Research

Abstract

In Taiwan, the way of university admissions has changed since 2019. The e-portfolio system was built to replace traditional paper documents. Students’ school grades, e-portfolio, and award records are uploaded layer by layer and are stored in the e-portfolio central database. This leaves the data of the students vulnerable to loss should the central database be attacked. With the change in university admissions, the focus of the review has also changed from entrance exam scores to long-term student learning performance. Students also actively participate in various activities to explore their interests. However, due to the lack of trust bridges between organizations, students cannot freely share and use their data. Students lack a self-sovereign identity that can sign up and control the authorization of their data.  We propose a self-sovereign identity based personal information security control infrastructure for the e-portfolio ecosystem. The decentralized identity chain and e-portfolio application chain are included in this system. The decentralized identity chain integrates the identities of users in different ecosystems and gives users a self-sovereign identity that can be fully controlled by themselves. The e-portfolio application chain records the authorization of the user. Besides, the trusted education unit audits the source of the review data together.

Abstract

This study uses deep learning and focuses on the improvement of medical processes. During X-ray examination, an RGB image of the patient’s position is pre-taken before X-ray exposure and then a convolutional neural network (CNN) model is used. After performing an image classification task on this RGB image, we’ll compare the classification result with the doctor’s order to see if it is consistent.

Abstract

Financial institutions are following an open banking (OB) trend for service innovation and integration. OB allows third-party service providers (TSPs) to access user financial data for purposes of finding the best deals and improving user experiences. The trust in third parties required for OB ecosystem success raises questions about digital identity integration, data sharing, and privacy preservation. Decentralized applications (DApps) for classifying and protecting data privacy, recording user consent, tracking TSP access actions, managing application programming interfaces (APIs), and preserving self-sovereign identities already exist, but in some countries such as Taiwan they have yet to be integrated into an operable three-phase OB approach. After identifying all major requirements of primary OB participants, we develop a blockchain-based identity management and access control (BIMAC) framework that shares some advantages of both traditional banking and blockchain technology. The BIMAC infrastructure applies smart contracts and a stateless authentication mechanism to form a reliable personal information transaction security control (PITSC) platform that offers such functionalities as decentralized third-party login (TPL), the ability to open bank accounts online, data authorization/revocation, integrated payouts, and TSP access monitoring. System performance evaluation results indicate that the frequent execution functions of the proposed framework have lower computation costs than the average transaction cost of public Ethereum.

Related Paper : Blockchain-based identity management and access control framework for open banking ecosystem

Abstract

Full-reference image quality assessment (FR-IQA) compares the differences between a test image and the reference image, and assigns an image quality score to the test image. Deep Image Structure and Texture Similarity (DISTS) compares texture and structure similarity of feature maps to evaluate image quality. Image Quality Transformer (IQT) applies Transformer to estimate image quality, which performs well on a novel image quality assessment (IQA) dataset, Perceptual Image Processing ALgorithms (PIPAL) dataset. This paper tries to integration them and improves on three key points: feature extraction backbones, integration of IQT and DISTS, and augmentation of IQA dataset through generative adversarial network (GAN). We propose Auxiliary Transformer with DISTS for Image Quality Assessment (ATDIQA) and Augmented Auxiliary Transformer with DISTS for Image Quality Assessment (Augmented ATDIQA) by adjusting weights of backbones in DISTS-based methods, improving IQT-based methods by extracting features of different scales, retraining the models using the augmented dataset, and finally weighted averaging the results of these methods. The results show that this method can improve the effect not only on PIPAL, but also on other IQA datasets. After further researching, we conclude that extracting different scale features and integrating these IQT-based methods can improve the performance on PIPAL.

Related Paper : Full-Reference Image Quality Assessment with Transformer and DISTS

Abstract

This study proved that students can develop computational thinking (CT) literacy through non programmed training methods. This study used Rummikub to conduct experiments, designed competitions and questionnaires, and measured students’ CT literacy by pre and post-test from Bebras’ Challenge. The result showed that for different level students, low-level students had significant improvement. This research showed that non-programmed training methods can effectively improve CT literacy for low-level students.

Abstract

Eye movement technology is highly valued for evaluating and improving digital learning content. In this paper, an educational innovation study of eye movement behaviors on digital learning content is presented. We proposed three new eye movement metrics to explain eye movement behaviors. In the proposed method, the digital content, which were slide-deck-like works, were classified into page categories according to the characteristics of each page. We interpreted the subjects’ eye movement behaviors on the digital slide decks. After data regularization and filtering, the results were analyzed to give directions for how to design an attractive digital learning content from the viewpoint of eye movement behaviors. The relationships between the subjects’ evaluation scores, page categories, and eye movement metrics are discussed. The results demonstrated that the proposed fixation time percentage (FTP) was a representative, strong, and stable eye movement metric to measure the subjects’ interest. Moreover, a reasonable portion of semantic content had a positive influence on the subjects’ interest.

Related Paper : Eye movement analysis of digital learning content for educational innovation

Abstract

More and more brands or companies have implemented loyalty programs in recent years. However, there are too many types of reward points to circulate, and most may have expired before they can be redeemed. In addition, small merchants cannot develop robust loyalty programs on their own like large corporations, and if they join a consortium loyalty program, they will be limited to a certain extent. To address these issues, we propose a blockchain-based collaborative loyalty program for a business consortium (BCLP). The system allows enterprises to integrate into the business alliance independently, without the need for the central party to dominate, and customers can exchange one point for various items at any time, without worrying about the point expiration. Ethereum smart contracts can record our core information: reward points (RP ) in a decentralized way, and provide anti-tampering, auditing events, and other functions to avoid attacks. The main contributions of this paper are as follows: Bank Point Liabilities: Banks recognize funds deposited by issuers as accounts payable, allowing funds to be used more freely; ERC-20 token standard: RP in our system are derived from the ERC-20 token standard; P2P transaction: Thanks to blockchain technology, every role can conduct peer-to-peer transactions securely in real time; User Point Exchange: Users can exchange reward points on our platform for other points issued by other loyalty program issuers and vice versa; Regulatory Authority: Regulatory authorities are required to monitor abnormal behavior in the system.

Related Paper : Blockchain-based cross-organizational integrated platform for issuing and redeeming reward points

Abstract

The prevalence of predatory journals has become more severe recently as this is harmful to science and technology development. For scholars publish papers more effectively and avoid publishers for profits, this research used a machine learning method to identify the predatory journals. The features like text content and keywords of the collected journals’ websites were extracted from mainstream predatory journal websites and normal journal websites. This research proposed a predatory journal classification system based on a new model. The results show that our model’s recall rate exceeds 90%, ensuring that the journals submitted by the researchers are not predatory.

Related Paper : An open automation system for predatory journal detection