김연성 교수

Existing metric learning losses can be categorized into two classes: pair-based and proxy-based losses. The former class can leverage fine-grained semantic relations between data points, but slows convergence in general due to its high training complexity. In contrast, the latter class enables fast and reliable convergence, but cannot consider the rich data-to-data relations. This paper presents a new proxy-based loss that takes advantages of both pair- and proxy-based methods and overcomes thei

Abstract: This study examined whether foreign language learner anxiety and motivational goal orientations remained stable across two different classroom contexts: a reading course and a conversation course. The researcher measured anxiety and four types of motivational goal orientations by surveying 59 Korean college students learning English in both courses. A repeated‐measures MANCOVA was used to analyze the responses. The findings indicated that levels of anxiety can vary according to instruc

Metric Learning for visual similarity has mostly adopted binary supervision indicating whether a pair of images are of the same class or not. Such a binary indicator covers only a limited subset of image relations, and is not sufficient to represent semantic similarity between images described by continuous and/or structured labels such as object poses, image captions, and scene graphs. Motivated by this, we present a novel method for deep metric learning using continuous labels. First, we propo

This study investigated classroom teacher anxiety with respect to teaching foreign languages in Korea, and sought to identify sources and symptoms of this anxiety. The subjects' anxiety was measured using the Foreign Language Teaching Anxiety Scale (FLTAS), a scale that was developed for this study. A number of participants acknowledged that they had experienced teaching anxiety in foreign-language classrooms. These results provide evidence for the construct of foreign language teaching anxiety,

This paper presents a novel method for embedding transfer, a task of transferring knowledge of a learned embedding model to another. Our method exploits pairwise similarities between samples in the source embedding space as the knowledge, and transfers them through a loss used for learning target embedding models. To this end, we design a new loss called relaxed contrastive loss, which employs the pairwise similarities as relaxed labels for intersample relations. Our loss provides a rich supervi

It has been more than 5 years since the TETE policy came into effect. Now is the time to assess the effectiveness of the policy, particularly from the teachers’ point of view. This study aims to identify Korean teachers’ responses to the TETE policy through written survey questionnaires. Their perspectives were analyzed with reference to the type of school they work for, the amount of teaching experience they have, and the frequency with which they use English. The findings indicate that most of

Abstract Teachers who know what, how and why to teach are essential for successful student learning. However, many preservice teachers (PSTs) lack teaching experience and the ability to integrate theory and practice. To help bridge this gap, this study employed a learning‐by‐design project approach in which 22 Korean PSTs developed lesson plans for middle school English classes, constructed virtual classrooms in the metaverse based on their English lesson plans, and conducted microteaching in th

We present a novel self-taught framework for unsuper-vised metric learning, which alternates between predicting class-equivalence relations between data through a moving average of an embedding model and learning the model with the predicted relations as pseudo labels. At the heart of our framework lies an algorithm that investigates contexts of data on the embedding space to predict their class-equivalence relations as pseudo labels. The algorithm enables efficient end-to-end training since it

Supervision for metric learning has long been given in the form of equivalence between human-labeled classes. Although this type of supervision has been a basis of metric learning for decades, we argue that it hinders further advances in the field. In this regard, we propose a new regularization method, dubbed HIER, to discover the latent semantic hierarchy of training data, and to deploy the hierarchy to provide richer and more fine-grained supervision than inter-class separability induced by c

Purpose This study aims to investigate how information quality and system quality influence the effectiveness of artificial intelligence (AI)-based recommendation service platforms. It integrates traditional information technology service quality (SQ) metrics with recommendation SQ measures, focusing on their impact on user satisfaction and behavior. This study further examines the moderating effects of standardization and customization on these relationships. Design/methodology/approach This st

This study examines the identities of Korean English teachers who have lived in various English-speaking countries but now serve as certified public school teachers in Korea. Data were collected from in-depth interviews with seven teachers and analysed via critical discourse analysis. Despite being rated as near-native, the teachers with transnational experience still undervalued their English proficiency, possibly because they compared their proficiency to that of native speakers. They were als

The goal of this study was to explore foreign language learner anxiety and strategy use associated with English listening and reading. For the purpose of the study, 98 high school freshmen took the test designed to measure their listening and reading comprehension. The participants were also asked to respond to the questionnaire items constructed to measure their anxiety and strategy use in English listening and reading. For data analysis, correlation and multiple regression analyses were perfor

Metric Learning for visual similarity has mostly adopted binary supervision indicating whether a pair of images are of the same class or not. Such a binary indicator covers only a limited subset of image relations, and is not sufficient to represent semantic similarity between images described by continuous and/or structured labels such as object poses, image captions, and scene graphs. Motivated by this, we present a novel method for deep metric learning using continuous labels. First, we propo