User service recommendation method and device, equipment, medium and product

Abstract

The invention provides a user service recommendation method and device, equipment and a medium, and the method comprises the steps: analyzing original data from a plurality of heterogeneous education data sources, and generating a standardized intermediate data stream; performing semantic alignment on entities in the intermediate data stream and entities in a pre-constructed education knowledge graph to generate an aligned knowledge graph; finely adjusting the pre-training language model by using the educational domain corpus to generate a domain enhancement model; wherein the aligned knowledge graph is used as a constraint in the fine tuning process; and in response to a service request of a current user, generating a personalized education intelligent service for the current user based on the service request, the aligned knowledge graph and the domain enhancement model.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of big data, in particular to a user service recommendation method and device, equipment, medium and product. BACKGROUND

[0002] There are various heterogeneous data sources in the current education informatization system, including structured data in student archives management system (such as student information, score records), semi-structured data in learning management system (such as learning behavior logs, homework submission records), and unstructured data in teaching resource platform (such as teaching videos, exercise analysis texts). These data are scattered in different systems, and there are significant semantic heterogeneity and mode conflict problems, which makes it difficult to accurately recommend personalized services for users. SUMMARY

[0003] In view of the above problems, the present application provides a user service recommendation method, device, equipment, medium and product, which at least solves one of the above problems.

[0004] According to a first aspect of the present application, a user service recommendation method is provided, comprising:

[0005] parsing raw data from multiple heterogeneous education data sources to generate standardized intermediate data streams;

[0006] aligning entities in the intermediate data streams with entities in a pre-built education knowledge graph semantically to generate an aligned knowledge graph;

[0007] using an education field corpus to fine-tune a pre-trained language model to generate a field enhanced model; wherein the fine-tuning process uses the aligned knowledge graph as a constraint;

[0008] in response to a service request of a current user, generating personalized education intelligent services for the current user based on the service request, the aligned knowledge graph and the field enhanced model.

[0009] In some embodiments, the aligning entities in the intermediate data streams with entities in a pre-built education knowledge graph semantically comprises: constructing a data mode graph corresponding to the intermediate data streams; based on a graph neural network, calculating the semantic similarity of nodes in the data mode graph and nodes in the education knowledge graph; according to the semantic similarity, establishing a mapping relationship between entities in the intermediate data streams and entities in the education knowledge graph.

[0010] In some embodiments, calculating the semantic similarity between nodes in the data pattern graph and nodes in the educational knowledge graph based on a graph neural network includes: using a heterogeneous graph attention network to encode the data pattern graph and the educational knowledge graph respectively to obtain node feature vectors; and calculating the semantic similarity between nodes based on the node feature vectors through a cross-graph attention mechanism.

[0011] In some embodiments, the method further includes: transferring the mapping learned from one data source to another data source by feature space projection; wherein, during the transfer process, the underlying parameters of the heterogeneous graph attention network are frozen, and the parameters of the top-level classifier are fine-tuned.

[0012] In some embodiments, using the aligned knowledge graph as a constraint includes: introducing a regularization term based on triples in the aligned knowledge graph in the loss function for fine-tuning the pre-trained language model, the regularization term being used to minimize the distance between the vector representations of the head entity, relation, and tail entity.

[0013] In some embodiments, the fine-tuning process further includes: constructing professional terms and their synonymous colloquial expressions as positive sample pairs; constructing professional terms and unrelated expressions as negative sample pairs; and increasing the semantic similarity of the positive sample pairs and decreasing the semantic similarity of the negative sample pairs by optimizing the contrastive learning loss function.

[0014] According to a second aspect of this application, a user service recommendation device is provided, comprising:

[0015] The data generation module is used to parse raw data from multiple heterogeneous educational data sources and generate standardized intermediate data streams.

[0016] The semantic alignment module is used to semantically align entities in the intermediate data stream with entities in the pre-built educational knowledge graph to generate an aligned knowledge graph.

[0017] The model generation module is used to fine-tune a pre-trained language model using educational domain corpus to generate a domain-enhanced model; wherein the fine-tuning process uses the aligned knowledge graph as a constraint.

[0018] The service generation module is used to respond to the service request of the current user and generate a personalized educational intelligent service for the current user based on the service request, the aligned knowledge graph and the domain enhancement model.

[0019] According to a third aspect of this application, an electronic device is provided, comprising: one or more processors; and a memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method according to the first aspect.

