Artificial Intelligence-Based Information Aggregation and Retrieval System and Method

Abstract

This invention belongs to the field of computer application technology. It discloses an information aggregation and retrieval system and method based on artificial intelligence, comprising: receiving user task objectives; combining associated resource sets and search preferences; calling a semantic model to generate semantic anchors; thereby generating a university-specific contextual blueprint to form a standardized information foundation within the university; parallel scheduling of various information interfaces within the university according to permissions; expanding query intents; filtering and sorting results to form a blueprint-bound retrieval result set; performing university-specific scenario analysis; generating dynamic aggregation summaries and difference comparison tables; grouping according to preset logic and performing visualization processing to form a structured presentation view; collecting user interaction data in the structured presentation view to generate a university user feedback dataset; and optimizing the university-specific contextual blueprint and university-specific knowledge unit network to form a closed-loop link.

Description

Technical Field

[0001] This invention relates to the field of computer application technology, and more specifically, to an information aggregation and retrieval system and method based on artificial intelligence. Background Technology

[0002] With the deepening of the digital transformation of higher education, information resources related to teaching, research, and administration in universities are experiencing explosive growth and are often isolated. Users, including faculty, students, and administrative staff, have an increasingly urgent need for precise, scenario-based information aggregation and retrieval services. While traditional retrieval systems can achieve basic information queries, they have significant shortcomings in adaptability to the specific scenarios of universities, making it difficult to meet users' core needs.

[0003] Existing information retrieval solutions for universities still have some shortcomings: low retrieval accuracy, traditional keyword retrieval mode relies on simple metadata matching, cannot deeply understand university-specific terminology (such as course chapters, research project numbers) and user task intent, lacks deep knowledge-level connections, and generally adopts static retrieval strategies and resource organization methods, which cannot be dynamically adjusted and iterated based on user interaction feedback. In the long run, service capabilities are difficult to continuously improve and cannot adapt to the ever-changing information needs in the university setting. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art and to achieve the above objectives, the present invention provides the following technical solution: an information aggregation and retrieval system based on artificial intelligence, comprising: Blueprint Standard Module: Receives user task objectives, combines related resource sets and search preferences, calls the semantic model to generate semantic anchors, and then generates a university-specific contextual blueprint with permission tags; simultaneously integrates multi-source information within the university to form a standardized information foundation within the university; Search Binding Module: Based on the standardized information foundation within the university, through the university-specific context blueprint, various information interfaces within the university are scheduled in parallel according to permissions. The semantic anchor points are used to expand the query intent. The results are filtered and sorted in combination with the search strategy. The system is linked to the university-specific knowledge unit network to form a blueprint-bound search result set. Analysis View Module: Performs university-specific analysis on the blueprint-bound search result set, generates dynamic aggregated summaries and difference comparison tables, groups them according to preset logic and performs visualization processing to form a structured presentation view; Interactive Feedback Module: Collects user interaction data in the structured presentation view, and associates it with the university-specific context blueprint and on-campus resources to generate a university user feedback dataset; Optimize the closed-loop module: Based on the university user feedback dataset, optimize the university-specific scenario blueprint and university-specific knowledge unit network to form a closed-loop link.

[0005] Furthermore, the generation method of the university-specific scenario blueprint includes: Receive user task objectives, combine related resource sets and search preferences, connect to the unified identity and permission authentication within the school, and generate exclusive permission tags; The system calls a pre-built semantic vector model to extract the core text of the task objective and associated resource set to generate a set of semantic anchor points. Simultaneously, it converts search preferences into executable search strategy parameters, adds scene tags, and integrates them to generate a structured, university-specific contextual blueprint.

[0006] Furthermore, the formation methods of the aforementioned standardized information infrastructure within the school include: Collect and temporarily store multi-source information from various distributed information systems within the campus according to a pre-defined hierarchical access policy; The core metadata of multi-source information is extracted by three-level metadata specifications, multi-dimensional data cleaning is performed to form unique and valid data, and then classified according to scenario and resource type and multi-dimensional structured index is built to form a standardized information foundation within the school.

[0007] Furthermore, the method of parallel scheduling of various information interfaces within the university according to permissions through a university-specific scenario blueprint includes: After a user initiates a search, the target university’s exclusive context blueprint is activated, and the scope of the search data is defined by combining the search strategy parameters with the university’s standardized information foundation. Then, based on the permission tags, a pre-set interface and permission mapping table are matched to filter out the schedulable internal information interfaces and form an interface list; Based on the interface list, the retrieval strategy parameters are refined into retrieval instructions, parallel scheduling parameters are configured, and the retrieval engine is initialized.

[0008] Furthermore, the blueprint binding retrieval result set is formed in the following ways: Based on the interface list and refined search instructions, the search is initiated by multi-threaded parallel scheduling of various information interfaces within the school. After the retrieval is completed, the semantic anchor set is expanded by combining the preset query expansion rules to generate accurate query terms, and a second matching filter is performed on the original retrieval results; The search results after secondary matching are subjected to multi-dimensional preliminary filtering according to the search strategy parameters. Then, the comprehensive score of each resource in the preliminary filtering search results is calculated and sorted. The resource-knowledge unit network of universities is associated to form the association pair between resources and knowledge units. After sorting the associated pairs of different interfaces, perform fusion and deduplication processing to generate a structured blueprint binding retrieval result set that binds blueprint IDs and permission identifiers.

[0009] Furthermore, the generation methods for the dynamic aggregated summary and difference comparison table include: Link the blueprint to the search result set and the corresponding university-specific scenario blueprint, filter effective resources and mark scenario analysis tags through multi-dimensional analysis rules for the corresponding scenario; The system analyzes tags and knowledge units based on scenarios to extract common information and generate dynamic aggregated summaries. It also identifies differences in annotations of similar resources and generates a difference comparison table. By integrating effective resources, dynamically aggregating summaries and difference comparison tables, a scenario-based analysis result package is formed.

[0010] Furthermore, the structured presentation view is formed in the following ways: The results of the contextual analysis are linked to the corresponding university-specific scenario blueprints. The logical grouping rules of the pre-set scenarios are matched to determine the dimension combination of this grouping. Then, the hierarchical logical grouping is executed first according to the main dimension to form a structured grouping and resource mapping relationship. By combining scene tags and grouping results, multi-view content is filled with corresponding scene-specific view templates to generate multi-type structured presentation views.

[0011] Furthermore, the generation method of the university user feedback dataset includes: Real-time data collection of user interaction in the structured presentation view; association of university-specific contextual blueprints, campus resources, and users according to four-dimensional association rules; integration of all associated data; generation of a structured university user feedback dataset.

