A teaching file digitization automatic indexing and multi-dimensional correlation analysis system

Abstract

The application discloses a kind of teaching archives digitization automatic indexing and multidimensional association analysis system, it is related to education information technology field, including: data acquisition module gathers multiple-source heterogeneous teaching archives data and is converted into unified format data to be handled;Intelligent indexing module is according to teaching field special glossary to the semantic analysis and multimodal adaptation of data to be handled, generates multidimensional indexing information;Multidimensional association analysis module identifies direct association according to multidimensional indexing information, excavates implicit association and carries out time series evolution analysis, reveals internal law of teaching;Knowledge graph construction module extracts teaching entity, establishes entity relationship based on association analysis result, constructs dynamic updating knowledge graph;Application service module supports retrieval based on knowledge graph, personalized push and visual analysis, and sets fine-grained access permission.The application can improve teaching resource management efficiency and utilization value, guarantee data security and support teaching decision optimization.

Description

Technical Field

[0001] This invention relates to the field of educational informatization technology, and more specifically to a system for automatic digital indexing and multi-dimensional correlation analysis of teaching records. Background Technology

[0002] With the accelerated advancement of digital transformation in education, the need for digital management of teaching archives—as core resources for recording the teaching process, accumulating teaching experience, and evaluating teaching quality—is becoming increasingly urgent. Current technologies have achieved the transformation from paper documents to electronic resources in digitized teaching archive solutions, and some systems have introduced automatic indexing and correlation analysis functions. However, several core issues still need to be addressed: 1) Insufficient Standardization in Intelligent Indexing: Current indexing of educational archives still largely relies on manual operation, resulting in inconsistent indexing quality and low efficiency. Although some studies have attempted to apply natural language processing techniques, general models struggle to fully understand the specialized terminology and context of the educational field, leading to limited indexing accuracy and a high rate of misclassification for polysemous words and implicit concepts. Furthermore, the lack of a dedicated thesaurus built specifically for the characteristics of the educational field makes it difficult to support accurate semantic mapping and multimodal adaptation.

[0003] 2) Limited Dimensions in Association Analysis: Most existing systems are limited to direct association analysis based on keyword matching, lacking the ability to mine implicit associations in teaching archives across multiple dimensions such as entities, semantics, and spatiotemporal relationships. Traditional methods struggle to reveal complex relationship networks such as "teacher-course-knowledge point," failing to effectively support the discovery of inherent teaching patterns and decision optimization. The phenomenon of data silos also exacerbates the difficulties of association analysis, making it difficult to integrate and analyze data from different sources and in different formats.

[0004] 3) Weak heterogeneous data fusion capability: Teaching archives come from diverse sources, including text, images, audio, video, and other modalities, resulting in highly heterogeneous data structures. Existing technologies lack the ability to uniformly represent and process multimodal data, particularly in the deep fusion of information after video keyframe extraction and OCR recognition. Conflict correction, quality control, and standardization during data fusion also pose challenges, impacting the quality of subsequent intelligent services.

[0005] 4) Lack of dynamic updating and service capabilities for knowledge graphs: While some advanced solutions have introduced knowledge graph technology, they are relatively weak in the knowledge reasoning stage after entity recognition and relation extraction. The knowledge graphs are often static after construction and are difficult to automatically update with new files or indexing system optimizations. Furthermore, existing systems still have significant room for improvement in providing personalized and visualized knowledge services to different users, particularly in terms of fine-grained access control and intelligent push capabilities.

[0006] Therefore, researching a system that can achieve unified processing of multi-source heterogeneous data, possess precise indexing and multi-dimensional correlation analysis capabilities, and provide personalized services based on dynamic knowledge graphs has significant theoretical and practical value and is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention provides an automatic indexing and multi-dimensional correlation analysis system for digital teaching archives, which solves the problems existing in the background technology.

[0008] To achieve the above objectives, the present invention provides the following technical solution: A system for automatic indexing and multi-dimensional correlation analysis of digital teaching archives, the system being implemented based on a multi-module collaborative architecture, wherein: The data acquisition module is used to collect multi-source heterogeneous teaching archive data and convert it into data to be processed in a unified format; The intelligent indexing module is used to perform semantic analysis and multimodal adaptation on the data to be processed based on a special vocabulary for the teaching field, and generate multi-dimensional indexing information. The multi-dimensional correlation analysis module is used to identify direct correlations between teaching archives based on multi-dimensional indexing information, mine implicit correlations between teaching elements and conduct time-series evolution analysis, reveal the inherent laws of teaching and provide data support for teaching decisions. The knowledge graph construction module is used to extract teaching entities, establish entity relationships based on the results of association analysis, and construct a dynamically updated knowledge graph. The application service module supports knowledge graph-based retrieval, personalized push notifications, and visual analysis, and allows for setting fine-grained access permissions.

