最新刊期
卷 11 ， 期 5 ， 2025
-
《大数据》十周年寄语 AI导读
-
- 摘要：Computer education is one of the requirements of modern information society education. With the development of large language models, there has been increasing attention on applying of large language models to the computer education process. However, the hallucination problem associated with large language models poses significant challenges to its application. To solve the challenges, RAG techniques by incorporating external knowledge bases can effectively enhance the quality of responses generated by large language models. However, the traditional RAG techniques lack a fine screening mechanism for the retrieved information, which leads to the retention of a large amount of low-correlation knowledge, and the interference of irrelevant information makes the model hallucination problem not effectively solved.We collected computer-related textbooks and knowledge documents, dividing them into knowledge document blocks according to the content structure to construct an external knowledge database. On this base, we introduced the large educational models question-answering enhancement method based on re-ranking and post-retrieval reflection, which utilized a high-performance multilingual re-ranking model based on a cross-encoder to capture deep semantic information, filter the retrieval information, filter out irrelevant information to improve the retrieval quality. The proposed method applied RAG techniques for model reflection so that the model can further enhance the quality of the model's answers through self-examination, and effectively improve the accuracy of the large language model in computer question-answering. This approach significantly improves the accuracy of large language models in computer question-answering tasks. The proposed method has been tested on several popular current generative models, achieving promising results on CS-Bench, with an approximate 5% increase in accuracy for computer question-answering tasks.关键词：large language model;retrieval-augmented generation;computer education280|288|0更新时间：2025-09-15
- 摘要：The natural language processing capabilities based on large language models have shown wide application potential in intelligent question-answering scenarios. To provide personalized computer learning support for university students, a computer public course question-answering system named "Shanshan" was developed, which was based on a code-oriented large language model. The system was designed with a front-end and back-end separation architecture, and sends user questions to the large language model through a message queue and listens to the model's return results. Leveraging the language understanding and generation capabilities of the large language model, the system was utilized to automatically answer computer-related questions posed by students, and continuous batch processing and retrieval-augmented generation techniques were employed for optimization. Performance evaluation experiments indicated that the system achieved superior performance in terms of concurrency, response time, and multi-turn dialogue compared to alternative methods.关键词：large language model;intelligent question-answering system;computer public course;batch processing;retrieval-augmented generation236|152|0更新时间：2025-09-15
- 摘要：The automatic solving of MWP not only provides students with accurate learning guidance but also helps alleviate teacher workload and promotes advancements of smart education. Existing MWP-solving methods face challenges in extracting deep semantic information and unifying the solutions for various types of MWP , and addressing the subtle differences in similar problems. To address these challenges, a general MWP solving model, BSCL, was proposed. Firstly, pre-trained language models were used to encode MWP in natural language form, employing contrastive learning methods to enhance the encoder's ability to understand different MWP types. Then, the model unified the decoding of various types of MWP and used a supervision task to ensure mathematical consistency between the problems and expressions. Extensive experiments on both Chinese and English datasets indicate that BSCL has the effectiveness and superiority in solving different types of MWP tasks.关键词：math word problem;natural language processing;contrastive learning231|158|0更新时间：2025-09-15
- 摘要：With the advent of large language models, their powerful language and reasoning abilities can mimic the question design methods of teachers, analyze the materials for questions, generate corresponding questions, and ensure the quality of the generated questions through self-checking. Inspired by this, a knowledge graph-enhanced large language model historical question generation system (KG-QGH) was designed, aiming to generate questions that meet teachers' needs, are content in accurate, aligned with the syllabus, and diverse in form. Taking short answer questions in the history subject as an example, the system automatically generated questions by integrating relational information from the historical knowledge graph with textbook knowledge points, a feedback mechanism was introduced to verify andcorrect the generated questions. The comparative experimental evaluations was demonstrated that the KG-QGH system significantly outperformed the single large language model approach in key metrics such as correctness, relevance, and diversity of question generation. This study not only validated the effectiveness of the knowledge graph enhancement method in the field of question generation but also offered new research insights and practical references for intelligent applications in the educational domain.关键词：large language model;knowledge graph;question generation;history education;intelligence education381|303|0更新时间：2025-09-15
