Assessing student engagement is critical for collaborative learning but remains a challenging task. Existing approaches often rely on controlled laboratory or online settings, which fail to capture the complexity of real-world classrooms. Furthermore, current datasets are scarce and rarely provide both individual-and group-level annotations, limiting the development of robust and generalizable models. To address these gaps, we propose CORE-Net, a multimodal architecture that integrates context modeling to capture group-level dynamics and ordinal supervision to account for the ordinal nature of engagement levels. We also present COLER, a large-scale dataset collected in authentic classroom environments with rich annotations at multiple levels. Experiments demonstrate that CORE-Net achieves 89.63% accuracy and 94.80 QWK, significantly outperforming strong baselines such as BlockGCN and MoViNet. Ablation studies further highlight the critical role of both context modeling and ordinal supervision. Our work establishes a robust and scalable foundation for automated engagement assessment, supporting timely feedback and enhancing the effectiveness of collaborative learning.

Legal queries are often expressed in unstructured, ambiguous, or noisy natural language, which poses significant challenges for accurate information retrieval. This paper presents a comprehensive framework for legal question answering that integrates semantic technologies—including an ontology platform, knowledge graph, and large language models (LLMs)—to improve question understanding and response accuracy. We propose a multi-step pre-processing pipeline that standardizes legal questions using spelling correction, abbreviation expansion, syntactic restructuring, and semantic summarization supported by LLMs. Besides, a query system is designed that maps pre-processed questions into graph-based representations and performs subgraph matching over a legal knowledge graph. The system is evaluated on real-world legal questions collected from online forums covering traffic law and social insurance. The results show that the proposed approach achieves a high semantic similarity score (avg. cosine similarity of 0.9078 after standardization) and outperforms baseline LLMs like ChatGPT and Gemini in query accuracy (up to 82.25% in certain question categories). These findings highlight the effectiveness of combining LLMs and semantic structures for robust legal information retrieval.

With the rapid development of Generative Artificial Intelligence (Generative AI), Sovereign AI—a vital vehicle for national digital sovereignty—has introduced new technical challenges for value alignment. Traditional alignment methods struggle to effectively address issues such as multicultural divergence, the abstract complexity of ethical principles, and the high computational cost known as the “alignment tax” under Sovereign AI. To tackle these challenges, a multi-granular semantic reasoning method is proposed for value alignment based on the DIKWP framework, designed explicitly for Sovereign AI. By mapping the interrelations among Data, Information, Knowledge, Wisdom, and Purpose (DIKWP), a dual system of intention and behavior is constructed to resolve semantic conflicts, fragmented value expressions, and dynamic adaptation in complex contexts. To reduce alignment costs, we design a dedicated, prompt template—DIKWP Chain of Thought(DIKWP-CoT)—that explicitly embeds value-behavior logic into the model’s reasoning process, significantly minimizing retraining needs. Experimental results show that, compared to traditional approaches, the proposed method significantly improves the accuracy and robustness of value alignment on the CoreValue and CMOS datasets, especially in complex scenarios involving implicit intentions or biased expressions. This provides a novel technical pathway for building safe, trustworthy, and sovereignty-aligned generative AI systems.

In Vietnam, accessing accurate and personalized legal advice remains a significant challenge due to the complexity of legal documents and limited public legal literacy. This paper introduces an intelligent legal consulting system that integrates three key components: the Legal-Onto knowledge representation model, the VietnamLegalText-SBERT embedding model, and a ReAct-based multi-agent framework. Legal-Onto organizes legal knowledge into semantic graphs of legal concepts, capturing internal relationships, definitions, and key attributes from legal texts. Each node in the knowledge graph is embedded using VietnamLegalText-SBERT—a BERT-based model fine-tuned on Vietnamese legal corpora—and stored in Elasticsearch to enable efficient vector-based retrieval. A multi-agent architecture facilitates step-by-step reasoning, query decomposition, document retrieval, legal interpretation, and output validation. The system is evaluated on real-world legal queries collected from public sources, showing superior performance over a general-purpose GPT-4o baseline in terms of answer accuracy, citation correctness, and contextual relevance. The results demonstrate that combining domain-specific semantic modeling with reasoning agents yields interpretable, reliable, and context-aware legal support, paving the way for broader public access to legal guidance in Vietnam.

Nowadays online learning is growing strongly, with many diverse options. Learners can utilize electronic devices and internet platforms for studying, searching for materials, and looking up knowledge. In this study, a method for design a knowledge chatbot system in education is proposed. This method includes a knowledge base, which is a knowledge model integrating ontology of relations and operators and knowledge graph, and the algorithms for solving problems on intellectual querying. The proposed method is utilized to build an intelligent chatbot system in the course of Foudation of Database. This chatbot can assist the online learning by querying subject knowledge content, categorizing concepts, and providing support for various types of exercises in the subject. The experimental results show that the built system meets the requirements of an intelligent educational system and outperforms than currently GenAI systems.