[0009] Optionally, the data acquisition module includes: a multi-source data access unit and a data format standardization unit; The multi-source data access unit is used to acquire multi-source heterogeneous teaching archive data through batch uploading, data collection via interface with the academic affairs system, and intelligent scanning of paper archives. The data format standardization unit is used to transform multi-source heterogeneous teaching archive data into data to be processed in a unified format through OCR recognition and conversion, video keyframe extraction and conversion, and audio-to-text conversion.

[0010] Optionally, multi-source heterogeneous teaching archive data includes structured data and unstructured data; among which, structured data includes teaching plans, course information, student grades and teacher information, and unstructured data includes lesson plan documents, test papers, courseware, teaching reflections, teaching videos, classroom audio and experimental images.

[0011] Optionally, the intelligent indexing module includes: a vocabulary construction unit, a semantic analysis unit, and a multimodal adaptation unit; The vocabulary building unit is used to expand upon the Chinese Archives Thesaurus as a base vocabulary, incorporating subject classifications, teaching procedures, and knowledge points to form a multi-dimensional, hierarchical vocabulary for teaching fields. The semantic analysis unit is used to preprocess and extract features from the data to be processed, and to perform semantic mapping and matching between the extracted features and a special vocabulary for the teaching field to complete semantic parsing. The multimodal adaptation unit is used to perform indexing adaptation on the feature data corresponding to the text, video, and audio data to be processed, and output multi-dimensional indexing information including topic, subject, and teaching link.

[0012] Optionally, the vocabulary building unit is also used to establish a dynamic optimization mechanism for a vocabulary specifically for the teaching field. Based on the search hit rate and indexing adaptation effect data fed back by the application service module, the vocabulary and hierarchical relationships in the vocabulary specifically for the teaching field are updated regularly.

[0013] Optional, the multi-dimensional association analysis module includes: a direct association identification unit, a latent association mining unit, a time-series evolution analysis unit, and a pattern extraction unit; The direct association identification unit is used to compare the multi-dimensional indexing information corresponding to different teaching files and determine that teaching files with indexing term overlap exceeding a preset threshold are directly associated. The implicit association mining unit is used to extract potential topics corresponding to multi-dimensional indexing information through topic models, and combine association rule algorithms to mine the association relationships between different potential topics to obtain implicit associations between teaching elements. The temporal evolution analysis unit is used to extract time attribute information from multi-dimensional indexing information, sort out the corresponding relationships of teaching archives according to the time dimension, analyze the evolution trend of teaching content and teaching relationships in different time periods, and complete the temporal evolution analysis. The pattern extraction unit is used to summarize and integrate the results of analysis on direct relationships between teaching files, implicit relationships between teaching elements, and time-series evolution, to screen out statistically significant relationships, and to extract the inherent teaching patterns related to teaching methods and teaching effects, and the suitability of knowledge points and exercises among teaching elements.

[0014] Optionally, the knowledge graph construction module includes: a teaching entity extraction unit, an entity relationship establishment unit, and a graph dynamic update unit; The teaching entity extraction unit is used to extract core information from multi-dimensional indexing information and multi-source heterogeneous teaching archive data to obtain teaching-related core entities, including teachers, students, courses, knowledge points, and teaching archives. The entity relationship establishment unit is used to establish and store entity relationships such as teacher-lecture-course, course-include-knowledge point, teaching file-association-course, and knowledge point-adaptation-exercise based on the direct association between teaching files and the implicit association between teaching elements. The knowledge graph dynamic update unit is used to automatically trigger the process of supplementing and extracting teaching entities and updating entity relationships after the indexing and association analysis of newly added teaching files are completed, or after the indexing system is optimized and adjusted. This synchronously updates the entity and relationship data in the knowledge graph, thereby realizing the dynamic update of the knowledge graph.

