Traditional Automated Speaking Assessment (ASA) systems exhibit inherent modality limitations: text-based approaches lack acoustic information while audio-based methods miss semantic context. Multimodal Large Language Models (MLLM) offer unprecedented opportunities for comprehensive ASA by simultaneously processing audio and text within unified frameworks. This paper presents a very first systematic study of MLLM for comprehensive ASA, demonstrating the superior performance of MLLM across the aspects of content and language use. However, assessment on the delivery aspect reveals unique challenges, which is deemed to require specialized training strategies. We thus propose Speech-First Multimodal Training (SFMT), leveraging a curriculum learning principle to establish more robust modeling foundations of speech before cross-modal synergetic fusion. A series of experiments on a benchmark dataset show MLLM-based systems can elevate the holistic assessment performance from a PCC value of 0.783 to 0.846. In particular, SFMT excels in the evaluation of the delivery aspect, achieving an absolute accuracy improvement of 4% over conventional training approaches, which also paves a new avenue for ASA.

A recent line of research on automated speaking assessment (ASA) has benefited from self-supervised learning (SSL) representations, which capture rich acoustic and linguistic patterns in non-native speech without underlying assumptions of feature curation. However, speech-based SSL models capture acoustic-related traits but overlook linguistic content, while text-based SSL models rely on ASR output and fail to encode prosodic nuances. Moreover, most prior arts treat proficiency levels as nominal classes, ignoring their ordinal structure and non-uniform intervals between proficiency labels. To address these limitations, we propose an effective ASA approach combining SSL with handcrafted indicator features via a novel modeling paradigm. We further introduce a multi-margin ordinal loss that jointly models both the score ordinality and non-uniform intervals of proficiency labels. Extensive experiments on the TEEMI corpus show that our method consistently outperforms strong baselines and generalizes well to unseen prompts.

Abstract Text readability assessment is a challenging interdisciplinary endeavor with rich practical implications. It has long drawn the attention of researchers internationally, and the readability models since developed have been widely applied to various fields. Previous readability models have only made use of linguistic features employed for general text analysis and have not been sufficiently accurate when used to gauge domain-specific texts. In view of this, this study proposes a latent-semantic-analysis (LSA)-constructed hierarchical conceptual space that can be used to train a readability model to accurately assess domain-specific texts. Compared with a baseline reference using a traditional model, the new model improves by 13.88% to achieve 68.98% of accuracy when leveling social science texts, and by 24.61% to achieve 73.96% of accuracy when assessing natural science texts. We then combine the readability features developed for the current study with general linguistic features, and the accuracy of leveling social science texts improves by an even higher degree of 31.58% to achieve 86.68%, and that of natural science texts by 26.56% to achieve 75.91%. These results indicate that the readability features developed in this study can be used both to train a readability model for leveling domain-specific texts and also in combination with the more common linguistic features to enhance the efficacy of the model. Future research can expand the generalizability of the model by assessing texts from different fields and grade levels using the proposed method, thus enhancing the practical applications of this new method.

It has been shown in numerous studies that the application of Augmented Reality (AR) to teaching and learning is beneficial, but determining the reasons behind its effectiveness, and in particular the characteristics of students for whom an AR is best suited, can bring forth new opportunities to integrate adaptive instruction and AR in the future. Through a quasi-experimental research design, our study recruited 66 participants in an 8-week long AR-assisted learning activity, and lag sequential analysis was used to analyze participants' behavior in an AR learning environment. We found that AR was more effective in enhancing the learning gains in elementary school science of learners who prefer a Kinesthetic approach to learning. We hypothesize that these effects are due to the increase in opportunity for hands-on activities, effectively increasing learners' concentration and passion for learning.

The study aims at investigating how the fitness between learners' learning styles and adopted learning modalities influences the learning of English as a foreign language (EFL) by elementary school students. A quantitative research was adopted. 87 fifth graders participated in this study. They were randomly assigned to three groups of learning modalities: video, gesture-based, and 3D virtual. They learned 18 English phrases related to motions for 3 times, once a week. Both performance test and learning style questionnaire were administered after the treatment completed. The analysis results of the current study show that the EFL learners' performances did not differ among the three learning modalities. However, EFL learners' learning styles also influence their performance. Implications of this study for language education and teaching are discussed.

Previous studies on Mandarin tone production indicate that there is no agreement on which tones are most difficult for L2 learners. Much of previous research on L2 learning of Mandarin tones has focused on monosyllables. In modern Mandarin, however, it is disyllabic words that dominate the vocabulary. This research investigates the production of Mandarin disyllabic tones by Japanese learners. In the current study, 25 Japanese learners of Mandarin were requested to produce 80 Mandarin disyllabic words with all tonal combinations (except for the neutral tone). The overall results showed a hierarchy of difficulty: Tone 3 > Tone 2 > Tone 1 = Tone 4. Most errors in the first syllable were found for Tone 2 and Tone 3 when followed by Tone 1 or Tone 4 (both start with a high pitch). In the second syllable, most errors were found for Tone 3 (misproduced as Tone 2). The findings are discussed in terms of the phonetic nature of Mandarin lexical tones and the interference from Japanese phonology.

This study presents a conceptual framework of applying mobile augmented reality technology on peer assessment to reduce the bounds of work reviews and assessment. According to mobile technology and augmented reality, students can show their design in various ways at di fferent places or in different situations. This paper proposes a novel mobile peer-assessment system which combines augmented reality with the reviewing and assessing processes. This framework enable s students to enhance work interpretation, frequently interact with peers, represent their thinking and reflect upon their own works. Furthermore, the mobile AR technique provides personalized and location-based adaptive contents that enable individual students to interact with the mixed reality environment and to observe how works are possibly applied to the real world in the future. The whole process assists students in reviewing works based on various dimensions, gaining proper knowledge, cultivating critical thinking skills and reflection as well as promoting meaningful learning.

