Dunwei Wen

The effectiveness of automatic text summarization in mobile learning contexts

Mobile learning benefits from the unique merits of mobile devices and mobile technology to give learners capability to access information anywhere and anytime. However, mobile learning also has many challenges, especially in the processing and delivery of learning content. With the aim of making the learning content suitable for the mobile environment, this study investigates automatic text summarization to provide a tool set that reduces the quantity of textual content for mobile learning support. Text summarization is used to condense texts into the most important ideas. However, reducing the amount of content transmitted may negatively impact the meaning conveyed within. Although many solutions of text summarization have been applied by intelligent tutoring systems for learning support, few of them have been quantitatively investigated for learning achievements of learners, especially in mobile learning context. This study focuses on a methodology for investigating the effectiveness of automatic text summarization used in mobile learning context. The experimental results demonstrate that our proposed summarization approach is able to generate summaries effectively, and those generated summaries are perceived as helpful to support mobile learning. The findings of this work indicate that properly summarized learning content is not only able to satisfy learning achievements, but also able to align content size with the unique characteristics and affordances of mobile devices.

A novel contextual topic model for multi-document summarization

A novel contextual topic model is proposed for multi-document summarization.The main idea is to leverage hierarchical topics and their correlations with respect to the lexical co-occurrences of words.The proposed contextual topic model can effectively determine the relevance of sentences. Information overload becomes a serious problem in the digital age. It negatively impacts understanding of useful information. How to alleviate this problem is the main concern of research on natural language processing, especially multi-document summarization. With the aim of seeking a new method to help justify the importance of similar sentences in multi-document summarizations, this study proposes a novel approach based on recent hierarchical Bayesian topic models. The proposed model incorporates the concepts of n-grams into hierarchically latent topics to capture the word dependencies that appear in the local context of a word. The quantitative and qualitative evaluation results show that this model has outperformed both hLDA and LDA in document modeling. In addition, the experimental results in practice demonstrate that our summarization system implementing this model can significantly improve the performance and make it comparable to the state-of-the-art summarization systems.

An Infrastructure for Developing Pervasive Learning Environments

This paper presents an infrastructure for developing problem-based pervasive learning environments. Building such environments necessitates having many autonomous components dealing with various tasks and heterogeneous distributed resources. Our proposed infrastructure is based on a multi-agent system architecture to integrate various components of the environments. The infrastructure includes a location- and context-awareness service, a question and answer service, an adaptive mechanism; problem based ubiquitous learning models, social networking issues, and the evaluation of multimedia inputs. Furthermore, student modeling issues among components are considered. The design of the infrastructure as well as its components is discussed. This paves the way towards the development of pervasive learning applications.

A novel topic feature for image scene classification

Abstract We propose a novel topic feature for image scene classification. The feature is defined based on the thematic representation of images constructed by using topics, i.e., the latent variables of LDA (latent Dirichlet allocation) and their learning algorithms. Different from the related works, the feature defined in this paper shares topics in different classes, and does not need class labels before classification, so that it can avoid the coupling between features and labels. For representing a new image, our approach directly extracts its topic feature by codewords linear mapping instead of the inference of latent variable. We compared our method with three other topic models under similar experimental condition, as well as with pooling methods on the 15 Scenes dataset. The results show that our approach is capable of classifying the scene classes with a higher accuracy than the other topic models and pooling methods without using spatial information. We also observe that the performance improvement is due to the proposed feature and our algorithm, rather than the other factors such as additional low-level image features and stronger preprocessing.