[0015] Optionally, the application service module includes: a knowledge graph retrieval unit, a personalized push unit, a visualization analysis unit, and a permission management unit; The knowledge graph retrieval unit supports retrieval methods based on index terms and entity relationships. It is used to traverse entity and relationship data in the knowledge graph according to the search conditions input by the user, accurately match and return the corresponding teaching files and related resources. The personalized push unit is used to push appropriate teaching resources to teachers and personalized learning resources to students based on the relationship between teaching entities in the knowledge graph and combined with user identity attributes and historical behavior data. The visualization analysis unit is used to display the distribution of teaching resources, the coverage of knowledge points, the relationship between teaching and the evolution of time sequence in the form of charts, intuitively presenting the inherent laws of teaching and providing visual support for teaching decisions. The access control unit is used to configure different access permissions for teachers, students, and administrators using a role-based hierarchical authorization mechanism, allowing only authorized users to access teaching files and analysis data within their corresponding permission scope.

[0016] As can be seen from the above technical solution, compared with the prior art, the present invention discloses a system for automatic indexing and multi-dimensional correlation analysis of teaching archives, which has the following beneficial effects: (1) Efficient and standardized data collection: Through multi-source access and format standardization processing, it fully covers structured and unstructured teaching archive data, breaks down heterogeneous data barriers, ensures the efficiency of unified data access, and lays a high-quality data foundation for subsequent processes. (2) Indexing is accurately adapted to teaching scenarios: Based on a special terminology for the teaching field and a multimodal adaptation process, it generates accurate indexing information including themes, subjects and other dimensions. The terminology can be dynamically optimized to continuously improve the indexing adaptability and accuracy, and solve the problem that traditional general indexing does not meet teaching needs. (3) In-depth and comprehensive association analysis: Through direct association identification, implicit association mining and time series evolution analysis, the association between teaching archives and elements is sorted out in all aspects, and the core teaching laws such as teaching methods and effects, and the suitability of knowledge points and exercises are accurately extracted, providing high-value data support for teaching decisions; (4) Dynamic and intuitive knowledge graph: accurately extract core teaching entities and establish multiple relationships to form a dynamically updated knowledge graph, clearly present entity relationships, improve the efficiency of resource association query, and ensure the timeliness and completeness of the knowledge graph; (5) Personalized and secure application services: Based on knowledge graphs, accurate retrieval and personalized resource push are realized. The teaching rules are presented intuitively through visual charts. Combined with the role-based authorization mechanism, it not only improves the efficiency of teaching resource utilization and management convenience, but also ensures data security and adapts to the differentiated needs of different users. (6) Full-process collaboration and high efficiency: The functions of each module are closely connected. The dynamic optimization mechanism of the indexing system and knowledge graph, as well as the collaborative linkage between modules, realize the full-process automated closed-loop operation of teaching archive processing from collection to application, significantly improve the overall processing efficiency and service accuracy, and help the digital transformation of education and the improvement of teaching quality. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0018] Figure 1 The architecture diagram of the teaching archive digital automatic indexing and multi-dimensional correlation analysis system provided by the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] This invention discloses a system for automatic indexing and multi-dimensional correlation analysis of digital teaching archives, such as... Figure 1 As shown, the system is implemented based on a multi-module collaborative architecture, wherein: The data acquisition module is used to collect multi-source heterogeneous teaching archive data and convert it into data to be processed in a unified format; The intelligent indexing module is used to perform semantic analysis and multimodal adaptation on the data to be processed based on a special vocabulary for the teaching field, and generate multi-dimensional indexing information. The multi-dimensional correlation analysis module is used to identify direct correlations between teaching archives based on multi-dimensional indexing information, mine implicit correlations between teaching elements and conduct time-series evolution analysis, reveal the inherent laws of teaching and provide data support for teaching decisions. The knowledge graph construction module is used to extract teaching entities, establish entity relationships based on the results of association analysis, and construct a dynamically updated knowledge graph. The application service module supports knowledge graph-based retrieval, personalized push notifications, and visual analysis, and allows for setting fine-grained access permissions.

[0021] In existing technologies, the pain points of data acquisition modules lie in: incomplete coverage of multi-source heterogeneous teaching archives, difficulty in unifying formats, and a lack of standardized processing procedures, resulting in a weak data foundation for subsequent indexing and analysis. Existing systems often only support single-type data acquisition and lack the ability to process unstructured data such as paper archives and audio / video, leading to chaotic data formats and severely impacting processing efficiency. Therefore, this embodiment designs the data acquisition module to include a multi-source data access unit and a data format standardization unit. The multi-source data access unit is used to acquire multi-source heterogeneous teaching archive data through batch uploading, data collection via interface with the academic affairs system, and intelligent scanning of paper archives. The data format standardization unit is used to transform multi-source heterogeneous teaching archive data into data to be processed in a unified format through OCR recognition and conversion, video keyframe extraction and conversion, and audio-to-text conversion.