[0012] Furthermore, the methods for optimizing the university-specific contextual blueprint and the university-specific knowledge unit network include: Based on a dataset of feedback from university users, we can explore the core needs and behavioral preferences of different user types in different scenarios within universities. Furthermore, by optimizing the rules based on university scenarios, the university-specific scenario blueprint and university-specific knowledge unit network are optimized and updated respectively.

[0013] Furthermore, the information aggregation and retrieval method based on artificial intelligence is characterized by including: S1: Receive user task objectives, combine associated resource sets and search preferences, call the semantic model to generate semantic anchors, and then generate a university-specific contextual blueprint with permission tags; simultaneously integrate multi-source information within the university to form a standardized information foundation within the university. S2: Based on the standardized information foundation within the university, through the university-specific context blueprint, various information interfaces within the university are scheduled in parallel according to permissions. The query intent is expanded with semantic anchors, and the results are filtered and sorted in combination with retrieval strategies. The university-specific knowledge unit network is associated to form a blueprint-bound retrieval result set. S3: Perform university-specific analysis on the blueprint-bound search result set, generate dynamic aggregated summaries and difference comparison tables, group them according to preset logic and perform visualization processing to form a structured presentation view; S4: Collect user interaction data in the structured presentation view, and associate it with the university-specific context blueprint and on-campus resources to generate a university user feedback dataset; S5: Based on the university user feedback dataset, optimize the university-specific scenario blueprint and university-specific knowledge unit network to form a closed-loop link.

[0014] The technical effects and advantages of the information aggregation and retrieval system and method based on artificial intelligence of this invention are as follows: This invention focuses on three core scenarios in higher education: teaching, research, and administration. It constructs an intelligent information aggregation and retrieval system that integrates user task input, retrieval result presentation, and system optimization. This system accurately addresses the core pain points of traditional retrieval systems and achieves an organic unity of precise retrieval, scenario-based services, and closed-loop system iteration. First, by accurately representing the user's task intent with the standardized integration of multi-source information within the university, combined with the extended query of semantic anchors in the intelligent retrieval and binding module, the multi-dimensional dynamic sorting mechanism, and the deep association of the university's exclusive knowledge unit network, the core problem of low accuracy in traditional retrieval is solved, and the accuracy, scenario adaptability, and systematic nature of the retrieval results are greatly improved. Secondly, by collecting user interaction behavior throughout the entire process, verifying compliance, and constructing a structured feedback dataset, combined with resource optimization and closed-loop iteration modules, the system dynamically optimizes the university-specific scenario blueprint and knowledge unit network based on the user feedback dataset, thus constructing a complete closed-loop iteration link of user feedback-resource optimization-precision service. This solves the core defects of traditional systems that are static and unable to be iteratively upgraded, and achieves continuous improvement in the system's service capabilities. Meanwhile, by performing scenario-based filtering of search results, dynamically aggregating and generating summaries, and presenting multiple types of visual views, this invention significantly reduces the time cost for users to obtain effective information and improves information utilization efficiency; combined with permission-based hierarchical and privacy-compliant design, it ensures secure access to university information resources and protection of user privacy. This invention, by constructing an AI-driven, end-to-end intelligent architecture and a dynamic closed-loop optimization mechanism, achieves precise aggregation, efficient utilization, and continuous optimization of university information resources throughout the entire process from data collection to service presentation. It is applicable to various scenarios such as university teaching, research, and administration, and improves the accuracy, efficiency, and user experience of information retrieval. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the information aggregation and retrieval system based on artificial intelligence according to the present invention; Figure 2 This is a schematic diagram illustrating the process of generating semantic anchor set in the artificial intelligence-based information aggregation and retrieval system of the present invention; Figure 3This is a schematic diagram of the information aggregation and retrieval method based on artificial intelligence according to the present invention. Detailed Implementation

[0016] 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.

[0017] Example 1, please refer to Figure 1 and Figure 2 As shown, the information aggregation and retrieval system based on artificial intelligence described in this embodiment includes: Blueprint Standard Module: Receives user task objectives, combines related resource sets and search preferences, calls the semantic model to generate semantic anchors, and then generates a university-specific contextual blueprint with permission tags; simultaneously integrates multi-source information within the university to form a standardized information foundation within the university; Search Binding Module: Based on the standardized information foundation within the university, through the university-specific context blueprint, various information interfaces within the university are scheduled in parallel according to permissions. The semantic anchor points are used to expand the query intent. The results are filtered and sorted in combination with the search strategy. The system is linked to the university-specific knowledge unit network to form a blueprint-bound search result set. Analysis View Module: Performs university-specific analysis on the blueprint-bound search result set, generates dynamic aggregated summaries and difference comparison tables, groups them according to preset logic and performs visualization processing to form a structured presentation view; Interactive Feedback Module: Collects user interaction data in the structured presentation view, and associates it with the university-specific context blueprint and on-campus resources to generate a university user feedback dataset; Optimize the closed-loop module: Based on the university user feedback dataset, optimize the university-specific scenario blueprint and university-specific knowledge unit network to form a closed-loop link.