The revelation of cognitive knowledge, development, and planning of distinct economic societies in human civilization have mostly been influenced by historically asymmetric availability or ownership of Data Asset, Information Asset, Knowledge Asset, and Wisdom Asset. The availability of asymmetric information and the asymmetry of demand for commercial goods or services serve as the foundation for many economic models and theories. However, from the DIKWP (Data, Information, Knowledge, Wisdom, and Purpose) Capital materialization and DIKWP Governance perspective, with the rapidly development of information technology and widely progressing digital communication facilities, Asymmetric Information Economy is increasingly replaced with essentially Symmetric Knowledge Economy and Symmetric Wisdom Economy. The inevitability of the replacement of Asymmetric Economy by Symmetric Economy in terms of DIKWP-12-Chains is formalized, and proposed uniformly semantic processing crossing DIKWP.

Nowadays the video game market is the highest in the entertainment industry. There are many architectures supporting the creation of a game, such as Unreal Engine and Unity. Multi-agent system is a popular and suitable solution to design platform games. It is employed to place streets, buildings and other items, resulting in a playable video game map. The system utilizes computational agents that act in conjunction with the human designer to produce maps that exhibit desirable characteristics. This paper proposes a new architecture for organizing in game development based on the multi-agents system (MAS). This architecture includes the structure of an agent with its attributes and internal behaviors, and the structure of relations between agents for action coordination in a determined strategy This structure of MAS is also applied to build a demonstration video game as an action game through Unreal Engine, which is a complete suite of development tools made for anyone working with real-time technology.

Buying a house is considered to be an increasing necessary demand. However, choosing the right one is a complex and time-costing decision. In this research, a method for knowledge representation of an expert system in real-estate investment is proposed. This method organizes the knowledge base of the system based on combining market surveying data in real-estate for each category of customers. Through that, a content-based searching system using the personal demographics, as an expert system, is proposed. This system supports both customer and real estate agent to find the suitable house fitting their desire and budget. This system is designed to extract information of the customer information to enhance the conventional searching method which only treat house property as the only input.

This paper proposes a new deep learning model to detect COVID-19 lesions in chest CT images. This method is based on the Attention U-net which uses the layer of Atrous Spatial Pyramid Pooling (ASPP) to capture the feature on various scales. It also contains an attention gate. The attention gate provides the ability to suppress irrelevant regions and focus on the useful feature in an input image. The experimental results show that this method can achieve 99.61% accuracy and 80.43% precision. They are more effectively than the baseline method on Chest CT images.

Vietnamese is a tonal and isolated language. Its highly ambiguity makes the designing of methods for sentiment analysis being difficult. For getting the most effectiveness, the designed method has to analyze sentiment of sentences based on combining the grammar and syllable structures of Vietnamese. In this paper, a method to build a Vietnamese dataset of product reviews with many sentiment levels, including very negative, negative, neutral, positive and very positive, is proposed. This method can be scaled to a large dataset using for analyzing sentiment of product reviews. Moreover, a solution to add more grammar rules of Vietnamese into the pre-processing of sentiment analysis is also constructed. Those rules simulate the sentiment recognition of humans and help to increase the accuracy of sentiment determination. The combination of grammar rules and some methods for sentiment analysis are experimented on the Vietnamese dataset of product reviews to classify them into sentiment-levels. The testing results show that their accuracy and F-measure are improved and suitable to apply in the practical business analyzing of customer behaviors.

Accurately assessing the air quality index (AQI) values and levels has become an attractive research topic during the last decades. It is a crucial aspect when studying the possible adverse health effects associated with current air quality conditions. This paper aims to utilize machine learning and an appropriate selection of attributes for the air quality estimation problem using various features, including sensor data (humidity, temperature), timestamp features, location features, and public weather data. We evaluated the performance of different learning models and features to study the problem using the data set “MNR-HCM II”. The experimental results show that adopting TLPW features with Stacking generalization yields higher overall performance than other techniques and features in RMSE, accuracy, and F1-score.

Abstract In this paper, we introduce methods of encoding propositional logic programs in vector spaces. Interpretations are represented by vectors and programs are represented by matrices. The least model of a definite program is computed by multiplying an interpretation vector and a program matrix. To optimize computation in vector spaces, we provide a method of partial evaluation of programs using linear algebra. Partial evaluation is done by unfolding rules in a program, and it is realized in a vector space by multiplying program matrices. We perform experiments using artificial data and real data, and show that partial evaluation has the potential for realizing efficient computation of huge scale of programs in vector spaces.

