End-to-End Automatic Speech Recognition (ASR) models have significantly advanced the field of speech processing by streamlining traditionally complex ASR system pipelines, promising enhanced accuracy and efficiency. Despite these advancements, there is a notable absence of freely available medical conversation speech corpora for Burmese, which is one of the low-resource languages. Addressing this gap, we present a manually curated Burmese Medical Speech Conversations (myMediCon) corpus, encapsulating conversations among medical doctors, nurses, and patients. Utilizing the ESPnet speech processing toolkit, we explore End-to-End ASR models for the Burmese language, focus on Transformer and Recurrent Neural Network (RNN) architectures. Our corpus comprises 12 speakers, including three males and nine females, with a total speech duration of nearly 11 hours within the medical domain. To assess the ASR performance, we applied word and syllable segmentation to the text corpus. ASR models were evaluated using Character Error Rate (CER), Word Error Rate (WER), and Translation Error Rate (TER). The experimental results indicate that the RNN-based Burmese speech recognition with syllable-level segmentation achieved the best performance, yielding a CER of 9.7%. Moreover, the RNN approach significantly outperformed the Transformer model.

ASR (Automatic Speech Recognition) is favorably chosen as a learning technology, which is used for English pronunciation practice. This research aims to build a personalized learning platform to improve English pronunciation at the word level for Thai EFL learners who learn English as a Foreign Language (EFL) by using ASR to detect mispronounced sounds. ASR models are built with an End-to-End learning approach with a Thai-English mispronounced words dataset. The practice of English pronunciation particularly focuses on eleven problematic consonant sounds of Thai EFL students according to the previous studies of English pronunciation in Thai contexts. These eleven consonant sounds are divided into five groups: 1) /ð/-/θ/-/tθ/, 2) /ʒ/-/ʃ/, 3) /dʒ/-/tʃ/, 4) /z/-/s/ and 5) /b/-/p/. The five of Grade 12 Thai Students who are native Thai speakers were selected as sampling process. The result of pre-test and post-test show that the samples have the most problem with the consonant sounds of /ö/-19/-1t9/ (29%), followed by /b/-/p/ (22%), /d3/-/tf/ (22%), /z/-/s/ (18%) and 13/-11/ (9%) respectively. In conclusion, this study reveals that 60% of the samples have improved their pronunciation after using our system.

In the media industry, there is a high demand for animation films, leading to the establishment of animation film courses in various universities to train future animators. However, writing animation scripts poses challenges for students, as it requires critical thinking skills to craft captivating and coherent story ideas, akin to other creative works that adhere to general principles for a complete and logical narrative. To solve this issue, we developed a tool as part of our project to help students systematically organize the essential components of a story based on the fundamental principles of animation, consisting of three acts. This tool creates a conducive learning environment by breaking down crucial elements into clear sections, prompting students to reflect on their own ideas about animation stories. Additionally, the tool encourages collaboration between students and instructors, enabling constructive feedback and reflections through error corrections. The results of our experiments showed significant improvements in students' development after using the tool. Notably, the recurring errors in omitting vital parts of the script did not reoccur after the initial use of the tool. Its implementation heightened students' awareness of the importance of each component. Furthermore, the assessment scores of all students demonstrated a significant improvement, with 34% of students displaying increased awareness in scriptwriting. The elements that students commonly missed were 'conflict' and 'progression of complications,' respectively.

This paper proposes a method to develop a machine translation (MT) system from Myanmar Sign Language (MSL) to Myanmar Written Language (MWL) and vice versa for the deaf community. Translation of MSL is a difficult task since only a small amount of a parallel corpus between MSL and MWL is available. To address the challenge for MT of the low-resource language, transfer learning is applied. An MT model is trained first for a high-resource language pair, American Sign Language (ASL) and English, then it is used as an initial model to train an MT model between MSL and MWL. The mT5 model is used as a base MT model in this transfer learning. Additionally, a self-training technique is applied to generate synthetic translation pairs of MSL and MWL from a large monolingual MWL corpus. Furthermore, since the segmentation of a sentence is required as preprocessing of MT for the Myanmar language, several segmentation schemes are empirically compared. Results of experiments show that both transfer learning and self-training can enhance the performance of the translation between MSL and MWL compared with a baseline model fine-tuned from a small MSL-MWL parallel corpus only.

