Personalized And Active Service Mode For Educational Resources

It is of great significance for every citizen to share high-quality educational resources so as to improve education quality and promote educational equity. In the context of the vigorous development of online education and the continuous advance of co-construction and sharing strategies, educational resources are getting diversified in form and richer in content with highly-improved services. However, due to the lack of initiative and personalized resource service mechanism, massive educational resources leads to information overload and learning trek, resulting in an awkward situation of more resources with higher difficulty of obtaining. Learners with different levels of information literacy vary a lot regarding the ability of acquiring educational resources. High information literacy learners can get personalized educational resources from the intricacies of big data, while it is extremely difficult for low information literacy learners to do so. Different abilities of obtaining educational resources will inevitably lead to different abilities of processing and using educational resources, which will then bring about new problems of educational inequity. The improving of information literacy lags far behind the growth of resources. Regardless of information literacy levels, learners have to invest much time and energy to search for the educational resources they need. In order to make learners of different levels access easily to high-quality educational resources, it is necessary to change the transfer mode of educational resources and introduce personalized active push service to deliver educational resources catering to the learners.In this study, a personalized active service system is established to enable learners to obtain personalized educational resources and services conveniently. The system includes three components, namely learner model, resource model and recommendation engine. On the one hand, actively pushing personalized educational resources to learners can greatly improve resource accessing efficiency, narrow knowledge gap to some extent, sufficiently meet learning needs and promote balanced individual development. On the other hand, the personalized active service system can enhance the exposure of resources, increase resource utilization ratio, and thus achieve a kind of educational equity which emphasizes efficiency priority. The major work of this research includes:(1) The construction of the learning model:On the basis of adequately investigating the properties of existing learner models, the research summarizes the relative features which influence learners' selection of resources and establishes a collection of candidate learner features. Then the study adopts Delphi method to determine the major features impacting resource selection most, and figures out the intrinsic relationship between learner features through the method of interpretive structural model to construct a hierarchical interpretation structure model. Based on the model, the core factors which influence resource selection are extracted, and a learner model is constructed from three aspects, namely learners' basic features and preferences, learning behavior record and learning situation. Finally, the study uses high-dimensional tensor to represent learners' dynamic features multi-dimensionally.(2) The construction of the educational resource model:Educational big data are characterized by being multi-sourced and heterogeneous. Efficient and unified representation method of educational resources can effectively promote developing quality and communication effect of educational resources. The study analyzes the advantages and disadvantages of representing educational resources with meta-data in-depth, introduces social annotation to generate educational resource tags, and adopts social network analysis to cluster the tags to get the hierarchical structure. The study proposes the method of establishing educational resource model by combining meta-data, social annotation and social network analysis. Meta-data framework restricts the scope of annotation. Social annotation discovers resource properties from different dimensions dynamically and comprehensively to find learners' needs timely. Social network analysis clusters the annotation results at a deep level. Finally, the study constructs an educational resource representation model based on tensor to multi-dimensionally and dynamically represent the properties of educational resources.(3) Personalized active service model and strategies of education resource:This study analyzes the educational information service demand in the big data environment, and proposes an education resource service model which embeds ubiquitous learning process based on the characteristics of ubiquitous learning and the four stages of cognitive process of learning. The study probes into the concept the four elements of the model and their mutual relationship, and dissects the features and significance of the model. A learning situation in which services are pointing directly to individuals can better meet the diversified and personalized demands of learners, and thus make individualized instruction possible. Accurately pushing the most suitable educational resources to learners can maximize the benefits of resources, improve resource quality and utilization ratio, and accelerate the flow, evolution and increment of knowledge. Then the study explores personalized active service strategies of educational resources from three aspects:dynamically update learner model to comprehensively obtain learners'actual demands, accurately represent resource model to actively build high-quality educational resources, flexibly select service strategies to provide real-time adaptive educational resource service.(4) The implementation of personalized active service of educational resources:This study proposes a learning-process-oriented and two-staged educational resource recommendation method. The first stage mainly adopts rule-based recommendation method. An educational resource recommendation system is established by designing the mapping rules between learners and resources so as to solve the problem of cold start. The second stage realizes individualized recommendation in certain context mainly based on learners' learning behavior record. The power of tensor in representing high-dimensional data is fully employed to construct learner-resource fusion tensor by tensor connection operations. Moreover, tensor-based higher-order singular value decomposition algorithm is applied to conduct multi-dimensional correlation analysis upon high-dimensional spatial data to offer an integrated solution for personalized educational resource recommendation and learning community recommendation. Through simulation experiment, the study verifies the recommendation performance of the algorithm on data collections of different scales, and compares it with the classical collaborative filtering algorithm from three evaluation indexes, namely precision ratio, recall ratio and F value. Results show that tensor-based higher-order singular value decomposition algorithm outperforms collaborative filtering algorithm regarding all the three indexes.This study takes the advantage of tensor in high-dimensional and multi-dimensional representing and proposes tensor-based learner model and resource model. It enables comprehensive and multi-dimensional representation of learners and educational resources, and enriches theoretical achievements in the researching field of learner model and resource model. It adopts interpretive structural model to discover the hierarchical relationship between learner features, which provides new reference for learner model building method. The study proposes an M-S-S resource model which integrates meta-data representation, social annotation and social network analysis to describe educational resource properties so as to improve the accuracy and comprehensiveness of learning resource representation. It introduces tensor-based higher-order singular value decomposition algorithm and conduct high-dimensional correlation analysis upon learning behavior record to excavate the complicated relationship between learners and resources, and thus realize personalized recommendation in certain situations. The study conducts high-dimensional clustering on learners and resources from different perspectives to form learning communities and resource clusters, and thus realize recommendation of learning companion in different situations. This study enriches the theory and practice of personalized educational resource recommendation.