[0020] According to a fourth aspect of this application, a computer-readable storage medium is provided having a computer program or instructions stored thereon, which, when executed by a processor, implement the steps of the method according to the first aspect. Attached Figure Description

[0021] The above-mentioned contents, other objects, features and advantages of this application will become clearer from the following description of embodiments with reference to the accompanying drawings, in which:

[0022] Figure 1 A flowchart illustrating a user service recommendation method according to an embodiment of this application is shown schematically.

[0023] Figure 2 The diagram illustrates a training graph of a domain enhancement model according to an embodiment of this application.

[0024] Figure 3 This schematic diagram illustrates the structure of a user service recommendation device according to an embodiment of this application;

[0025] Figure 4 A schematic diagram of an electronic device according to an embodiment of this application is shown. Detailed Implementation

[0026] The embodiments of this application will now be described with reference to the accompanying drawings. However, it should be understood that these descriptions are exemplary only and are not intended to limit the scope of this application. In the following detailed description, numerous specific details are set forth to provide a thorough understanding of the embodiments of this application for ease of explanation. However, it will be apparent that one or more embodiments may be implemented without these specific details. Furthermore, descriptions of well-known structures and technologies are omitted in the following description to avoid unnecessarily obscuring the concepts of this application.

[0027] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0028] All terms used herein (including technical and scientific terms) have the meanings commonly understood by those skilled in the art, unless otherwise defined. It should be noted that the terms used herein are to be interpreted in a manner consistent with the context of this specification, and not in an idealized or overly rigid way.

[0029] When using expressions such as "at least one of A, B and C", they should generally be interpreted in accordance with the meaning that is commonly understood by those skilled in the art (e.g., "a system having at least one of A, B and C" should include, but is not limited to, a system having A alone, a system having B alone, a system having C alone, a system having A and B, a system having A and C, a system having B and C, and / or a system having A, B and C, etc.).

[0030] In the technical solution disclosed herein, the user information (including but not limited to user personal information, user image information, user transaction record information, user device information such as location information) and data (including but not limited to data used for analysis, stored data, and displayed data) involved are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, storage, use, processing, transmission, provision, disclosure, and application of related data all comply with relevant laws, regulations, and standards, take necessary confidentiality measures, do not violate public order and good morals, and provide corresponding operation entry points for users to choose to authorize or refuse.

[0031] In scenarios involving automated decision-making using personal information, the methods, devices, and systems provided in this disclosure all offer users corresponding entry points for choosing to agree to or reject the automated decision-making results. If the user chooses to reject, the process proceeds to the expert decision-making stage. Here, "automated decision-making" refers to the activity of automatically analyzing and evaluating an individual's behavioral habits, interests, or economic, health, and credit status through computer programs, and then making a decision. Here, "expert decision-making" refers to the activity of making decisions by personnel who specialize in a particular field, possess specialized experience, knowledge, and skills, and have reached a certain level of professional expertise.

[0032] Figure 1 A flowchart illustrating a user service recommendation method according to an embodiment of this application is shown schematically. Figure 2 The diagram illustrates a training graph of a domain enhancement model according to an embodiment of this application.

[0033] like Figure 1 As shown, the user service recommendation method includes operations S110-S140.

[0034] In operation S110, raw data from multiple heterogeneous educational data sources are parsed to generate standardized intermediate data streams.

[0035] According to embodiments of the present invention, heterogeneous educational data sources can refer to multiple educational information systems that differ in data structure, storage format, and semantic definition. Heterogeneous educational data sources may include structured data sources, semi-structured data sources, and unstructured data sources.

[0036] Structured data sources, such as relational databases (e.g., MySQL tables) in a Student Information Management System (SIS), contain table records with fields such as student ID, name, and subject grades.

[0037] Semi-structured data sources, such as JSON-formatted behavior logs output by learning management systems (LMS), record raw data such as students' video viewing time, assignment submission time, and forum posts.

[0038] Unstructured data sources, such as PDF electronic lesson plans, exercise explanation texts, or teaching videos with subtitles in digital teaching resource platforms.

[0039] According to an embodiment of the present invention, a standardized intermediate data stream can refer to a unified data format output after parsing. In this embodiment, the data stream can be a data sequence in an attribute graph model format. Each node represents an entity (e.g., a student, a test question), each edge represents a relationship (e.g., "belongs to", "answered"), and both nodes and edges have attributes (e.g., the student's "grade", the test question's "difficulty").