- 摘要：With the rapid advancement of artificial intelligence technology, LLMs have achieved significant success across variousfields. However, their application in the field of education domain still faces challenges such as difficulties in processing multimodal data, insufficient response accuracy, and limited information delivery methods. To address these issues, a VELM was proposed. VELM was trained on multimodal public educational datasets and specialized educational datasets, and combined with model optimization techniques, VELM not only enhances response quality in educational scenarios but also optimizes and reduces computational resource consumption. Additionally, RAG technology was utilized to ensure accuracy and richness in generated content. In terms of deployment and application, VELM was implemented through the Dify platform, enabling flexible multi-end deployment, including WeChat mini programs, Web cloud platforms, and localized deployment, meeting the diverse needs of different educational scenarios. Evaluation experiments demonstrated that VELM significantly outperformed open-source large models such as MiniCPM-V, DeepSeek-VL, and Yi-VL on standard benchmark datasets like Mathvista, OCRBench, and MMMU. On specialized educational evaluation datasets, the accuracy of VELM was improved by 9.78% compared to the base model Qwen2-VL.关键词：large language model;multimodal;smart education;RAG technology326|240|0更新时间：2025-09-15
- 摘要：Knowledge tracing aims to track the transformation of learners’ knowledge states during the learning process, thereby enhancing learners’ learning efficiency. Recently, due to its significant importance in education, knowledge tracing has attracted considerable attention from researchers. However, deep learning-based knowledge tracing models primarily focus on achieving high accuracy in predicting learners’ performance, but they significantly lack interpretability. Balancing the interpretability of knowledge tracing models while enhancing their performance remains a challenge. Based on this, a meticulously designed, high-performance, interpretable knowledge tracing model TIKT, was firstly proposed using item response theory (IRT) and Transformer. Subsequently, a fine-tuned pretrained model to predict the implicit difficulty levels of question texts to enhance data-side interpretability was used, and incorporated thesedifficulty levels into a contrastive learning framework to boost the overall performance of the knowledge tracing model. Experiments conducted on three real-world benchmark datasets conclusively demonstrate that our proposed model achieves improvements in both interpretability and performance relative to other models.关键词：knowledge tracing;contrastive learning;pretrained model231|150|0更新时间：2025-09-15
- 摘要：With the continuous advancement of quality education in our country, the influence of aesthetic education in subject education is becoming increasingly important. Appreciation of artworks is one of the important contents of aesthetic education, which can cultivate students' artistic ability and literacy. However, the lack of excellent art teachers and the imbalance of the development level of art education in various regions has led to many students being unable to receive high-quality art appreciation education. In this case, using multimodal large language models to tutor students in art appreciation has become a potential alternative. Using a multimodal large language model, this paper proposes a method based on graph retrieval augmented generation and few-shot learning to guide the model to generate artappreciation content that meets the needs of high school education. Experimental results show that compared with the comparison methods, this method can effectively improve the quality of the appreciation content generated by the multimodal large language model.关键词：graph retrieval augmented generation;art education;multimodal large language model201|166|0更新时间：2025-09-15
TOPIC
- 摘要：Wuxi, located at the geometric center of the Yangtze River Delta, has been historically known for its role as a dock for rice, cloth, silk, and finance. The "1+8+X" model for constructing data element markets was developed by Wuxi, leveraging its unique resource endowments, industrial strengths, and historical experience with docks. This model was designed to establish a "Chinese Digital Dock", which was characterized by significant advantages in high marketization, robust security protections, and low circulation costs, collectively known as "two highs and one low". The initiative was aimed at attracting both domestic and international data for storage, processing, circulation, and trade within Wuxi. Common challenges in the domestic market, such as the lack of standardized data market regulations, suboptimal data supply, poor platform interoperability, and insufficient industrial carriers, were overcome through Wuxi's efforts. Positive progress and results were achieved. These experience provide valuable insights for other large and medium-sized cities seeking to establish their own data element markets.关键词：data element market;large and medium-sized city;Chinese digital dock;innovative practice;model255|186|0更新时间：2025-09-15