Dictionary learning for VQ feature extraction in ECG beats classification

We improve dictionary learning algorithm for vector quantization of ECG.The algorithm is employed to extract feature of ECG.The algorithm can avoid interference from dirty data.The algorithm is capable of increasing classification accuracy.An initial cluster centers selecting method is utilized to speed up the algorithm. Vector quantization(VQ) can perform efficient feature extraction from electrocardiogram (ECG) with the advantages of dimensionality reduction and accuracy increase. However, the existing dictionary learning algorithms for vector quantization are sensitive to dirty data, which compromises the classification accuracy. To tackle the problem, we propose a novel dictionary learning algorithm that employs k-medoids cluster optimized by k-means++ and builds dictionaries by searching and using representative samples, which can avoid the interference of dirty data, and thus boost the classification performance of ECG systems based on vector quantization features. We apply our algorithm to vector quantization feature extraction for ECG beats classification, and compare it with popular features such as sampling point feature, fast Fourier transform feature, discrete wavelet transform feature, and with our previous beats vector quantization feature. The results show that the proposed method yields the highest accuracy and is capable of reducing the computational complexity of ECG beats classification system. The proposed dictionary learning algorithm provides more efficient encoding for ECG beats, and can improve ECG classification systems based on encoded feature.

Personalized Text Content Summarizer for Mobile Learning: An Automatic Text Summarization System with Relevance Based Language Model

Although millions of text contents and multimedia published on the Web have potential to be shared as the learning contents for mobile learning, effectively extracting useful information from them is an extremely difficult problem. Oft-decried information overloading is the main issue to impede this potential. Many approaches have been proposed to revise and reinforce content to provide the appropriate delivery for mobile learning. However, approaches of manually converting content to suit the mobile learning require a huge effort on the part of the teachers and the instructional designers. Automatic text summarization can reduce this cost significantly, but it may have negative impact on the understanding of the meaning conveyed, as well as the risk of producing a standard summary for all learners without reflecting their interests and preferences. In this paper, a personalized text-based content summarizer is introduced to address an approach to help mobile learners to retrieve and process information more quickly, based on their interests and preferences. In this work, probabilistic language modeling techniques are adapted to build a user model and an extractive text summarization system to generate the personalized and automatic summary for mobile learning. Experimental results have indicated that the proposed solution provides a proper and efficient approach to help mobile learners by summarizing important content quickly and adaptively.

Transition from e-Learning to u-Learning: Innovations and Personalization Issues

Use of mobile and sensor technologies in learning has emerged as a growing research area, and has given rise to a lot of research that takes advantage of learners location, environment, proximity and situation to contextualizing the learning process. The adaptivity and personalization in these scenarios have taken a new meaning by bringing authentic learning much closer to holistic learning by seamlessly integrating physical objects available in the learners vicinity with virtual information in real-time. Such ubiquitous environments not only break the barriers for education by widening the access to those who cannot come to a physical classroom but also increase the richness of the instruction by integrating multiple sources of instruction, contextualization and real-time location-aware learning, hence overcoming the limitations of classroom learning. This paper focuses on various contexts that arise in such environments where seamless immersion of formal and informal activities and interactions has potential to contribute to the learning process.

A new Web-service-based architecture for question answering

A natural language dialogue system is composed of many parts: natural language process, dialogue management, data query, etc. Adding a new service function to traditional complex dialogue systems, which is based on a mixed initiative paradigm, for example question answering (QA) service for e-commerce Web site, requires in general a big effort for redesigning dialogue management modules and other important modules. The new architecture suggested in this paper is based on registration of Web services, and all the services are distributed on agents. It can balance between overall dialogue management and particular agent service. The new services to be added in the system need only register as agents with a central managing server. That makes application of complex QA system much easier for Web site service application compared to traditional architecture. We discuss the advantages brought by the new architecture, including more extensibility, more effective, platform insensitivity, wider adaptability etc., and describe in detail how to implement the system modules.

Adaptive Assessment in Web-Based Learning

Web-based assessment is used in different contexts with the aim to support students and help to make learning easier and more effective for them. Typically, the individual characteristics and needs of students are used to personalize and customize existing approaches to assessment. In this paper, we show the potential of adaptive web-based assessment in different learning applications. We introduce adaptive systems in the area of readiness self-assessment, performance self-assessment, and peer assessment. We show how they incorporate individual differences and create adaptive applications.

Automatic Twitter Topic Summarization

This paper aims to generate digests of tweets from live trending and ongoing topics. The primary purpose is to group the tweets by importance or usefulness so that an end user can be presented with a reasonable extract of the most important content from the Twitter stream. Summarization is accomplished using a non-parametric Bayesian model applied to Hidden Markov Models and a novel observation model designed to allow ranking based on selected predictive characteristics of individual tweets.