[0040] In operation S120, entities in the intermediate data stream are semantically aligned with entities in the pre-built educational knowledge graph to generate an aligned knowledge graph.

[0041] According to embodiments of the present invention, a pre-constructed educational knowledge graph can refer to a pre-constructed semantic network containing core concepts, entities, and relationships between them within the field of education. For example, the graph may contain triples such as "<Pythagorean theorem, is A, mathematical theorem>" or "<Newton's second law, related to, acceleration>".

[0042] According to embodiments of the present invention, the aligned knowledge graph can refer to a graph formed by incorporating specific instance data from intermediate data streams on the basis of the original educational knowledge graph. For example, it no longer only contains the abstract concept of "student", but can include the specific student individual "Zhang San", and associate the fact that "Zhang San" "score 98 points in the final math exam" as an attribute in the graph.

[0043] In operation S130, a pre-trained language model is fine-tuned using an educational domain corpus to generate a domain-enhanced model; the fine-tuning process uses the aligned knowledge graph as a constraint.

[0044] According to embodiments of the present invention, a pre-trained model can refer to a model that is pre-trained on a large-scale general corpus and has strong semantic understanding capabilities, such as BERT, RoBERTa, etc.

[0045] According to embodiments of the present invention, educational corpus can refer to a collection of professional texts from the field of education, including but not limited to textbooks, teaching aids, curriculum standards, academic papers, and high-quality lesson plans.

[0046] According to an embodiment of the present invention, the fine-tuning process using the aligned knowledge graph as a constraint can be achieved by introducing a knowledge graph regularization term into the fine-tuning loss function. This regularization term enforces that the model's representation of entity relations must be consistent with the triple structure in the aligned knowledge graph. Its loss function can be simplified to:

[0047] Total loss = Language model task loss + λ * Σ||vector(head entity) + vector(relation) - vector(tail entity)||²

[0048] Here, λ is a hyperparameter that sums up the triples in the graph. Through this constraint, the model can learn structured and logical educational knowledge from the graph while learning the text.

[0049] In operation S140, in response to the current user's service request, a personalized educational intelligent service is generated for the current user based on the service request, the aligned knowledge graph, and the domain enhancement model.

[0050] According to embodiments of the present invention, the service request of the current user can refer to a query or instruction initiated in real time by the end user (student or teacher) seeking a specific service. For example, a student inputting "Please recommend suitable physics exercises for me", or a teacher operation captured by the system "Generate a learning analysis report for class A".

[0051] According to embodiments of the present invention, generating personalized educational intelligent services can refer to the collaborative invocation of a domain enhancement model and an aligned knowledge graph through a multi-task inference engine. The domain enhancement model can be used to deeply understand the semantics of service requests (e.g., accurately understanding that "suitable for me" means considering the student's knowledge weaknesses). Queries and inferences are performed in the aligned knowledge graph (e.g., retrieving the student's weak knowledge points and finding exercises of moderate difficulty related to these knowledge points). Combining the information from both sources, a final service is generated. For example, a personalized list of exercises or a visualized learning progress diagnostic report may be output.

[0052] According to the embodiments of the present invention, it is possible to achieve automatic and dynamic fusion of multi-source heterogeneous educational data, generate a high-quality aligned knowledge graph, and train an enhanced model with deep understanding of the education field, and finally be able to provide accurate and personalized intelligent education services for users.

[0053] In some embodiments of the present invention, a data schema graph corresponding to the intermediate data stream can be constructed; based on a graph neural network, the semantic similarity between the nodes in the data schema graph and the nodes in the educational knowledge graph is calculated; according to the semantic similarity, a mapping relationship between the entities in the intermediate data stream and the entities in the educational knowledge graph is established.

[0054] A data schema graph is a method of representing the schema information of the data source itself in a graph structure. In this graph, nodes are defined as fields in the data schema (such as "score") and knowledge graph entities (such as "math score"). Edges represent field co-occurrence relationships (such as "score" and "class" being in the same table in the database) and entity semantic relationships (such as "math score" having a parent-child relationship with "algebra").