[0022] The advantage of the above design is that it achieves full-scenario coverage of teaching archive data. The three collection methods correspond to existing electronic data, native data of the academic affairs system, and paper-based data respectively, ensuring no omissions. At the same time, through targeted format conversion technology, different types of heterogeneous data are unified into a standardized format, completely breaking down data silos and providing a high-quality and consistent data foundation for subsequent intelligent indexing and correlation analysis, which greatly improves the smoothness and accuracy of the entire process.

[0023] In existing technologies, the classification of teaching record data is vague, and the specific scope of structured and unstructured data is not clearly defined, resulting in insufficient targeting during data processing, omission or misprocessing of some data types, and affecting the overall system operation. Therefore, this embodiment discloses multi-source heterogeneous teaching record data, including structured and unstructured data; wherein, structured data includes teaching plans, course information, student grades, and teacher information, and unstructured data includes lesson plans, test papers, courseware, teaching reflections, teaching videos, classroom audio, and experimental images.

[0024] The above definitions clarify the specific boundaries of data collection, making the work of the data collection module more targeted and avoiding collection omissions or redundancies caused by unclear definition of data types. At the same time, the clear division of structured and unstructured data provides a clear processing basis for the subsequent data format standardization unit. Different types of data adopt adapted conversion technologies, further improving the accuracy and efficiency of data processing, and ensuring that all kinds of core teaching data can be effectively incorporated into the system.

[0025] In the prior art, the core problems faced by the intelligent indexing module are as follows: the existing indexing lacks the support of a standardized vocabulary exclusive to the teaching field, the general vocabulary is difficult to adapt to the professional terms and contexts of the teaching scenario, and the multi-modal data indexing ability is missing, resulting in low accuracy and poor consistency of the indexing results. Currently, general text indexing algorithms are mostly used, which cannot meet the indexing requirements of different modal data such as text, audio and video, and the indexing dimension is also relatively single. Therefore, in this embodiment, the intelligent indexing module is designed to include a vocabulary construction unit, a semantic analysis unit and a multi-modal adaptation unit; The vocabulary construction unit is used to take the Chinese Archives Thesaurus as the basic vocabulary, and expand and incorporate the exclusive attribute vocabulary of the teaching field such as subject classification, teaching links, and knowledge points to form a multi-dimensional hierarchical special vocabulary for the teaching field; The semantic analysis unit is used to preprocess and extract features from the data to be processed, and perform semantic mapping and matching between the extracted features and the special vocabulary of the teaching field to complete semantic parsing. Among them, the preprocessing includes Chinese word segmentation (using the jieba word segmentation algorithm), stop word filtering (filtering general stop words irrelevant to the teaching field, such as "of, and, for", etc.), synonym normalization (based on the synonym mapping of the special vocabulary of the teaching field), and at the same time, the core features are extracted by the fusion method of TF-IDF and TextRank algorithms; The multi-modal adaptation unit is used to respectively complete indexing adaptation for the feature data corresponding to the data to be processed of text type, video type, and audio type, and output multi-dimensional indexing information including theme, subject, and teaching link. Among them, the text type data is directly indexed based on the semantic analysis result, the video type data is indexed by combining the scene information of key frame recognition and the text information of audio-to-text conversion, and the audio type data is assisted in indexing based on the text + speech emotion features (such as classroom interaction activity) of the converted text.

[0026] The core advantage of the above design lies in the construction of a dedicated indexing system tailored to teaching scenarios. The basic vocabulary ensures the standardization of indexing, while the expansion of vocabulary specific to the teaching field enhances the professionalism and accuracy of indexing. The semantic analysis unit achieves deep semantic-level indexing through feature extraction and vocabulary mapping, avoiding the limitations of superficial keyword matching. The multimodal adaptation unit fills the gap in existing technology for indexing non-text data such as audio and video, and the multi-dimensional indexing information provides rich retrieval dimensions and association basis for subsequent multi-dimensional association analysis, significantly improving the practical value of the indexing results.

[0027] Furthermore, existing indexing vocabularies are statically fixed once built, failing to adapt to the updates and iterations of teaching content and the actual feedback on indexing effectiveness. This leads to a gradual decline in indexing accuracy over time, making it difficult to continuously adapt to changes in teaching needs. Therefore, the vocabulary construction unit in this embodiment is also used to establish a dynamic optimization mechanism for a dedicated vocabulary for the teaching field. Based on the retrieval hit rate and indexing adaptation effect data fed back by the application service module, the vocabulary and hierarchical relationships in the dedicated vocabulary for the teaching field are updated regularly. Specifically, there is an update cycle of 3 months. When the retrieval hit rate of a certain word is less than 30% of the average hit rate of similar words, or when there are more than 5 indexing adaptation error feedbacks, a word adjustment is triggered. Newly added words must be manually reviewed before being included in the vocabulary.