Metacognition is difficult to train, and most of students lack the knowhow to practice it. This work proposes a thought analysis tool from reading to help training in understand logical connections between sentences. The tool allows learners to explicate types of logical expression of the writing article as the example of strategy to convince readers. By analysis and annotating the text in a controlled environment, it is expected that learners can learn from thinking about author’s cognitive process, and apply them to think about their own thought and make a strategic planning when they are in a role to write. The tool is designed for learners to assign several kinds of annotations including logical statement type, keyword and logical linking between sentence to the proper writing text by self-analysis under the supervision of coaches. The experiment results show that the tool helps to apparently improve learners’ cognitive performance in composing proper essay via the training on analysis reading. In comparison to other sample groups with lecturing and coaching without the tool, the improvement of the tool users was significantly greater in term of logical relatedness, logical completeness and convincible power.

Open government data (OGD) is a global initiative to promote transparency, service innovation and citizen participation. The most common means for publishing OGD is usually in forms of datasets made available on OGD catalogs. Although publishing open data as datasets is straightforward and requires minimal technological skills, it is not ideal for the users who want to use the data in a more dynamic fashion. This paper proposes a scalable framework for creating OGD services from OGD catalog. Our framework emphasizes the need to convert existing OGD datasets to RDF data and value-added services that provides more convenient access to the users. Data querying APIs are among the OGD services to promote application development from the OGD datasets. Our framework is unique in that it does not require additional user intervention in the dataset publishing process and hides the complexity of the linked data technology from the data publishers and users. In this framework, the datasets listed in the OGD catalog were collected and validated for its well-formedness of the tabular data. The service building system converted the data to the RDF format and utilized SPARQL query templates in building the OGD services for each dataset. The framework was applied with the datasets on Data.go.th. The results on overall successful conversion rate of the datasets into the OGD services are reported. The case study exemplifies a scalable approach to augmenting access to OGD and provide a first step in moving OGD towards linked data.

In this paper, we propose a method to detect a similarity of competency to generate a career path. Career path is important for students and workers for their planning in career. In this work, data of competencies from Thailand Professional Qualification Institute are used for generating a relation among units of competency (UoC) as a crossable path for career transfer. Similarity Score using ngram precision is exploited to find the commonness in UoC context to indicate the possibility in career relation. From an experiment, the proposed method gained 80% accuracy. As a result, career path is generated as a map for a person in career planning for both promotional path and crossable path.

In education sector, teaching style has been adapted to the online content platform. Moreover, MOOCs (Massive Online Open Courses) and flipped learning become high potential tools to support student learning process. These e-learning tools have been designed for education leading countries based on individual students’ learning style. It is quite difficult to apply for other countries. In Thailand, there are insufficient number of teachers in the rural schools and teachers have to teach a lot of subjects both experienced and inexperienced subjects. This paper proposes a new design of MOOCs hybrid learning model which is suitable and effective for rural areas students and analyse the important features to identify the factor which have influence on student ability.

Although the Semantic Web data standards are established, ontology-based applications built on the standards are relatively limited. This is partly due to high learning curve and efforts demanded in building ontology-based Semantic Web applications. In this paper, we describe an ontology application management (OAM) framework that aims to simplify creation and adoption of ontology-based application that is based on the Semantic Web technology. OAM introduces an intermediate layer between user application and programming and development environment in order to support ontology-based data publishing and access, abstraction and interoperability. The framework focuses on providing reusable and configurable data and application templates, which allow the users to create the applications without programming skill required. Three forms of templates are introduced: database to ontology mapping configuration, recommendation rule and application templates. We describe two case studies that adopted the framework: activity recognition in smart home domain and thalassemia clinical support system, and how the framework was used in simplifying development in both projects. In addition, we provide some performance evaluation results to show that, by limiting expressiveness of the rule language, a specialized form of recommendation processor can be developed for more efficient performance. Some advantages and limitations of the application framework in ontology-based applications are also discussed.