[0018] The methods for generating university-specific scenario blueprints include: After logging in using the unified campus identity authentication, users enter the blueprint creation section. The visual interface displays a pop-up window by default, offering three creation paths to suit different needs and search requirements. Specifically: Mode 1: Quick Template Creation: Users can select a corresponding preset template (such as "Teaching - Courseware Integration", "Scientific Research - Experimental Data Retrieval", "Administration - Service Process Inquiry") by entering keywords (such as courseware integration, literature research, etc.) and loading the basic search preferences built into the template; users only need to fill in the task objective (such as "Integration of courseware for the 2025 Computer Science Major 'Machine Learning' course") to enter the subsequent related resource supplementation stage, without having to manually set complex parameters; Mode 2: Custom Creation: Users select a custom blank template, input detailed task objectives through a visual form, then select task scenario tags (such as machine learning, course construction, etc.), and manually set search preferences (time range, resource type priority, format requirements, permission range (such as public only, department-specific resources, classified resources, etc., classified options require subsequent permission verification)). After completing the tag and search preference settings, the user enters the associated resource supplementation stage. Mode 3: Reuse existing blueprints (for recurring tasks): Users enter their personal blueprint library and quickly find all created blueprints by searching by keywords and filtering by scenario category; they can choose "Direct Reuse" (retain original parameters) or "Quick Adjustment" (only modify search preferences and supplement related resources, without re-entering task objectives), and then proceed to the next step; Resource Supplementation: Pre-set resource supplementation methods are provided in advance. Users can choose one or a combination of them. All resources are associated with a unique ID within the school (ID refers to a unique identifier), so there is no need to upload repeatedly. Among them, related resources refer to core resources that fit the three scenarios of teaching, scientific research and administration in universities, such as seed literature, course outlines, project proposals, draft experimental records, courseware, exercises, and service guides. The methods for supplementing associated resources include local file upload (applicable to initial resources stored locally), direct selection from the campus resource library (applicable to archived resources within the campus), and inheritance of historical resources (applicable to resources that users have previously collected or associated). After receiving the user's input of task objectives, associated resource sets, and search preferences, the system extracts the user's identity identifier (employee ID, student ID, etc.) during login, calls the unified identity authentication interface within the university, verifies the user's identity (student, teacher, researcher, or administrative staff) based on their employee ID or student ID, matches the permission level, completes permission authentication, and generates a unique permission tag (e.g., student-teaching public resource access permission); if the verification fails (e.g., no access permission for classified resources), the system restricts the resource search scope and provides a real-time prompt. Example of identity and permission level matching: Student - Access to public teaching resources, the corresponding scope of accessible resources includes public courseware in the academic affairs system, open literature in the library, public administration guides, etc.; Teacher - Full access to teaching resources + basic research permissions, the corresponding scope of accessible resources includes resources accessible to students + course-specific resources, departmental research shared resources, etc. After the permission verification is passed, two types of core text are extracted, including the task target text (complete extraction) and the core text of the associated resource set (extracting abstracts and keywords from literature, extracting titles and chapter names from courseware, extracting core knowledge points from course outlines, and extracting project terms and experimental objectives from experimental data). The extracted core text is input into a pre-integrated semantic vector model, which outputs a semantic vector with a dimension of 768. For multi-source text (such as the core text of multiple related resources), the average of the vectors is taken as the comprehensive semantic vector to ensure that all core information is covered. It should be noted that this semantic vector model is based on the BERT-base pre-trained model architecture, with lightweight fine-tuning for university scenarios. Its core function is to transform the target text of the task and the core text of the associated resource set into accurate semantic vectors. The main body of the model retains the 12-layer Transformer encoder, 768-dimensional output vector dimension, and 12 attention heads of BERT-base, and only fine-tunes the word embedding layer and output layer. It incorporates a university-specific terminology dictionary (containing 2000+ teaching terms such as "course outline" and "knowledge points", scientific research terms such as "experimental parameters" and "project proposal" and administrative terms such as "procedures"). The training dataset uses a self-built text dataset from universities, containing 50,000 texts (50-200 characters each) from three scenarios: teaching, research, and administration. The dataset was annotated by three university experts according to the dual dimensions of "task intent - core terminology". The data preprocessing steps include text cleaning (removing special symbols and redundant spaces), word segmentation (using jieba word segmentation combined with a university terminology dictionary), and normalization (the text length is uniformly 256 tokens, with zeros added if insufficient and truncation if too long). Training was performed using the PyTorch 2.0 framework, with Adam as the optimizer. The initial learning rate was 0.0001 (decreasing by 0.1 every 10 epochs), the batch size was 32, and the training epochs were 50. Training was stopped if the validation set loss did not decrease for 5 consecutive epochs. The cross-entropy loss function was used, combined with semantic similarity constraints to improve the accuracy of university terminology representation. This design specifically addresses the problem of inaccurate representation of university-specific terms by general semantic models, ensuring that the generated semantic vectors can accurately match the university's task intent and provide a reliable foundation for subsequent semantic anchor selection. The obtained comprehensive semantic vectors are clustered based on similarity, and core terms with a similarity of ≥0.85 are selected as candidate anchors. Combined with task scenario labels, terms irrelevant to the scenario are removed from the candidate anchors (e.g., removing research terms from teaching scenarios). Finally, the 3-5 semantic anchors with the highest similarity are retained to form a semantic anchor set, in the format of core term - semantic similarity. The generated semantic anchor set is displayed through a visual interface, and users can manually delete irrelevant anchors or add missing anchors (e.g., adding "supervised learning") to improve anchor accuracy. The search preferences are simultaneously converted into structured executable parameters, forming search strategy parameters, for example: Time range: "Past 3 years" is converted to a timestamp range (current timestamp - 3×365×24×3600 to the current timestamp); Resource type weights: "Courseware > Exercises > Literature" are converted to structured parameters: {"Courseware": 0.5, "Exercises": 0.3, "Literature": 0.2}; The task objectives, semantic anchor set, transformed retrieval strategy parameters, permission tags, associated resource set and user identity identifier are integrated, and scenario tags (teaching, scientific research, administration) are added to generate a structured, university-specific scenario blueprint and generate a unique blueprint ID in the format of user identity identifier-time stamp-scenario type code. It should be noted that blueprints created quickly using templates are marked as temporary blueprints by default, but can be upgraded to official blueprints through the upgrade mechanism; The upgrade mechanism for a temporary blueprint is defined as follows: it is called ≥3 times within 7 days, or the user favorites the blueprint or the user favorites ≥5 resources in the search results for the blueprint; Custom-created and reused blueprints are marked as official blueprints by default, supporting full editing, sharing, and archiving operations.