[0028] The aforementioned dynamic optimization mechanism enables the dedicated thesaurus to have self-iterative capabilities, allowing it to be continuously improved based on actual application results: words with low retrieval hit rates may be adjusted or supplemented, and word relationships with poor indexing adaptation effects can be optimized, ensuring that the thesaurus always meets actual teaching needs; at the same time, the dynamic updates of the thesaurus will in turn improve the accuracy of the entire process, including indexing, retrieval, and association analysis, forming a closed loop of "indexing-application-feedback-optimization," enabling the system's indexing capabilities to continuously evolve and maintain high adaptability and accuracy in the long term.

[0029] In existing technologies, the key technical pain point of multi-dimensional association analysis modules is that the analysis is limited to direct associations based on keyword matching, lacking the ability to mine implicit associations and conduct time-series dimension analysis. This makes it difficult to reveal the deep-seated patterns between teaching elements and to support the optimization of teaching decisions. Existing systems suffer from single-dimensional and insufficient-depth association analysis, failing to uncover potential associations such as those between knowledge points and exercises, or between teaching methods and grades, and also unable to track the temporal evolution trends of these relationships. Therefore, this embodiment designs the multi-dimensional association analysis module to include a direct association identification unit, an implicit association mining unit, a time-series evolution analysis unit, and a pattern extraction unit. The direct association identification unit is used to compare the multi-dimensional indexing information corresponding to different teaching files and determine that teaching files with an indexing term overlap exceeding a preset threshold are directly associated. For example, an indexing term overlap exceeding 60% is considered to be directly associated, and an overlap exceeding 40% for core indexing terms (such as subject and knowledge point) is considered to be directly associated. The implicit association mining unit is used to extract latent topics corresponding to multi-dimensional indexing information through topic modeling, and to mine the association relationships between different latent topics by combining association rule algorithms to obtain implicit associations between teaching elements; among them, the topic model adopts the LDA algorithm and the association rule algorithm adopts the Apriori algorithm. The temporal evolution analysis unit is used to extract time attribute information from multi-dimensional indexing information, sort out the corresponding relationships of teaching archives according to the time dimension, analyze the evolution trend of teaching content and teaching relationships in different time periods, and complete the temporal evolution analysis. The pattern extraction unit is used to summarize and integrate the results of direct correlations between teaching files, implicit correlations between teaching elements, and time-series evolution analysis. It then filters out statistically significant relationships and extracts the inherent teaching patterns related to teaching methods and effectiveness, and the suitability of knowledge points and exercises within the teaching elements. The criteria for determining statistical significance are: using a p-value test, a relationship with p < 0.05 is considered statistically significant; simultaneously, combined with a frequency threshold, only relationships that occur ≥ 5 times cumulatively are retained for pattern extraction.

[0030] The aforementioned design constructs a complete analytical system encompassing "direct association, implicit association, temporal evolution, and pattern extraction." This system enables rapid identification of direct associations and uncovers deep implicit associations among teaching elements through topic models and association rule algorithms, filling a gap in existing technologies. Temporal evolution analysis reveals the dynamic trends of these associations, making the analysis results more in-depth and timely. Finally, the pattern extraction unit filters out high-value intrinsic teaching patterns, providing precise and comprehensive data support for teaching decisions and contributing to the optimization of teaching quality.

[0031] In existing technologies, the core problems faced by knowledge graph construction modules are: inaccurate entity extraction, limited entity relationships, and static graphs that cannot iterate synchronously with data updates and indexing optimizations, resulting in insufficient practicality and timeliness. Currently, it is difficult to comprehensively extract core teaching entities, entity relationship establishment lacks support from association analysis results, and static graphs cannot adapt to the dynamic growth of teaching archives. Therefore, this embodiment designs the knowledge graph construction module to include a teaching entity extraction unit, an entity relationship establishment unit, and a dynamic graph update unit. The teaching entity extraction unit is used to extract core information from multi-dimensional indexing information and multi-source heterogeneous teaching archive data to obtain teaching-related core entities, including teachers, students, courses, knowledge points, and teaching archives. Specifically, it directly extracts entity identifiers such as teacher ID, student ID, and course number from structured data, extracts entities such as topics and knowledge points from the indexing information of unstructured data, and uses named entity recognition algorithms to assist in the extraction of entities that are not explicitly indexed. The entity relationship establishment unit is used to establish and store entity relationships such as teacher-lecture-course, course-include-knowledge point, teaching file-association-course, and knowledge point-adaptation-exercise based on the direct association between teaching files and the implicit association between teaching elements. The knowledge graph dynamic update unit is used to automatically trigger the process of supplementing and extracting teaching entities and updating entity relationships after new teaching files have been indexed and analyzed, or after the indexing system has been optimized and adjusted. This synchronously updates the entity and relationship data in the knowledge graph, enabling dynamic updates of the knowledge graph. Specifically, a knowledge graph update is automatically triggered when the number of newly added teaching files in a single instance is ≥10, or when the cumulative number of newly added files in a single month is ≥50; and a synchronous update of the knowledge graph is completed within 24 hours after the indexing system is optimized and adjusted.

