Objective] By analyzing the academic blog recommendations and comments, this research explores interdisciplinary interaction in the online academic community. [Methods] The whole data set in 2013 of ScienceNet.cn Blog is collected firstly. Then the overall network, the local network, and the individual network about disciplines recommendation networks and disciplines comments networks are analyzed by social network analysis method. [Results] The research discovers that recommendations behavior is more frequently than comments behavior on the interaction mode. On the centrality of interaction network, disciplines recommendation matrix has significant differences from disciplines comments matrix in the in-degree centrality, out-degree centrality and in-closeness centrality, and there is no significant difference from disciplines comments matrix in betweenness centrality and out-closeness centrality. On the interaction parameter correlation, the correlation of disciplines recommendation numbers and disciplines recommendation network parameters is similar to the correlation of disciplines comments numbers and disciplines comments network parameters. [Limitations] The deep explanation of interaction is lack between discipline nodes with high degree centrality and the other nodes. [Conclusions] The academic blog is diversified and has a low threshold, large-span effect of exchange between scholars of different disciplines.
[Objective] Try to combine information recommendation and co-occurrence into a new informational relation, namely information co-recommendation in online academic blogs. [Methods] Taking ScienceNet.cn Blog as an example, use network analysis as the basis of quantitative analysis to explore the features of co-recommendation in academic blogs. [Results] The empirical research of ScienceNet.cn Blog shows that compared to the other types of networks, the case has the structural characteristics of high cohesiveness, active interaction and balanced strength; the network takes node group as the network core, and the relative balance occurs in the core group. [Limitations] Co-recommendations have different motivations and functions in different application fields. However, this paper only gives an empirical research on ScienceNet.cn. [Conclusions] The co-recommendation can be an option for future studies of academic blogs. This behavior presents more equality in the structure.
[Objective] This paper studies the features of h-degree in recommendation network of academic blogs. [Methods] Based on the datasets of blogs in ScienceNet.cn in 2013, construct the recommendation network of academic blogs, calculate the h-degree and related measures, and enter discussion by information visualization. [Results] In recommendation network of academic blogs, the generation of nodes with high h-degree is not only caused by academic knowledge connotations which are held by the information source (bloggers), but also because of the interest from topic the information source provided. This paper explores an approximate functional relationship (NA=b×hA2) between h-degree (hA) and node weighted degree (NA). Nodes with high h-degree typically become the organizer of subgroup in the center of a network. [Limitations] H-degree is not a perfect indicator, and the future studies will expand the improved h-degree. [Conclusions] H-degree can be one of the measurements for recommendation network analysis of academic blogs, and h-degree is also important for community management of this kind community.
[Objective] To analyze the research status and future research direction of Entity Resolution (ER) over relational databases. [Methods] Systematical researches are made on the accuracy and efficiency aspects of ER. The accuracy of ER is based on incremental methods, statistical methods and related information. The efficiency of ER is based on blocking, string similarity and other ideas. [Results] Maximizing precision and efficiency are the main goals of ER, but the research on dynamic evolution, heterogeneity of data sources and inexact string matching still faces significant challenges. [Limitations] Only precision and efficiency in the process of ER are discussed, but the characteristics and limitations of ER model don't get the same level of attentions. [Conclusions] This paper gives a comprehensive overview of the process of ER over relational databases, research status and future research direction.
[Objective] This study aims to build a general semantic model for the organization and description of research data in the e-Science environment, providing semantic layer of data organization for building semantic computing environment in digital repositories. [Methods] Based on the analysis on organization patterns for research data, this paper designs semantic organization architecture of research data and builds Ontology models for each component by concept analysis and Ontology modeling. At last, this Ontology model is applied to the design of a prototype system and experiments are made on some application. [Results] This model can achieve relatively better result in semantic linked organization of research data. [Limitations] The semantic supporting effect of this model remains to be further validated, which needs to be based on other modules' application experiments. [Conclusions] This Ontology model can achieve semantic organization of research data to provide the foundation of semantic knowledge organization for resource building and services of knowledge platform.
[Objective] To solve the Oracle Bone Inscriptions (OBI) interpretation of ambiguity and semantic problems by constructing extenics OBI language model and OBI HowNet knowledge base. [Methods] Based on HowNet framework, a semantic knowledge base combined with OBI and modern Chinese is proposed. The OBI extenics language model is introduced and fusing the HowNet knowledge representation method, and it can implement semantic similarity computation and related extension strategy application. [Results] Small-scale experiments show that the method can reach more than 90% on OBI of the correctness of the definition of ambiguity recognition, and more than 75% on OBI for the residual alteration of the righteous. It can effectively solve the problem of OBI interpretation. [Limitations] The limit of OBI HowNet knowledge base scale and extension strategy maturity restricts the semantic information of automatic understanding on OBI. [Conclusions] The OBI information processing research is based on extenics OBI language model and OBI HowNet knowledge base. It provides a new solution for OBI characters semantic deduction and incomplete OBI rubbing text integration.