[0019] The formation methods of standardized information infrastructure within the school include: It connects to all distributed information systems within the university, with two types of interfaces pre-configured: a real-time scheduling interface, adapted to dynamically updated information systems such as the academic affairs management system and the university's collaboration platform; and a batch synchronization interface, adapted to statically updated information systems such as the library resource database and the administrative service guide database. At the same time, it configures hierarchical access policies for each interface and links with the university's unified identity authentication system to ensure that only data within the corresponding permission scope can be collected. After a user logs in through the unified campus identity authentication, the campus-specific contextual blueprint is constructed simultaneously, triggering a multi-source information collection process to form a standardized information foundation within the campus. Specifically: According to the pre-defined hierarchical access policy of the interface, collect multi-source information (including internal data such as academic affairs, scientific research, and departmental resources) from various distributed information systems within the school, and classify and temporarily store it in a temporary database according to scenario type, resource type and interface type. The temporarily stored multi-source information is parsed, and the required metadata fields of each data resource are extracted according to the three-level metadata specification as core metadata. For optional metadata fields, they are extracted as completely as possible. For metadata that cannot be automatically extracted (such as experimental parameters of experimental data), they are marked as to be supplemented, and a reminder notification is sent to the uploader of the data resource to ensure the integrity of the metadata. The three-level metadata specification is categorized according to the structure of first-level scenario - second-level resource type - third-level core fields, as exemplified below: Level 1 Scenario: Teaching - Level 2 Resource Type: Courseware - Level 3 Core Fields: Unique Resource ID, Course Name, Applicable Grade, Uploader, Permission Level - Optional Metadata Fields: Upload Time, Update Log, Download Count; Level 1 Scenario: Scientific Research - Level 2 Resource Type: Experimental Data - Level 3 Core Fields: Unique Resource ID, Project Number, Experimental Parameters, Data Format, Permission Level - Optional Metadata Fields: Experiment Time, Data Source, Person in Charge; Level 1 Scenario: Administration - Level 2 Resource Type: Service Guide - Level 3 Core Fields: Unique Resource ID, Process Name, Required Materials, Processing Time Limit, Permission Level - Optional Metadata Fields: Processing Location, Contact Number, Update Time; The extracted core metadata and raw data are standardized in a unified format, including entity normalization (converting entity information such as course name, department name, and project number into a unified format to eliminate differences caused by synonyms, abbreviations, etc.), format unification (converting time format to ISO8601 format, text encoding to UTF-8, and file format to a supported format), and length standardization (unifying the length of core metadata fields to a preset maximum value). Perform multi-dimensional data cleaning on the standardized core metadata and raw data, including validity verification (marking and removing invalid data with missing required metadata, mismatched permission levels, and data in unsupported formats) and double deduplication. The dual deduplication process is as follows: if two data resources have the same unique ID, the latest updated one is retained; if two data resources have different unique IDs but the same content hash value (i.e. the content is completely consistent), the data resource with more complete metadata is retained, the unretained data resources are removed, and the retained data resources are marked as unique valid data. The unique, valid data after deduplication is categorized according to the scenario, resource type, and core fields of the three-level metadata specification, and a multi-dimensional structured index is constructed, including a scenario index (categorized by teaching, research, and administrative scenarios), a resource type index (categorized by resource types such as courseware, literature, and experimental data), a permission level index (categorized by permission levels such as public resources, department-specific resources, and research shared resources), and a core term index (based on the core metadata fields of resources, core terms are extracted to construct the index). This data is then stored in the university's standardized information database to form the university's standardized information foundation.

[0020] The methods for parallel scheduling of various information interfaces within the university according to permissions, using a university-specific scenario blueprint, include: After a user selects or constructs a specific scenario blueprint for a university and initiates a search, the specific scenario blueprint for the target university is activated, core elements are extracted and loaded into the search engine's memory, including a set of semantic anchors, search strategy parameters, permission tags, and a list of associated resource IDs for the associated resource set. Then, by combining the loaded search strategy parameters with the standardized information foundation within the school, the scope of search data is defined, including the time range (limiting the time range of data in the standardized information foundation within the school according to the timestamp interval converted in the blueprint), the resource type range (prioritizing the search of high-weight resource types according to resource type weight), and the format range (limiting the search to only resources that meet the requirements according to format filtering rules). Furthermore, based on the permission tags in the target university's exclusive scenario blueprint, the system matches the pre-set interface and permission mapping table, filters out the schedulable internal information interfaces according to the scenario (such as filtering the academic affairs system interface in the teaching scenario and filtering the scientific research project platform interface in the scientific research scenario), excludes unusable interfaces, and thus integrates and generates a filtered interface list, marking the priority of each interface (high-weight resource types correspond to high interface priority). The interface-permission mapping table is a pre-defined three-dimensional mapping table of permission tags, internal information interfaces, and accessible resource ranges. This table clarifies the schedulable interfaces and data access boundaries corresponding to different permission levels. An example is: Access Control Tags: Student-Teaching Public Resource Access Permissions; Scheduled Internal Information Interfaces: Academic Affairs System Public Resource Interface, Library Open Document Interface, Public Administration Guide Interface; Accessible Resource Scope: Open Courseware, Open Journal Articles, Public Administrative Procedures; Permission Tag: Researcher - Project-Specific Permissions; Accessible Internal Information Interfaces: Research Project Management Platform Interface, Experimental Equipment Data Interface, Departmental Research Shared Resource Interface; Accessible Resource Scope: Corresponding Research Project Data, Experimental Records, Departmental Research Literature; Based on the interface list, the loaded search strategy parameters are further refined into search instructions that the interfaces can recognize: for time range parameters, they are converted into timestamp query conditions that are uniformly supported by each interface; for resource type parameters, the interface search priority is generated according to weight (high-weight resource type interfaces are scheduled first); for format parameters, they are converted into file format filtering fields of the interfaces. Call the pre-configured retrieval strategy parameter template and configure the parallel scheduling parameters: For thread allocation, classify threads according to the number of filtered interfaces, and allocate an independent thread to the interface with the highest priority; For timeout and retry configuration, set a timeout threshold (e.g., 5 seconds) for each interface and a 3-retry mechanism (retry interval of 2 seconds); For caching configuration, enable temporary caching of retrieval results to avoid receiving duplicate data repeatedly. The pre-configuration logic of the retrieval strategy parameter template is as follows: the retrieval strategy parameter template is pre-configured according to teaching, scientific research and administrative scenarios, including basic parameters such as the number of parallel scheduling threads (default 8 threads, which can be dynamically adjusted according to the number of interfaces), interface timeout threshold (default 5 seconds), number of retries (default 3 times), and data receiving buffer size (default 100MB). Finally, the search engine is started, and the interface list, refined search instructions, and parallel scheduling parameters are loaded to form the initialization process of the search engine, thus completing the full preparation before the search and waiting for the subsequent parallel search process to be triggered.