[0055] The data schema graph and the educational knowledge graph are jointly input into a graph neural network model. Through learning, the model can generate a low-dimensional feature vector (i.e., node embedding) for each node in the two graphs. These vectors can capture the structural information and name semantic information of the nodes. Subsequently, the cosine similarity between the node vectors of the data schema graph and the node vectors of the knowledge graph is calculated to quantify the semantic association strength between them. For example, the vector of the node "stu_perf" in the data schema graph may have a high cosine similarity with the vector of the node "academic performance" in the educational knowledge graph. Finally, establishing the mapping relationship means finding and associating one or more nodes with the highest similarity in the knowledge graph for each node in the data schema graph, forming mapping pairs such as "stu_perf" → "academic performance".

[0056] In some embodiments of the present invention, a heterogeneous graph attention network can be used to encode the data schema graph and the educational knowledge graph respectively to obtain node feature vectors; based on the node feature vectors, the semantic similarity between nodes is calculated through a cross-graph attention mechanism.

[0057] A heterogeneous graph attention network is a dedicated network designed for the characteristics that the data schema graph and the knowledge graph may have different node types and relationship types (i.e., "heterogeneous"). Its specific implementation may include node type-aware encoding, which can refer to the network differentiating whether a node comes from the data schema graph (such as "field" type) or the knowledge graph (such as "concept" type) when processing nodes, and using different parameters to encode different types of nodes and relationships, so as to more accurately capture their specific attributes.

[0058] Cross-graph attention mechanisms allow nodes in one graph to directly attend to all relevant nodes in another graph, aggregating this information in a weighted manner. Semantic similarity is directly reflected by the calculated attention coefficient. The higher the coefficient, the stronger the semantic similarity between nodes. This mechanism can dynamically and discriminatively assess the strength of associations between nodes across graphs, making it more accurate and powerful than simple cosine similarity calculation.

[0059] In some embodiments, the mapping learned from one data source can be transferred to another data source through feature space projection; wherein, during the transfer process, the underlying parameters of the heterogeneous graph attention network are frozen and the top-level classifier parameters are fine-tuned.

[0060] The alignment model trained on a system with relatively complete data patterns and annotations (e.g., a successfully aligned "student registration management system") is considered a "pre-trained model" with general semantic alignment capabilities. The migration process is initiated when a new data source with very little labeled data is needed (e.g., a new "online experimental platform").

[0061] The Heterogeneous Graph Attention Network (HGAT) is responsible for learning basic, general graph structure features and semantic features (e.g., how to understand "field name co-occurrence" or "concept hierarchy relationship"). These features are shared across different data sources, so fixing (freezing) their parameters preserves the general alignment capabilities that the model has already learned and significantly reduces the number of parameters that need to be trained.

[0062] The top layer of the network is typically a classifier used for the final mapping relationship classification. These parameters are more inclined to learn features related to specific data source patterns. Therefore, a small number of labeled samples from the target data source (e.g., only 50 manually labeled fields from an "online experimental platform") can be used to fine-tune these parameters, allowing them to quickly adapt to the characteristics of the new data source.

[0063] In some embodiments, a regularization term based on triples in the aligned knowledge graph can be introduced into the loss function for fine-tuning the pre-trained language model. This regularization term is used to minimize the distance between the vector representations of the head entity, relation, and tail entity.

[0064] The idea of ​​the classic TransE algorithm in knowledge graph embedding can be adopted. For each triple (h, r, t) in the aligned knowledge graph (e.g., "<trigonometric function, belongs to, algebra>"), the vector representation learned by the model is required to satisfy: vector(h) + vector(r) ≈ vector(t).

[0065] Therefore, the regularization loss L_KG of the knowledge graph is defined as the sum of distances over all triples:

[0066] L_KG = Σ_{(h,r,t) in KG} ||vector(h) + vector(r) - vector(t)||²

[0067] L_MLM is the loss of the masked language model, L_total = L_MLM + λ * L_KG. By minimizing L_total, the model learns educational texts while its parameter space is constrained and guided by structured knowledge, thus effectively injecting factual knowledge into the model.

[0068] In some embodiments, the fine-tuning process may also construct professional terms and their synonymous colloquial expressions as positive sample pairs; construct professional terms and unrelated expressions as negative sample pairs; and increase the semantic similarity of the positive sample pairs and decrease the semantic similarity of the negative sample pairs by optimizing the contrastive learning loss function.