[Objective] This paper is to analyze evolution path of network public opinion in the emergency management of unexpected events, and discover the relation between evolution of netizen crowd behavior and public opinion of unexpected events. [Methods] Design a Multi-Agent model which involving Agent properties, interaction and game rules among Agents, cross validation method between online and offline, and simulate this Multi-Agent model based on NetLogo system. [Results] With an empirical study, the feasibility of the Multi-Agent is verified. [Limitations] The interaction and game rules of Multi-Agents need to be optimized based on more empirical study in special domain. [Conclusions] Agent-Based Modeling can combine netizen crowd behavior and real environments for modeling and simulating, and can discover the inner rule of the public opinion evolution in the unexpected events.
[Objective] Make a comparison between three typical mobile reading applications, including iReader, DuoKan and Kindle. [Methods] Design evaluation metrics and a user experiment for mobile reading applications. After finishing the reading assignments, participants use a five-point scale questionnaire to make ratings based on their own experiences. And statistical methods are used to analyze the results of the user experiment. [Results] DuoKan is the best in nine indicators, including the layout of the interface, the experience of probation, the reading function, the communication, the user friendly and easy to use, the aesthetic design, the practicability, the necessaries, and the overall experience. iReader is the best in four indicators, including the installation process of applications, the individual center, the response speed, and the experience of paying. And Kindle is the best in the indicator of content quality. [Limitations] There is lack of different types of enough users who participating the experiment. [Conclusions] DuoKan is the mobile application with the best user experience. To improve the user experience, the mobile reading application needs to refine the classification of users, to enhance the quality of its electronic books and to improve the aesthetic design.
[Objective] This paper aims to analyse user behavior based on search engine log. [Methods] Analyse user behavior from query string, query methods, query subjects, user click behavior and user types by word segmentation, statistical analysis, clustering analysis and visualization. [Results] Search users prefer to use 2-5 Chinese noun phrases; Use less colloquial query strings; Dislike using advanced search functions; Perfer to use various query strings; There are peaks and valleys in the number of users. Up-tail phenomenon is confirmed once again in this research. [Limitations] The amount of data used in this paper is not big enough and details of user information is not considered. [Conclusions] Analysis on search engine log is beneficial to acquisition of user behavior characteristics and improving search performance.
[Objective] This study targets to improve the ability to discover research layout of funding agencies based on their funding application documents. [Methods] The K-means++ clustering method is proposed to analyse the funding direction and main focus based on multiple sources of funding application documents. [Results] After validation and a case study based on the funding application abstracts from NSF and FP, it is discovered that single-word feature is more accurate than multi-word feature in the K-means++ clustering. If only keep the essential contents of application abstrcts as analysis documents, the accuracy of the K-means++ clustering is significantly improved. [Limitations] Data cleaning of the funding application documents is not fully automated. Adjustment of clustering parameters need to be manually controlled. [Conclusions] The K-means++ clustering of funding application documents is a practicable method by validation and case study. Research layout differences in agencies' funding trends are discovered and could be helpful for scientific management and policy decision.
[Objective] For complaint text has the characteristics of informative, unstructured, weak regularity etc., the current information management of city complaint needs an efficient classification method to improve the efficiency of the management staff.[Methods] Analyze the characteristics of complaints and go for text preprocessing; Then use the parser, synonyms forest, and through the contribution of the document to filter guide word; At last, calculate the guide word weighting coefficients with TF-IDF, use VSM model to represent guide words and use SVM model to classify the complaint text. [Results] In multiple categories of complaint text, the average precision of the method is up to 82.1% and the average recall is up to 82.3%. [Limitations] Thesparsity of complaint text affects the classification results to a certain extent.[Conclusions] The experiment results show that the method is effective and feasible in the text classification of complaints, and it can improve categorization effect of thecomplaint text.
[Objective] In order to help consumers distinguish high quality reviews from enormous review sets.[Methods] Using LDA topic model to classify the themes and referring to the thoughts of improved automatic summarization, this paper puts forward Subject-Oriented High Quality Reviews Mining Model.[Results] The model extracts high quality reviews automatically under each topic. The results of the experiment show that its precision, recall and F1 score reach 80.73%, 64.90% and 71.95% respectively, proving the model's effectiveness and superiority.[Limitations] Just compared the model with some typical models, but there are some other methods exist but have not been verified. [Conclusions] The model can effectively mine high quality reviews under different themes from the review sets, thus help customers in making more effective purchase decision.