[0021] The blueprint binding search result set is formed in the following ways: Based on the interface list and refined search instructions, each interface is assigned a thread with corresponding weight (3 threads for the highest priority interface (courseware), 2 threads for the medium priority interface (exercise), and 1 thread for the ordinary priority interface (document). The search engine uses parallel scheduling parameters to schedule the various information interfaces within the school to send search requests simultaneously in parallel, triggering parallel retrieval; thus, it receives the original search results returned by each interface and temporarily stores them in a temporary cache. After the search is completed, the semantic anchor set is expanded by combining the pre-set university-specific query expansion rules to generate accurate query terms; Among them, the pre-built query expansion rules are: a pre-built semantic anchor-university terminology expansion dictionary, including exclusive expansion rules for three types of scenarios: teaching (e.g., "courseware" expanded to "lecture PPT, case analysis, exercise set"), scientific research (e.g., "experimental data" expanded to "experimental original records, data analysis reports, parameter configuration tables"), and administration (e.g., "service process" expanded to "application materials, approval nodes, processing time limit"). Example of precise query term expansion generation: If the semantic anchor set is "Machine Learning - Course Materials - Computer Science - 2025 Class", then it is expanded to: Core query terms: Machine Learning, Course Materials; Scenario expanded terms: Computer Science 2025 Class, Lecture PPT, Case Analysis; Excluded terms: Non-Computer Science Major, Old Course Materials Before 2025 Class; Finally, a precise query term set of core terms + expanded terms + excluded terms is formed; Using the generated set of precise query terms, a secondary matching filter is performed on the temporarily stored original search results. Resources whose titles, summaries, and core metadata contain at least one core term and one extended term, and which do not contain any excluded terms, are retained. Irrelevant resources (such as machine learning-related resources for non-computer science majors) are removed to improve search accuracy. The search results after secondary matching are subjected to multi-dimensional preliminary filtering based on the search strategy parameters, including time filtering (removing resources that exceed the time range set by the blueprint), format filtering (retaining resources that meet the format requirements), permission filtering (re-verifying the permission level of the resources to ensure that they match the user's permission tags), and validity filtering (removing resources with invalid addresses or incomplete content). After filtering is completed, the comprehensive score of each resource in the preliminary filtered search results is calculated according to the resource type weight in the search strategy parameters. The formula is: Comprehensive score = Basic score × Time coefficient × Format coefficient; The base score is calculated as: base score = resource type weight × semantic similarity between resource and anchor point; the time coefficient is matched based on the resource update time (default setting: 1.0 for the timeliness of resources in the past year, 0.8 for the past two years, and so on); the format coefficient is preset based on the resource format (e.g., PPT format courseware coefficient = 1.0, DOCX format courseware coefficient = 0.8). Sort resources by overall score from highest to lowest; if resources have the same score, sort them by resource update time in descending order; if resources of the same type have the same overall score, sort them by semantic similarity in descending order. Based on the prioritized resources, a pre-built network of knowledge units specific to universities is invoked. Semantic matching is performed between the core metadata of the resources and the core terms of the knowledge units to associate the corresponding knowledge units. Attributes such as knowledge unit ID, association basis, and knowledge unit confidence are added to each resource to form resource-knowledge unit association pairs. Among them, the university-specific knowledge unit network is a pre-constructed structured knowledge unit network covering three types of scenarios: teaching, scientific research, and administration. Each knowledge unit contains attributes such as knowledge unit ID, core terms, associated resource types, scenario tags, and confidence levels; and pre-set association rules between knowledge units and internal resources (such as associating the "supervised learning" knowledge unit with machine learning courseware and related scientific research literature). After sorting the associated pairs returned by different interfaces, perform fusion and deduplication processing: double deduplication is performed by resource ID + content hash value (only the one with the highest comprehensive score is retained if the resource ID is the same or the content is completely identical); multi-source information of the same resource is merged (e.g., if a courseware comes from both the academic affairs system and the department resource library, the metadata of both is merged to complete the information); the blueprint association criteria for each resource are marked (e.g., "matching semantic anchor 'machine learning - courseware'" "meets the 'resources of the past 3 years' preference", etc.). After fusion and deduplication, all resources are sorted by comprehensive score and categorized by knowledge unit. Corresponding blueprint IDs, user permission tags, and search timestamps are added as identifiers to generate a structured blueprint-bound search result set, which is stored in the search result database and simultaneously pushed to the subsequent scenario-based analysis stage.

[0022] The methods for generating dynamic aggregated summaries and difference comparison tables include: The system associates the blueprint with the search results set and the corresponding university-specific scenario blueprint. Based on the scenario tags of the blueprint, it calls the multi-dimensional analysis rules of the corresponding scenario from the pre-set scenario-based analysis rule library. According to the called analysis rules, it filters the effective resources in the blueprint-bound search results set and marks them with scenario analysis tags. Among them, the scenario-based analysis rule base is an analysis rule base constructed according to three scenarios: teaching, scientific research and administration, which maps scenario tags, analysis dimensions, filtering rules and related knowledge units. An exemplary scenario-based analysis rule base (scenario type - analysis dimension - filtering rules - related knowledge unit requirements): Teaching - Course adaptability, knowledge point matching - Resources applicable to grade or major and blueprint Figure 1 The core knowledge points cover ≥3 key nodes of the course outline, the resource format meets the teaching and presentation requirements - related to knowledge units of "course chapters and knowledge points", with a confidence level of ≥0.8; Research - Project Relevance and Data Completeness - Resource project number matches research project associated with blueprint; experimental data includes complete parameters, results, and analysis conclusions; literature publication time is within the project cycle - associated with knowledge units of "research direction, experimental method, and core indicators"; Administration - Process matching and material completeness - Resources correspond to the service process and blueprint task objectives, including a complete list of application materials and approval node descriptions. The information is the latest version (updated within the last year) - associated with knowledge units of "service process, approval node, and material type"; The method for filtering valid resources is as follows: Teaching scenario: Extract teaching attribute information such as applicable grade and major from the resource metadata, and verify its consistency with the teaching scenario attributes (including applicable grade, major, curriculum system, etc.) marked by the blueprint, and eliminate mismatched resources; Based on the core knowledge point system of the corresponding course in the university-specific knowledge unit network, compare the knowledge point coverage of the core content of the resource, and retain resources with a coverage that reaches a preset threshold (such as ≥3 key nodes). Research scenario: Remove research resources whose project identification information is not related to or does not match the project identification bound to the blueprint; verify the completeness of resource data fields and retain only resources whose completeness meets the preset standard (e.g., ≥80%). In administrative scenarios: Remove administrative resources whose names correspond to the service process and the blueprint task objectives are inconsistent; verify whether the resources contain service processes, application materials and approval nodes, and retain those with ≤1 missing items; For the selected valid resources, tag them with scenario analysis labels (such as "teaching-machine learning-supervised learning") to provide a basis for subsequent grouping; for the resources that are removed from the selection, record the reasons for removal (such as "project number mismatch") to facilitate user traceability. Analyze tags and associated knowledge units according to scenarios, and structure effective resources into groups (e.g., the "Teaching-Machine Learning-Supervised Learning" group contains all courseware and exercises that match the tag). For each set of resources, extract the unified and compatible scenarios and the core knowledge points they cover as common information, and extract differences in resource types and content focus as individual characteristics. Integrate the extracted common and individual characteristics to generate a dynamic aggregated summary (Example (teaching scenario): "This set of resources is adapted for the supervised learning chapter of 'Machine Learning' for the 2025 Computer Science major, including 3 PPT courseware and 2 exercise sets, covering the three core knowledge points of logistic regression model construction, parameter tuning, and model evaluation. Courseware 1 focuses on practical cases, while courseware 2 focuses on theoretical derivation. They can be used together to improve teaching effectiveness"). Simultaneously, it identifies similar resources within the same group (such as courseware from the supervised learning chapter of "Machine Learning" or experimental data from the same research project). Based on the identified resource type, it extracts the comparison dimension information of similar resources through preset comparison dimensions, marks the differences, and generates a difference comparison table. An example difference comparison table (experimental data comparison): [Resource Name; Experimental Parameters; Sample Size and Detection Method; Result Error; Validity Level]; [Experimental data -202506; learning rate 0.01, 500 iterations; 1000; gradient descent; 2.3%; excellent] [Experimental data -202507; learning rate 0.02, 500 iterations; 1200; gradient descent method; 1.8%; excellent] The preset comparison dimensions are for different resource types. Examples include: comparison dimensions for courseware: knowledge point coverage, applicable semester, update time, and format; comparison dimensions for experimental data: experimental parameters, sample size, detection method, and result error; and comparison dimensions for administrative services: processing time limit, application materials, approval level, and processing location. By integrating effective resources, dynamically aggregating summaries and difference comparison tables, a scenario-based analysis result package is formed.