- 摘要：With the rapid development of deep learning technology, fake audio generation technology, especially deep fake audio generation, has become a major challenge in the field of digital media security. The audio generated by these technologies is highly realistic and difficult to distinguish from real speech, which puts forward higher requirements for social security and personal privacy protection. In view of this, the research of deep forgery audio authentication technology was particularly urgent, although some progress has been made in this field, but with the continuous emergence of data volume, computing power and new models, new challenges also follow. Firstly, the evolution of fake audio generation techniques was outlined, transitioning from traditional audio editing to deep learning-based generation models. Subsequently, an in-depth analysis was conducted on both acoustic feature-based and end-to-end model-based fake audio detection strategies, delving into details such as deep acoustic feature detection, pre-trainedneural network feature detection, end-to-end model optimization, generalization enhancement techniques, and the enhancement of real-time detection. Lastly, future work was envisioned, highlighting the key research directions that require attention, in order to advance the development of deep fake audio detection techniques, more effectively counter increasingly complex fake audio threats, and ensure the security and reliability of digital media.关键词：deep fake audio;audio forgery detection;deep audio generative model;detection technique365|306|0更新时间：2025-09-15
- 摘要：In view of problems such as objects are interfered by complex background, objects are small and densely distributed, objects are multi-scale and their directions are random in high resolution remote sensing image data, an object detection algorithm for high resolution remote sensing image based on improved YOLOv8 was proposed. Firstly, dynamic snake convolution was incorporated to make the algorithm detect objects with different scales and directions better; Secondly, in order to enable the algorithm to capture the global context information in the image with complex background, iRMB was combined with Shift-Wise convolution. Attention-based Intra-scale feature interaction was used. Bi-directional feature pyramid network and semantics and detail infusion module were combined as the algorithm's feature fusion network. An improved efficient decoupled detection head was designed. Finally, combining NWD and WIoU loss functions, a new loss function NWD-WIoU was designed to accelerate the convergence speed of algorithm detectionand improve the detection performance of the algorithm for small objects. Experimental results on three public remote sensing image object detection datasets show that the high resolution remote sensing image detection algorithm based on improved YOLOv8 performs better than other mainstream high resolution remote sensing image detection algorithms.关键词：deep learning;high resolution remote sensing image;object detection;YOLOv8;lightweight502|221|0更新时间：2025-09-15
STUDY
- 摘要：With the explosive growth of data volume in the era of the digital economy, data platforms are facing increasing challenges in data storage, processing, and calculation. This paper proposes multiple Data Lakes under a unified Data Middle Platform, which is applied in industrial big data. The architecture, based on the integration of lake and warehouse technology, aims to build an efficient, flexible, and scalable data management platform to support cross-level, cross-organizational, and cross-regional data-sharing applications. With the centralized deployment of the three-level data lake and the middle platform deployed uniformly, the five core functions are constructed, forming the dataacquisition capabilities of multi-source heterogeneous data, multi-mode data storage, data processing, data governance, data service, and data security protection. Through multi-data-lakes, collaborative interaction, and unified data control ability, to realize the overall management and open sharing of data of super-large group companies, application scenarios in various business fields are supported, and the value of data elements is released.关键词：Data Lake;Lakehouse;data middle platform;industrial big data;data governance;data service304|254|0更新时间：2025-09-15
- 摘要：As a key production factor, data assets are gradually evolving into a crucial engine driving China’s stable growth and economic transformation. However, the circulation and trading of data assets face multiple challenges, including issues related to data authenticity and privacy before transactions, security and trustworthiness during transactions, and reliability and traceability after transactions. In response to these challenges, the study proposes an on-chain and off-chain data asset circulation and trading scheme that integrates blockchain and federated learning technologies. The study focuses on establishing a trustworthy data interaction framework, technologies for ownership confirmation and privacy protection of multi-source sensitive data, as well as full-lifecycle data trustworthiness assurance and abnormal transaction traceability mechanisms. This approach ensures high security, privacy, and traceability in the circulation and trading of data assets, aiming to build an efficient, secure, and trustworthy data asset trading system that promotes thesafe, orderly, and sustainable development of the data element market, thereby providing technical support for the high-quality development of China’s digital economy.关键词：data asset;data trading;blockchain;federated learning260|147|0更新时间：2025-09-15
APPLICATION
- Postal code：100079
- Tel：（010）53879208、53859533 Email：bdr@bjxintong.com.cn
- Technical support is provided by Beijing Founder electronics co., LTD 京ICP备09064830号-19
京公网安备11010802024621 - It is recommended to read the content of this site in Chrome&IE9+. Please switch to extreme mode in browser 360.
- Cookies We use cookies to help provide and enhance our service and tailor content. By continuing, you agree to the use of cookies.
- Total Visits：273110 Visits Today：16
0