[0032] The advantages of the above design scheme are as follows: entity extraction combines indexing information and raw data to ensure that no core entities are missed and extraction is more accurate; entity relationship establishment is based on the results of association analysis, with rich relationship types that fit the actual teaching situation, enabling the graph to truly reflect the complex relationships between teaching elements; the dynamic update mechanism ensures the timeliness of the graph, and the graph can be automatically updated synchronously after new data or indexing optimization, avoiding the problem of static graphs becoming outdated and invalid. At the same time, the accuracy and completeness of the graph also provide strong support for the retrieval and push functions of the application service module.

[0033] In existing technologies, the main problems with application service modules are: low retrieval accuracy, lack of personalized push notifications, poor visualization analysis, and insufficient data security, failing to meet the diverse needs of different users. Currently, most systems rely on simple keyword searches, resulting in highly homogenized push content, monotonous visualization, and rudimentary permission management, posing a risk of data leakage. Therefore, this embodiment designs the application service module to include a knowledge graph retrieval unit, a personalized push notification unit, a visualization analysis unit, and a permission management unit. The knowledge graph retrieval unit supports retrieval methods based on index terms and entity relationships. It is used to traverse entity and relationship data in the knowledge graph according to the search conditions input by the user, accurately match and return the corresponding teaching files and related resources. The personalized push unit is used to push appropriate teaching resources to teachers and personalized learning resources to students based on the relationship between teaching entities in the knowledge graph and combined with user identity attributes and historical behavior data. The visualization analysis unit is used to display the distribution of teaching resources, the coverage of knowledge points, the relationships between teaching points, and the time-series evolution trend in the form of charts, intuitively presenting the inherent laws of teaching and providing visual support for teaching decisions; among them, the distribution of teaching resources is presented in pie charts (by subject / data type), the coverage of knowledge points is presented in heat maps (by grade / chapter), and the time-series evolution trend is presented in line charts (by month / semester). The access control unit is used to configure different access permissions for teachers, students, and administrators using a role-based hierarchical authorization mechanism, allowing only authorized users to access teaching files and analysis data within their corresponding permission scope.

[0034] The above design scheme achieves precision, personalization, and security in application services: compared with traditional keyword retrieval, the knowledge graph-based retrieval method is more accurate in matching and more comprehensive in results, and can directly link to relevant resources; personalized push combines user identity and behavioral data to meet the differentiated needs of teachers' teaching and students' learning, and improve resource utilization efficiency; visual analysis presents complex relationships and teaching patterns in intuitive charts, reducing the difficulty of decision-making; and the role-based hierarchical authorization mechanism achieves fine-grained permission control, effectively ensuring the security of teaching archive data and adapting to the access needs of different users.

[0035] Therefore, based on the structural design of the "Automatic Indexing and Multi-Dimensional Association Analysis System for Digital Teaching Archives," this embodiment can obtain a complete interactive process of "data acquisition - indexing - association analysis - map construction - application services," the main contents of which are as follows: S1. After the data acquisition module completes the format standardization, it sends a "data ready signal" to the intelligent indexing module and synchronously transmits the data to be processed in a unified format (the data format is JSON, which includes fields such as data ID, type, content, and acquisition time). S2. After receiving the data, the intelligent indexing module first calls the latest teaching domain-specific vocabulary through the vocabulary construction unit, and then the semantic analysis unit completes the preprocessing and feature extraction. The multimodal adaptation unit calls the corresponding adaptation rules according to the data type (text / video / audio), generates indexing information, and then pushes "indexing completion signal + indexing data" to the multidimensional association analysis module. After receiving the indexing data, the S3 multi-dimensional association analysis module sequentially initiates direct association identification, implicit association mining, and temporal evolution analysis. Finally, the pattern extraction unit outputs the association results and teaching patterns, which are synchronized to the knowledge graph construction module and application service module. S4. After updating entities and relationships based on the association results, the knowledge graph construction module sends a "graph update complete signal" to the application service module. The application service module responds to user requests for retrieval, push, etc., based on the updated graph.