[0023] The methods for creating structured view representations include: The contextual analysis results package is linked to the corresponding university-specific scenario blueprint. Based on the scenario tags, the full set of grouping rules for the corresponding scenario (including main dimensions, sub-dimensions, and priorities) is called from the preset logical grouping rules. At the same time, the specific dimension combination for this grouping is determined by combining the difference comparison table, scenario analysis tags, and related knowledge units in the contextual analysis results package, ensuring that the grouping logic and task intent are accurately matched. Among them, the pre-defined logical grouping rules are hierarchical rules of grouping main dimension - grouping sub-dimension - grouping priority constructed according to the scenario, which are used to clarify the core grouping logic of different scenarios; An example of a default logical grouping rule: Teaching - Related Knowledge Units (Course Chapters, Knowledge Points) - Resource Type, Adapted Teaching Stage, Update Time - Knowledge Unit > Resource Type > Teaching Stage > Update Time; Scientific research - related knowledge units (research direction, experimental process) - resource type, project cycle stage, data integrity level - knowledge unit > project cycle > resource type > integrity level; Administration - Related Knowledge Units (Service Process, Approval Nodes) - Resource Type, Update Time, Processing Channel - Knowledge Unit > Service Process > Resource Type > Update Time; Based on the priority of logical grouping rules, the grouping dimensions are parsed and sorted (e.g., teaching scenarios are sorted by knowledge unit > resource type > teaching stage) to ensure that high-priority dimensions dominate the grouping structure and low-priority dimensions assist in refining the grouping. Based on the parsed grouping dimensions, the effective resources in the scenario-based analysis results package are hierarchically grouped: first-level grouping is completed according to the highest priority main dimension, generating first-level group labels; for each first-level group, second-level grouping is completed according to the next lower priority sub-dimensions, generating second-level group labels; the grouping results are validated. If a group has too few resources (e.g., ≤2) and its core attributes are consistent with those of adjacent groups, it is merged into the same group; if a group has too many resources (e.g., ≥50), it is further split according to lower priority dimensions to ensure that the group size is adapted to users' reading and search needs. Add identification information such as group tags, resource quantity, dynamic aggregation summary and knowledge unit to each group to form a structured mapping relationship between groups and resources, providing data structure support for subsequent view generation; By combining scene tags and grouping results, the system matches and calls the corresponding view template combination (such as calling "scene overview view + knowledge unit grouping view + teaching course adaptation view + similar difference comparison view" in the teaching scenario) through the preset corresponding scene-specific view template combination, automatically fills the multi-view content, generates multi-type structured presentation views, and supports users to manually switch or add view types to meet personalized presentation needs. Among them, the preset view template combination is a set of differentiated view templates customized according to the scenario, which is used to cover the full-dimensional presentation needs of overview, details and comparison, and to clarify the core display elements and format specifications of each view; An example of a multi-type structured presentation view template: [View type; suitable scenarios; core display elements; format specifications]; [Scene overview view; full scene; blueprint task objectives, number of groups, total resources, core summary, high-value resource recommendations; card layout]; [Knowledge Unit Grouped View; Full Scene; Group Tags, Resource List within Groups, Dynamic Aggregated Summary, Resource Access Entry; Hierarchical Collapsible Layout]; [Comparison view of similar differences; full-scenario comparison table, visualization charts, applicable conclusions, and advantageous resource indicators; table + chart layout]; [Teaching course adaptation view; teaching scenarios; course chapter and resource matching relationship, knowledge point coverage map, teaching usage suggestions; map + list combination layout]; [Research project lifecycle view; research scenario; project lifecycle timeline, resource distribution at each stage, data integrity statistics; timeline + statistical chart layout]; [Administrative process navigation view; administrative scenarios; service process node diagram, corresponding material list for each node, and processing guide; flowchart + card combination layout].

[0024] The methods for generating university user feedback datasets include: When a user performs any interactive operation in the generated structured view, interactive operation data is collected in real time, including basic operation information (including operation type, operation timestamp, terminal type initiating the operation, etc.), associated identification information (including user ID, associated blueprint ID, associated view ID, resource ID pointed to by the operation, etc.), and operation details (operation duration (e.g., resource preview duration), operation result (e.g., download successful or failed, secondary search successfully initiated, etc.), auxiliary operation filtering and search conditions (e.g., filtering by resources in the past year, etc.). The interactive operation data is associated with the university's exclusive scenario blueprint, campus resources and users according to the pre-set four-dimensional association rules; Among them, the four-dimensional association rule is used to clarify the association key and matching logic of each data dimension by associating users, interactive operations, blueprints and resources; The specific association method is as follows: Interactive Operation - University-Specific Context Blueprint Association: Match detailed information of the corresponding blueprint (task objectives, scene tags, semantic anchor set, retrieval strategy parameters) through blueprint ID and supplement it to the collected data; Interactive Operation - Internal Resource Association: Match the metadata (resource type, resource name, associated knowledge unit, permission level) of the corresponding internal resource by resource ID and supplement it to the collected data; Interactive Operations - User Association: By matching the user's basic scenario attributes (student or teacher, department, grade or title) with the anonymized user ID, the data is supplemented and a basis is provided for subsequent personalized optimization. All associated data after matching and association are integrated to generate a structured dataset of university user feedback, including but not limited to fields such as user ID, blueprint ID, blueprint task goal, scene tag, interaction operation type, operation timestamp, associated resource ID, resource type, operation result, and operation duration.