Social network is one of efficient tools for spreading information. The evaluation of the content creation of a user is a useful feature to improve the ability of information propagation on social network. In this paper, the measures for evaluating the user’s content creation are proposed. They include the passion point of a user with a brand and the quality of the user’s posts. The passion point is computed based on the sentiment score of posting and the activity of the user. The quality of the user’s posts is computed through the analyzing of the post’s content. Those measures are combined to analyze the interesting of posts. The proposed method has been tested and get the positive experimental results.

Nowadays building the intelligent systems for education plays an important role to construct the smart city. Knowledge engineering provides technology to build the Intelligent Tutoring System (ITS) supporting for learning. In this paper, a model for representing the knowledge of courses about algorithms is presented. This knowledge model is the integrating between ontology and frames, called Integ-model. Based on the structure of Integ-model, problems about querying on the knowledge and visualizing of algorithms have been proposed and solved. The Integ-model is applied to organize the knowledge base of the intelligent system for learning Data Structures and Algorithms. This system can illustrate algorithms's process and the working of data structures in the course. It also helps learners to query the content of lessons in the course.

The active learning methods bring enthusiasm for studying of students, especially in learning of algorithms and programming. Graphical and dynamic web based tools as visualization in learning these courses make more attractive to learners for spending much more their study time. In this paper, a method for designing an Intelligent system in learning of algorithms is proposed. This method proposes the model for representation the knowledge of an algorithm, called Algo-model. This model has the structure of tasks which are actions in the algorithm's processing. The proposed method also includes the processing to visualize the happen change in an algorithm's running. It will visualize on three contents synchronously: displaying the current state of the algorithm, results of data structures at this state, and visualizing this current state on the graphic environment. Based on the proposed method, an Intelligent learning system for Graph Theory at university has been constructed. This system meets requirements of an Intelligent system in learning of algorithms. It is useful for students to study and understand how the algorithm runs. It can interact to the student step-by-step based on the visualization of algorithms in learning tutors.

Intelligent software has been very useful in education, especially those support studying knowledge. The knowledge of Discrete Mathematics and Graph Theory are essential in STEM education. There have been many computer programs that support students to study in the above knowledge domains such as Wolfram Alpha, Maple, etc. However, they do not satisfy students' requirements. This article will present knowledge representation methods, reasoning and querying techniques to design and to implement an intelligent software for effectively studying Discrete Mathematics and Graph Theory. The system consists of two subsystems for knowledge retrieval, visualization of methods or algorithms, and for solving problems. It can be easily used for studying the knowledge as well as for solving problems.

Knowledge about relations plays a crucial role in human’s knowledge. Different methods for representing this type of knowledge have been proposed. However, due to the lack of theoretical foundations, these methods cannot guarantee criteria in knowledge representation such as formality, universality, usability and practicality. They are not adequate to represent the knowledge domains in practice which have many components. Based on formal ontology approach, a knowledge model about relations, called Rela-model, is presented in this paper. It has the components such as concepts, relations between concepts, and rules. The concepts in this model consist of attributes, facts and rules of itself. Each object in a concept has also equipped its behavior to solve problems on it. The methods for solving problems based on Rela-model are also studied. The general problems on this model are the following: Given some objects and facts on them, determine the closure of set of attributes and facts on the objects or determine an object or consider a relation between the objects. The algorithms to solve problems are designed and their properties, such as finiteness, effectiveness, have also been proved. Besides the solid mathematical foundation, Rela-model also has a simple specification language which can effectively represent the knowledge, thus it can be used in many real situations. Our approach is also applied to build two systems: the intelligent problem solver about solid geometry in high school mathematics, and the expert system to diagnose diseases in diabetic microvascular complication.

A grand challenge for artificial intelligence in education is building the Intelligent Problem Solver (IPS) for Science Technology Engineering and Math (STEM) Education. The IPS system has to be able to solve the exercises of the course automatically. It has the following criteria: the knowledge base is sufficient, the program can solve the common exercises in the curriculum of the course based on the knowledge base, the solutions are readable, pedagogical and suitable for the learner's level. Discrete Mathematics is an important course for the undergrad technological curriculum at the university. In this course, knowledge about logic and Boolean algebra is the foundation of logical thinking, it helps students improve their skills in logical reasoning, solving the problems. There are many programs for solving problems in propositional logic and first-order logic; nevertheless, they cannot meet the requirements of a learning support system. In this paper, an IPS system in knowledge domain about logic and Boolean algebra has been proposed. This system satisfies the criteria of the STEM education. It helps students to understand the methods for solving basic and advanced problems: simplify the logical expression in propositional logic, reasoning checking, determine the value or the negative expression of of a logical expression in predicate logic, find the minimization expression of a Boolean function with parameter and non-parameter. In this system, the knowledge base about propositional logic, predicate logic and Boolean algebra at the university for undergraduates has been built based on knowledge model of operators. Via this knowledge base, the inference engine has been designed to solve the kinds of general problems in this knowledge domain. The program has been also tested by the students in University of Information Technology, VNU-HCM.