[0036] Below, we provide two examples to further illustrate the practical needs of educational scenarios and help you understand this technical solution.

[0037] Scenario 1: Teacher's Retrieval and Adaptation of Teaching Resources

[0038] 1) Teachers can enter "High School Mathematics - Function Monotonicity - Classroom Exercises" (index term + entity relation combination search) into the knowledge graph search unit of the application service module; 2) The knowledge graph retrieval unit traverses the graph, matches the entity relationship of "course (high school mathematics) - includes - knowledge point (function monotonicity) - adapt - exercises (classroom exercises)", and associates the directly associated files of "teaching file - association - course". 3) The system returns matching lesson plan documents, classroom exercise test papers, teaching videos and other resources, and supplements and recommends teaching method cases related to "function monotonicity" through personalized push units (based on the "teaching method-knowledge point" association results of implicit association mining); 4) The access control unit verifies the teacher's identity and only grants access to resources corresponding to the grade level taught by that teacher.

[0039] Scenario 2: End-to-end processing of new teaching archives

[0040] 1) The administrator uploaded 15 high school physics experiment teaching videos (unstructured data) using the batch upload function of the data collection module. 2) The data format standardization unit converts video data into a unified JSON format through video keyframe extraction and conversion (extracting one keyframe every 5 seconds and identifying information such as experimental equipment and steps in the frame) and audio-to-text conversion (using the ASR algorithm with a recognition accuracy of ≥95%). 3) The intelligent indexing module calls up a special terminology for the teaching field (including terms such as "physics experiment", "mechanics experiment", and "experimental steps") to generate indexing information containing "topic (mechanics experiment), subject (high school physics), and teaching link (experimental operation)". 4) The multi-dimensional correlation analysis module identifies the direct correlation between these 15 videos and existing "high school physics-mechanics experiment" related files (index term overlap of 75%), explores the implicit correlation between "experiment steps-student grades", and the temporal evolution analysis module sorts out the correlation relationship according to the collection time (September 2024). 5) The knowledge graph construction module automatically supplements the entity relationships of "teaching archives (experimental videos) - association - course (high school physics)" and "course - contain - knowledge points (mechanics experimental principles)" to complete the graph update; 6) The application service module pushes newly added experimental video resources to high school physics teachers, and the visualization analysis unit displays a "Mechanics Experiment Resource Distribution" chart in the administrator backend.

[0041] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0042] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A system for automatic indexing and multi-dimensional correlation analysis of digital teaching archives, characterized in that, The system is implemented based on a multi-module collaborative architecture, wherein: The data acquisition module is used to collect multi-source heterogeneous teaching archive data and convert it into data to be processed in a unified format; The intelligent indexing module is used to perform semantic analysis and multimodal adaptation on the data to be processed based on a special vocabulary for the teaching field, and generate multi-dimensional indexing information. The multi-dimensional correlation analysis module is used to identify direct correlations between teaching archives based on multi-dimensional indexing information, mine implicit correlations between teaching elements and conduct time-series evolution analysis, reveal the inherent laws of teaching and provide data support for teaching decisions. The knowledge graph construction module is used to extract teaching entities, establish entity relationships based on the results of association analysis, and construct a dynamically updated knowledge graph. The application service module supports knowledge graph-based retrieval, personalized push notifications, and visual analysis, and allows for setting fine-grained access permissions.

2. The teaching archive digital automatic indexing and multi-dimensional correlation analysis system according to claim 1, characterized in that, The data acquisition module includes: a multi-source data access unit and a data format standardization unit; The multi-source data access unit is used to acquire multi-source heterogeneous teaching archive data through batch uploading, data collection via interface with the academic affairs system, and intelligent scanning of paper archives. The data format standardization unit is used to transform multi-source heterogeneous teaching archive data into data to be processed in a unified format through OCR recognition and conversion, video keyframe extraction and conversion, and audio-to-text conversion.

3. The teaching archive digital automatic indexing and multi-dimensional correlation analysis system according to claim 1, characterized in that, Multi-source heterogeneous teaching archive data includes structured and unstructured data; among them, structured data includes teaching plans, course information, student grades and teacher information, while unstructured data includes lesson plan documents, test papers, courseware, teaching reflections, teaching videos, classroom audio and experimental images.