[Objective] Use review sentiment feature sets extracted by dictionary matching method and machine learning method to predict review's helpfulness. [Methods] This paper adopts sentiment dictionary matching method and machine learning classification method to extract review sentiment feature sets, including building sentiment dictionary, designing appropriate matching algorithm and choosing the best sentiment classifier. Random forest algorithm is applied to predict review's helpfulness with different sentiment feature sets. [Results] The combination of two sentiment analysis methods performs best in predicting review helpfulness. Review's average sentiment score and deviation score derived from sentiment dictionary method have better prediction performance to review helpfulness. [Limitations] Only focused on reviews of search product but neglected the reviews of experience product. The research dataset is limited. [Conclusions] The combination of sentiment dictionary matching method and machine learning method can predict review helpfulness effectively.
[Objective] A multi-dimensionality reputation model, named e-BRM, is proposed to surpass (1, 0, -1) scoring system used in e-Barter (one type of emerging online C2C market). [Methods] Based on Wilson score interval and uniform distribution, e-BRM can calculate positive ratio and positive coverage ratio respectively. Meanwhile, time-delay factor, negative punishing factor and real name authentication factor are designed in e-BRM to be further aggregated as barter's transaction value. [Results] The positive ratio, coverage ratio and transaction value are aggregated as barter's reputation degree by e-BRM. The aggregated value can describe barter's true reputation degree better than that of (1, 0, -1) scoring system. [Limitations] For the application of e-BRM, an online updating mechanism should be designed for improving real-time performance. [Conclusions] Simulation experimental results show the validity of e-BRM, thus barters can make reasonable deal decisions based on reputation degree for decreasing transaction risk.
[Objective] In order to achieve the micro drilling analysis of technical efficiency map and the recognition of specific patent involved in technical efficiency map. [Methods] This paper proposes a Patent Data Warehouse-based technical efficiency map mining method, which achieves the construction and multidimensional analysis of technical efficiency map by cleaning patent structured information and extracting feature words of unstructured information, combined with the Data Warehouse. [Results] The experiment results show that this method can achieve the objective fastly. [Limitations] However, if the amount of patent data is large, the star model used may reduce efficiency. And the patent feature extraction can't be automated. [Conclusions] This proposed method provides a new way for constructing and mining technical efficiency map.
[Objective] Practice on construction of Linked Data of book bibliography data for college library. [Context] Try to publish bibliography data from college library as Linked Data to discover knowledge. [Methods] Construct a lightweight book bibliography Ontology with additional classes from social network of the most popular criterion in the view of users. Then reorganize bibliography data with Ontology, and publish Linked Data through the D2R server. [Results] Try to publish the computer subject bibliography data as Linked Data. Aggregate data through the version, languages, subject and link with the resource out of library through authorial links. [Conclusions] It is practicable that using the lightweight book bibliography Ontology to reorganize bibliography data and publish it into Linked Data through D2R. However, cataloging without authority controlling make it difficult to form the aggregated linkages based on author and title.
[Objective] Integrate My-Library System with Shanghai Library WeChat System by QR Code to improve the level of library information service. [Context] With the application of HTTP Persistent Connection, WeChat, QQ, Sina Weibo provides QR Code login service. [Methods] Using WeChat QR Code and HTTP Persistent Connection to provide Cross-Validation Service betweend WeChat System and My-Library System, so readers can login My-Library System by scanning WeChat QR Code. [Results] Combining WeChat QR Code and authentication system, readers can login My-Library System by scanning QR Code. [Conclusions] Provide a new way to login My-Library System for the reader, and also expand the application of QR Code in library.
[Objective] To address the problems of Web data collection, difficult to integrate multiple types of digital resources etc. in characteristic database construction. [Context] The life of characteristic digital resources information is short, each heterogeneous database platform in Shaanxi has great difference, supports limited RSS interface, contains complex data formats. [Methods] Using Web data collection technology such as Drupal Feeds, XPath Parser, Crawls, Image Grabber, combined with data cleaning and removing, to achieve specialization and systematization for Web data collection. [Results] Explore feeds RSS collection, HTML/XML automatic acquisition, rules for different characteristics of resource modification specially, and Web streaming media collection. [Conclusions] This study can rich platform data sources, partially provide solutions to difficult data collection, data formats unstandardized, data source route limited and so on.
[Objective] The seat reservation system software based on the WeChat public platform can extend the functions of traditional seat reservation system. [Context] Traditional seat reservation system is not applied to the field of mobile platforms. The WeChat is the main entrance of mobile terminal with a large number of users, so it is a good choice for seat reservation system. [Methods] Based on ASP.NET, this system is designed by HTML5, CSS3 and AJAX etc. Futher, embed the basic function of seat reservation system into the WeChat platform by the message interface. [Results] Users can check the seats condition, search their friends' exact positions and book the seats by WetChat. [Conclusions] This application explores the development pattern on the mobile platforms, and it has a significance for developing the WeChat public accounts.