[0025] Optimizing university-specific contextual blueprints and university-specific knowledge unit networks includes: The dataset of feedback from university users was cleaned and filtered to remove invalid data such as erroneous operations and duplicate submissions. The data was stratified by scenario and user type (teachers, students, etc.) to ensure that the data sample covers the entire scenario and user group. The data was aggregated by blueprint ID-scenario tag-resource type to count the interaction frequency, operation preferences and reasons for invalid operations for various scenarios and resources. Based on the cleaned and filtered data, statistical analysis and association rule mining methods were used to extract the core needs and preferences of different scenarios and user types in universities. Specifically: First, identify the frequently accessed blueprint types, frequently interacted resource types, and frequently used search strategy parameters (e.g., courseware blueprints in teaching scenarios have the highest access volume); then analyze the core reasons for secondary searches and invalid clicks (e.g., inaccurate semantic anchors in blueprints leading to resource matching deviations, incorrect knowledge unit associations leading to group confusion, etc.); summarize the operational preferences of different user types (e.g., teachers prefer "document download + difference comparison" operations, while students prefer "courseware preview + abstract copy" operations). Furthermore, by optimizing the rules based on university scenarios, the university-specific scenario blueprint and university-specific knowledge unit network are optimized and updated respectively; Among them, optimizing the rules to create a blueprint for a specific scenario for universities includes optimizing blueprint generation parameters and optimizing the adaptation of blueprint templates to scenarios; Blueprint generation parameter optimization: Based on the correlation analysis of semantic anchors and resource matching degree in the data of feedback from university users, the semantic anchor clustering threshold was adjusted (e.g., the similarity threshold was lowered from 0.85 to 0.8 in the teaching scenario to improve the comprehensiveness of knowledge point coverage) and the core text extraction weight was increased (the extraction weight of high-frequency effective fields such as "course name" and "project number" was increased). Based on resource click preferences in different scenarios, adjust the weight of resource type (e.g., in scientific research scenarios, increase the weight of "experimental data" from 0.3 to 0.4, which is higher than the weight of "literature" at 0.2) and the time coefficient threshold (e.g. in administrative scenarios, increase the time coefficient of "updated in the last year" from 1.0 to 1.2 to strengthen the priority of time-sensitive resources). Blueprint template scene adaptation optimization: Core field iteration: Adjust the core fields of the blueprint template according to the needs of different scenarios (e.g., add the "Adapt to teaching stage" field in the teaching scenario and the "Project cycle stage" field in the scientific research scenario). Default parameter optimization: Contextualized default parameters are set based on high-frequency preferences (e.g., the default search time range for teaching scenarios is "last 3 years", and the default format requirement for administrative scenarios is "PDF"), reducing manual configuration operations for users; Methods for optimizing university-specific knowledge unit networks through rule optimization include: Based on the university user feedback dataset, identify frequently accessed resources that do not have matching knowledge units (such as "AI course training case" resources in teaching scenarios), add corresponding knowledge units, and supplement attributes such as "core terms, scenario tags, and associated resource types"; For knowledge units with an association error rate of ≥20%, correct the core terminology and scene tags; for knowledge units with an association resource click rate of <10%, supplement the association relationships of high-frequency and effective resources. Based on the user approval rate of the centralized resource and knowledge unit association in the university user feedback dataset (number of clicks on associated resources ÷ total number of times associated resources are displayed), the association confidence threshold for different scenarios is adjusted (e.g., the threshold for the teaching scenario is increased from 0.8 to 0.85 to improve the accuracy of association). For different scenarios, supplementary association rules are added (e.g., in scientific research scenarios, a priority association logic for "experimental data - experimental method knowledge unit" is added, and in administrative scenarios, a mandatory association logic for "service guide - approval node knowledge unit" is added). Improve the retrieval priority of frequently accessed knowledge units-resources and knowledge units-knowledge unit association links in the network, and optimize the efficiency of subsequent grouping and analysis; The optimized university-specific scenario blueprint and university-specific knowledge unit network will be fully updated to the core resource database, and the corresponding rules will be updated synchronously to ensure that subsequent services can directly call the optimized resources; Through a cyclical iteration mechanism of monthly minor iterations and quarterly full optimizations, new user feedback datasets from universities are read regularly, and the process of repeated mining and optimization is repeated to continuously iterate the core data resource system. Through continuous monitoring and iteration mechanisms, a closed-loop link of user feedback-resource optimization-precise service is formed, providing a basis for the next round of iterations.

[0026] Example 2, please refer to Figure 3 As shown, parts not described in detail in this embodiment are described in Embodiment 1. An information aggregation and retrieval method based on artificial intelligence is provided, including: S1: Receive user task objectives, combine associated resource sets and search preferences, call the semantic model to generate semantic anchors, and then generate a university-specific contextual blueprint with permission tags; simultaneously integrate multi-source information within the university to form a standardized information foundation within the university. S2: Based on the standardized information foundation within the university, through the university-specific context blueprint, various information interfaces within the university are scheduled in parallel according to permissions. The query intent is expanded with semantic anchors, and the results are filtered and sorted in combination with retrieval strategies. The university-specific knowledge unit network is associated to form a blueprint-bound retrieval result set. S3: Perform university-specific analysis on the blueprint-bound search result set, generate dynamic aggregated summaries and difference comparison tables, group them according to preset logic and perform visualization processing to form a structured presentation view; S4: Collect user interaction data in the structured presentation view, and associate it with the university-specific context blueprint and on-campus resources to generate a university user feedback dataset; S5: Based on the university user feedback dataset, optimize the university-specific scenario blueprint and university-specific knowledge unit network to form a closed-loop link.

[0027] Example 3: This example discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the operation mode of the artificial intelligence-based information aggregation and retrieval system described above.

[0028] Since the electronic device described in this embodiment is the electronic device used to implement the information aggregation and retrieval method based on artificial intelligence in the embodiments of this application, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the information aggregation and retrieval method based on artificial intelligence described in the embodiments of this application. Therefore, how the electronic device implements the method in the embodiments of this application will not be described in detail here. Any electronic device used by those skilled in the art to implement the information aggregation and retrieval method based on artificial intelligence in the embodiments of this application falls within the scope of protection of this application.

[0029] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters and thresholds in the formulas are set by those skilled in the art according to the actual situation.