4. The teaching archive digital automatic indexing and multi-dimensional correlation analysis system according to claim 1, characterized in that, The intelligent indexing module includes: a vocabulary construction unit, a semantic analysis unit, and a multimodal adaptation unit; The vocabulary building unit is used to expand upon the Chinese Archives Thesaurus as a base vocabulary, incorporating subject classifications, teaching procedures, and knowledge points to form a multi-dimensional, hierarchical vocabulary for teaching fields. The semantic analysis unit is used to preprocess and extract features from the data to be processed, and to perform semantic mapping and matching between the extracted features and a special vocabulary for the teaching field to complete semantic parsing. The multimodal adaptation unit is used to perform indexing adaptation on the feature data corresponding to the text, video, and audio data to be processed, and output multi-dimensional indexing information including topic, subject, and teaching link.

5. The teaching archive digital automatic indexing and multi-dimensional correlation analysis system according to claim 4, characterized in that, The vocabulary building unit is also used to establish a dynamic optimization mechanism for a vocabulary specifically for the teaching field. Based on the search hit rate and indexing adaptation effect data fed back by the application service module, the vocabulary and hierarchical relationships in the vocabulary specifically for the teaching field are updated regularly.

6. The teaching archive digital automatic indexing and multi-dimensional correlation analysis system according to claim 1, characterized in that, The multi-dimensional correlation analysis module includes: a direct correlation identification unit, a latent correlation mining unit, a time-series evolution analysis unit, and a pattern extraction unit; The direct association identification unit is used to compare the multi-dimensional indexing information corresponding to different teaching files and determine that teaching files with indexing term overlap exceeding a preset threshold are directly associated. The implicit association mining unit is used to extract potential topics corresponding to multi-dimensional indexing information through topic models, and combine association rule algorithms to mine the association relationships between different potential topics to obtain implicit associations between teaching elements. The temporal evolution analysis unit is used to extract time attribute information from multi-dimensional indexing information, sort out the corresponding relationships of teaching archives according to the time dimension, analyze the evolution trend of teaching content and teaching relationships in different time periods, and complete the temporal evolution analysis. The pattern extraction unit is used to summarize and integrate the results of analysis on direct relationships between teaching files, implicit relationships between teaching elements, and time-series evolution, to screen out statistically significant relationships, and to extract the inherent teaching patterns related to teaching methods and teaching effects, and the suitability of knowledge points and exercises among teaching elements.

7. The teaching archive digital automatic indexing and multi-dimensional correlation analysis system according to claim 1, characterized in that, The knowledge graph construction module includes: a teaching entity extraction unit, an entity relationship establishment unit, and a graph dynamic update unit; The teaching entity extraction unit is used to extract core information from multi-dimensional indexing information and multi-source heterogeneous teaching archive data to obtain teaching-related core entities, including teachers, students, courses, knowledge points, and teaching archives. The entity relationship establishment unit is used to establish and store entity relationships such as teacher-lecture-course, course-include-knowledge point, teaching file-association-course, and knowledge point-adaptation-exercise based on the direct association between teaching files and the implicit association between teaching elements. The knowledge graph dynamic update unit is used to automatically trigger the process of supplementing and extracting teaching entities and updating entity relationships after the indexing and association analysis of newly added teaching files are completed, or after the indexing system is optimized and adjusted. This synchronously updates the entity and relationship data in the knowledge graph, thereby realizing the dynamic update of the knowledge graph.

8. The teaching archive digital automatic indexing and multi-dimensional correlation analysis system according to claim 1, characterized in that, The application service module includes: a knowledge graph retrieval unit, a personalized push unit, a visualization analysis unit, and a permission management unit; The knowledge graph retrieval unit supports retrieval methods based on index terms and entity relationships. It is used to traverse entity and relationship data in the knowledge graph according to the search conditions input by the user, accurately match and return the corresponding teaching files and related resources. The personalized push unit is used to push appropriate teaching resources to teachers and personalized learning resources to students based on the relationship between teaching entities in the knowledge graph and combined with user identity attributes and historical behavior data. The visualization analysis unit is used to display the distribution of teaching resources, the coverage of knowledge points, the relationship between teaching and the evolution of time sequence in the form of charts, intuitively presenting the inherent laws of teaching and providing visual support for teaching decisions. The access control unit is used to configure different access permissions for teachers, students, and administrators using a role-based hierarchical authorization mechanism, allowing only authorized users to access teaching files and analysis data within their corresponding permission scope.

Images