This article reveals the design of an ongoing research that investigates the effectiveness of an alternative learning environment Circuitously Collaborative Learning Environment (CirCLE), which is designed to enhance metacognitive awareness on the learning processes in algebraic mathematical word problem (MWP) solving environments. We perform the research based on the hypothesis that a student will be encouraged and can reflect his own thinking when he practicing a role of an inspector together with receiving appropriate feedback to revise his solutions.

In this paper, a system to classify a readability level of English reading passage and to match student personal interest is purposed. Student model is applied to collect student information for selecting their preferable passage topic. Statistical passage level checker is implemented to match student readability level with passage difficulty by using neural network. Three linguistic features, syllable, vocabulary and sentence complexity, are chosen to distinguish a difficulty difference among passage level. The best accuracy gained by the system is 86.25% and the constantly reliable feature for this task is a sentence complexity of the passage.

This paper describes the design and development of EAGLE, an Error tAGger for Learners of English. EAGLE combines all the processes necessary for the analysis of learners' language, such as creating an error tagset, tagging learners' writing, and reporting error tagging results, within the same tool. EAGLE has been developed to allow more flexibility in the creation of tagsets as well as to support many tagsets. These functionalities are achieved with the use of a hierarchical tree tagset and offset annotation. Researchers can develop their own tagsets from different theoretical frameworks and even apply them to the same document. EAGLE also provides multiple ways of viewing and comparing the statistics of tagged errors. All these features allow taggers to compare their ideas and work together to create an error tagset to render the analysis more reliable and accurate. Since EAGLE was not designed specifically for any learner language, it could be applied to tag errors produced by learners of other second languages as well.

In this paper, we develop a preliminary research on passage grading system. We propose an approach to examine an English reading passage that meets students' ability and level. CRF has been applied to create a level characteristic model from passage corpus. The system calculates by using three features; word, syllable and sentence complexity. The system does not require manual criteria to grade a passage, but passages are automatically graded by comparing their scores to a model. The output of the system shows a level of passage based on Thai academic school level.

In this paper, we develop a framework to increase student's motivation by applying automatic translations which provide Thai translated output to students. Students have to edit or correct the translation results from the system and receive comments from teacher. Based on this framework, the system enables them to increase their motiction in reading. We evaluate results of our system at Thammasart Klongluang School. The students get a significance improvement and satisfy with the system.

Patient self-management is an important component in improving quality of chronic disease healthcare. The promising benefits of the Interactive Behavior Change Technology (IBCT) on diabetes patient self-management are increasingly recognized. In this paper, we describe development of a knowledge portal prototype designed for enhancing self-management support among patients with type-2 diabetes. The portal focuses on personalization of the provided services: self-regulation, self-monitoring and evaluation, social support, virtual home visit and reminder. One of the development challenges is in designing a core framework that coordinates the related data, knowledge, interactions and personalized services in facilitating patients' self-care.

One of the problems in using English in Thai students is vocabulary limitation and the structure differences between English-Thai. This paper proposes a methodology to ease learning curve for English content by adding an English-Thai translation feature, including machine translation, in Thai open source learning management system called LearnSquare. This feature enables users to translate both in sentence level and word level.

One of the problems on using English in Thai students is vocabulary limitation. This paper proposes a methodology to ease learning curve for English content by adding an English-Thai dictionary based translation features in Thai open source learning management system called LearnSquare. Each vocabulary is automatic translated by placing a pointing device over them.

The necessity on exchanging information among countries become a major task in information based society. Machine translation is an application that enables users to communicate each other without language barrier problem. With the great support on computer's efficiency, corpus-based technology becomes a fundamental concept for developing software based on a large amount of data. We introduce the first example-based English to Thai machine translation using n-gram model and implemented the system. Some advantages and disadvantages of this method are discussed.

Rule-Based Machine Translation (RBMT) [1] approach is a major approach in MT research. It needs linguistic knowledge to create appropriate rules of translation. However, we cannot completely add all linguistic rules to the system because adding new rules may cause a conflict with the old ones. So, we propose a memory based approach to improve the translation quality without modifying the existing linguistic rules. This paper analyses the translation problems and shows how this approach works.