The studies of readability aim to measure the level of text difficulty. Although traditional formulae such as the Flesch-Kincaid formula can properly predict text readability, they are only effective for English text. Other formulae with very few features may result in inaccurate text classification. The study takes into account multiple linguistic features, and attempts to increase the level of accuracy in text classification by adopting a new model which integrates Principal Component Analysis (PCA) with Genetic Programming (GP). Empirical data are utilized to demonstrate the performance of the proposed model.

The aim of this study is to develop a diagnostic system for mathematical concepts. By adopting a Bayesian network for its high recognition rate in artificial intelligence and diagnosis, and combining and applying deduction methods in computerized tests, this system helps students to understand the difficulties they encounter in mathematical learning, and subsequently helps in implementing immediate remedies. The computerized diagnostic tests established in this research module can diagnose the types of mistakes students make; and in addition to helping students realize their erroneous concepts, this system also helps teachers to grasp the types of mistakes students make, and to implement group remedial teaching accordingly. The study result indicates that the mean recognition rates of the computerized diagnostic system developed in this study are 95.72 %, 99.10 %, 98.73 %, 99.02 %, and 98.96 %; this system can effectively and automatically detect the types of mistakes that students make.

The grasp of main ideas as a learning strategy and activity can have various benefits on learners in different instructional objectives, such as reading comprehension, memory, and discerning the main points of reading content. A necessary condition for grasping main ideas is determining the importance of sentences [2]. Thus, this study attempts to explore the interaction of a reciprocal teaching digital annotation automatic matching system with teachers and learners, and how it affects learning activities. This study hopes to assist teachers in using computer systems to quickly assess how student groups are marking important points, and uses the computer to automatically evaluate the similarity between class markings and teacher markings, so that the teacher can have a general idea of how students have different markings from the teacher. This is used to understand students’ learning conditions and the distribution of their markings for later instructional discussion of important points, so that learners can receive better learning guidance in a limited period of time. The concept of the system in this study can allow the teachers and teaching assistants to provide timely assistance to learners in the teaching and learning process, and provide suitable support, with the objective of “internalizing” the development of learner ability development. The whole system concept proceeds in a digital learning environment, and serve as a learning motivation for learners and a direction for instructors to improve their teaching, so that they can achieve the purpose of offering individualized teaching.

This paper describes the field experience of two prospective teachers of Chinese as a Second Language (CSL) in the teacher-training programme, regarding the differences in pedagogic skills between the face-to-face and the virtual world of Second Life. The student teachers recounted how they made pedagogical changes in relation to the extent of the consciousness raising, decision making and perception towards the different environments. A qualitative approach was primarily adopted to elicit this information from the data that consisted of post-teaching interviews and reflection reports. The preliminary results confirm that, to some extent, the prospective teachers’ pedagogical knowledge was influenced by their previous experience (e.g., foreign language learning). The findings derived from a self-rating scale further reveal that the student teachers felt confident about their pedagogical knowledge and skills in the face-to-face context, e.g., conducting activities, but not confident in the virtual classroom, e.g., managing students’ conduct. Both the student teachers opined that teaching in the virtual world was feasible to engender the involvement and interest of learners. The teachers’ knowledge transferred from the conventional classroom to the virtual world has implications for teacher education.

The aim of this research is to develop a hands-on spatial geometry learning system to facilitate students’ geometry learning. The system is developed with Duval’s four critical elements of geometry apprehension including perceptual apprehension, sequential apprehension, operational apprehension and discursive apprehension. The system supports senior high school students in the process of spatial geometry problem-solving, allowing them to hands-on manipulate spatial geometry graphics and develop their visualization and mental imagery. In total, 58 participants from different classes were recruited. The experimental group used the hands-on learning system, whereas the control group followed the traditional paper-based approach. The study investigates the effects of the hands-on geometry learning system on students’ perceptual apprehension, sequential apprehension, operational apprehension, overall spatial geometry scores and learning attitude. The results revealed more learning attitude, and higher perceptual apprehension, sequential apprehension, operational apprehension in the experimental group.

This paper provides an overview of implementing self-directed learning by utilizing weblogs in elementary studentspsila independent study. The key feature of studentspsila independent study in Taiwanese elementary settings was used to teacher-directed learning. By utilizing weblogs in studentspsila science study, it is found that students seem to behave more self-directed learning behaviors and more active in pursuing their study goals because of the feedbacks and participation by the peer community.

This study aims to explore the time-management patterns of learners' interactive behaviors during online project-based learning (PBL) from an empirical observation without teachers' interception. According to the derived pattern, the study found how learners allocated their time and how these interactive activities influence learners to allocate their time more appropriately. Finally, we also gave some suggestions for teachers' interception and guidance for promoting the quality of students' projects.

In this paper, we designed a wireless peer-assisted leaning system for early EFL reading to enhance online collaborative EFL reading and provide distributed EFL learners with an opportunity to teach and learn with each other. Compare with the control group, the result shows that wireless EFL reading is emerging as a flexible and portable solution to the pedagogical challenges exist in conventional EFL reading teaching instruction.

Concept maps have been widely put to educational uses. Concept maps in the past were limited in their assessment standards. Some only take concept nodes into consideration; other evaluation criteria use concept nodes as the basis, with relation links playing a minor role. To address this problem, this study proposes a new style of concept map, called weighted concept map, which assigns a weight to each proposition in a concept map to represent its importance. This study proposes a new assessment with qualitative comparison founded on a weighted concept map.