[0030] The above description is merely a preferred embodiment of the present invention, and the scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for users of ordinary technical skills, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. An information aggregation and retrieval system based on artificial intelligence, characterized in that, include: Blueprint Standard Module: Receives user task objectives, combines related resource sets and search preferences, calls the semantic model to generate semantic anchors, and then generates a university-specific contextual blueprint with permission tags; simultaneously integrates multi-source information within the university to form a standardized information foundation within the university; Search Binding Module: Based on the standardized information foundation within the university, through the university-specific context blueprint, various information interfaces within the university are scheduled in parallel according to permissions. The semantic anchor points are used to expand the query intent. The results are filtered and sorted in combination with the search strategy. The system is linked to the university-specific knowledge unit network to form a blueprint-bound search result set. Analysis View Module: Performs university-specific analysis on the blueprint-bound search result set, generates dynamic aggregated summaries and difference comparison tables, groups them according to preset logic and performs visualization processing to form a structured presentation view; Interactive Feedback Module: Collects user interaction data in the structured presentation view, and associates it with the university-specific context blueprint and on-campus resources to generate a university user feedback dataset; Optimize the closed-loop module: Based on the university user feedback dataset, optimize the university-specific scenario blueprint and university-specific knowledge unit network to form a closed-loop link.

2. The information aggregation and retrieval system based on artificial intelligence according to claim 1, characterized in that, The methods for generating the university-specific scenario blueprint include: Receive user task objectives, combine related resource sets and search preferences, connect to the unified identity and permission authentication within the school, and generate exclusive permission tags; The system calls a pre-built semantic vector model to extract the core text of the task objective and associated resource set to generate a set of semantic anchor points. Simultaneously, it converts search preferences into executable search strategy parameters, adds scene tags, and integrates them to generate a structured, university-specific contextual blueprint.

3. The information aggregation and retrieval system based on artificial intelligence according to claim 1, characterized in that, The formation methods of the aforementioned standardized information infrastructure within the school include: Collect and temporarily store multi-source information from various distributed information systems within the campus according to a pre-defined hierarchical access policy; The core metadata of multi-source information is extracted by three-level metadata specifications, multi-dimensional data cleaning is performed to form unique and valid data, and then classified according to scenario and resource type and multi-dimensional structured index is built to form a standardized information foundation within the school.

4. The information aggregation and retrieval system based on artificial intelligence according to claim 3, characterized in that, The method of scheduling various information interfaces within the school in parallel according to permissions includes: After a user initiates a search, the target university’s exclusive context blueprint is activated, and the scope of the search data is defined by combining the search strategy parameters with the university’s standardized information foundation. Then, based on the permission tags, a pre-set interface and permission mapping table are matched to filter out the schedulable internal information interfaces and form an interface list; Based on the interface list, the retrieval strategy parameters are refined into retrieval instructions, parallel scheduling parameters are configured, and the retrieval engine is initialized.

5. The information aggregation and retrieval system based on artificial intelligence according to claim 4, characterized in that, The blueprint binding retrieval result set is formed in the following ways: Based on the interface list and refined search instructions, the search is initiated by multi-threaded parallel scheduling of various information interfaces within the school. After the retrieval is completed, the semantic anchor set is expanded by combining the preset query expansion rules to generate accurate query terms, and a second matching filter is performed on the original retrieval results; The search results after secondary matching are subjected to multi-dimensional preliminary filtering according to the search strategy parameters. Then, the comprehensive score of each resource in the preliminary filtering search results is calculated and sorted. The resource-knowledge unit network of universities is associated to form the association pair between resources and knowledge units. After sorting the associated pairs of different interfaces, perform fusion and deduplication processing to generate a structured blueprint binding retrieval result set that binds blueprint IDs and permission identifiers.

6. The information aggregation and retrieval system based on artificial intelligence according to claim 5, characterized in that, The methods for generating the dynamic aggregated summary and difference comparison table include: Link the blueprint to the search result set and the corresponding university-specific scenario blueprint, filter effective resources and mark scenario analysis tags through multi-dimensional analysis rules for the corresponding scenario; The system analyzes tags and knowledge units based on scenarios to extract common information and generate dynamic aggregated summaries. It also identifies differences in annotations of similar resources and generates a difference comparison table. By integrating effective resources, dynamically aggregating summaries and difference comparison tables, a scenario-based analysis result package is formed.

7. The information aggregation and retrieval system based on artificial intelligence according to claim 6, characterized in that, The structured presentation view is formed in the following ways: The results of the contextual analysis are linked to the corresponding university-specific scenario blueprints. The logical grouping rules of the pre-set scenarios are matched to determine the dimension combination of this grouping. Then, the hierarchical logical grouping is executed first according to the main dimension to form a structured grouping and resource mapping relationship. By combining scene tags and grouping results, multi-view content is filled with corresponding scene-specific view templates to generate multi-type structured presentation views.

8. The information aggregation and retrieval system based on artificial intelligence according to claim 7, characterized in that, The methods for generating the university user feedback dataset include: Real-time data collection of user interaction in the structured presentation view; association of university-specific contextual blueprints, campus resources, and users according to four-dimensional association rules; integration of all associated data; generation of a structured university user feedback dataset.

9. The information aggregation and retrieval system based on artificial intelligence according to claim 8, characterized in that, The methods for optimizing university-specific contextual blueprints and university-specific knowledge unit networks include: Based on a dataset of feedback from university users, we can explore the core needs and behavioral preferences of different user types in different scenarios within universities. Furthermore, by optimizing the rules based on university scenarios, the university-specific scenario blueprint and university-specific knowledge unit network are optimized and updated respectively.

10. An information aggregation and retrieval method based on artificial intelligence, implemented based on the information aggregation and retrieval system based on artificial intelligence as described in claims 1 to 9, characterized in that, include: S1: Receive user task objectives, combine associated resource sets and search preferences, call the semantic model to generate semantic anchors, and then generate a university-specific contextual blueprint with permission tags; simultaneously integrate multi-source information within the university to form a standardized information foundation within the university. S2: Based on the standardized information foundation within the university, through the university-specific context blueprint, various information interfaces within the university are scheduled in parallel according to permissions. The query intent is expanded with semantic anchors, and the results are filtered and sorted in combination with retrieval strategies. The university-specific knowledge unit network is associated to form a blueprint-bound retrieval result set. S3: Perform university-specific analysis on the blueprint-bound search result set, generate dynamic aggregated summaries and difference comparison tables, group them according to preset logic and perform visualization processing to form a structured presentation view; S4: Collect user interaction data in the structured presentation view, and associate it with the university-specific context blueprint and on-campus resources to generate a university user feedback dataset; S5: Based on the university user feedback dataset, optimize the university-specific scenario blueprint and university-specific knowledge unit network to form a closed-loop link.

Images