[0069] Construct professional terms and their synonymous, colloquial expressions as positive sample pairs. For example: Positive sample pair 1: ("Pythagorean theorem", "the square relationship between the three sides of a right triangle"); Positive sample pair 2: ("cognitive load", "the burden felt by the brain during learning").

[0070] Construct professional terms and unrelated expressions as negative sample pairs. For example: negative sample pair 1: ("Pythagorean theorem", "quadratic equation formula") (different concepts within the same field); negative sample pair 2: ("cognitive load", "classroom attendance") (unrelated concepts).

[0071] Based on the above-described user service recommendation method, this application also provides a user service recommendation apparatus. The following will be combined with... Figure 3 The device is described in detail.

[0072] Figure 3 A schematic block diagram of a user service recommendation device according to an embodiment of this application is shown.

[0073] like Figure 3 As shown, the user service recommendation device in this embodiment includes a data generation module 310, a semantic alignment module 320, a model generation module 330, and a service generation module 340.

[0074] The data generation module 310 is used to parse raw data from multiple heterogeneous educational data sources and generate a standardized intermediate data stream. In one embodiment, the feature extraction module 310 can be used to perform the operation S110 described above, which will not be repeated here.

[0075] The semantic alignment module 320 is used to semantically align entities in the intermediate data stream with entities in the pre-built educational knowledge graph to generate an aligned knowledge graph. In one embodiment, the mapping module 320 can be used to perform the operation S120 described above, which will not be repeated here.

[0076] The model generation module 330 is used to fine-tune a pre-trained language model using educational domain corpus to generate a domain-enhanced model; wherein the fine-tuning process uses the aligned knowledge graph as a constraint. In one embodiment, the computation module 330 can be used to perform the operation S130 described above, which will not be repeated here.

[0077] The service generation module 340 is used to respond to the service request of the current user and generate a personalized educational intelligent service for the current user based on the service request, the aligned knowledge graph, and the domain enhancement model. In one embodiment, the fusion module 340 can be used to perform the operation S140 described above, which will not be repeated here.

[0078] In some embodiments, semantic alignment of entities in the intermediate data stream with entities in a pre-built educational knowledge graph includes: constructing a data pattern graph corresponding to the intermediate data stream; calculating the semantic similarity between nodes in the data pattern graph and nodes in the educational knowledge graph based on a graph neural network; and establishing a mapping relationship between entities in the intermediate data stream and entities in the educational knowledge graph based on the semantic similarity.

[0079] In some embodiments, the semantic similarity calculation between nodes in the data pattern graph and nodes in the educational knowledge graph based on graph neural networks includes: using a heterogeneous graph attention network to encode the data pattern graph and the educational knowledge graph respectively to obtain node feature vectors; and calculating the semantic similarity between nodes based on the node feature vectors through a cross-graph attention mechanism.

[0080] In some embodiments, the method further includes: transferring the mapping learned from one data source to another data source by feature space projection; wherein, during the transfer process, the underlying parameters of the heterogeneous graph attention network are frozen and the top-level classifier parameters are fine-tuned.

[0081] In some embodiments, using the aligned knowledge graph as a constraint includes: introducing a regularization term based on triples in the aligned knowledge graph in the loss function for fine-tuning the pre-trained language model, the regularization term being used to minimize the distance between the vector representations of the head entity, relation, and tail entity.

[0082] In some embodiments, the fine-tuning process further includes: constructing professional terms and their synonymous colloquial expressions as positive sample pairs; constructing professional terms and unrelated expressions as negative sample pairs; and increasing the semantic similarity of the positive sample pairs and decreasing the semantic similarity of the negative sample pairs by optimizing the contrastive learning loss function.

[0083] According to embodiments of this application, any multiple modules among the data generation module 310, semantic alignment module 320, model generation module 330, and service generation module 340 can be merged into one module, or any one of these modules can be split into multiple modules. Alternatively, at least some of the functions of one or more of these modules can be combined with at least some of the functions of other modules and implemented in one module. According to embodiments of this application, at least one of the data generation module 310, semantic alignment module 320, model generation module 330, and service generation module 340 can be at least partially implemented as hardware circuitry, such as a field-programmable gate array (FPGA), a programmable logic array (PLA), a system-on-a-chip, a system-on-a-substrate, a system-on-package, an application-specific integrated circuit (ASIC), or implemented in hardware or firmware by any other reasonable means of integrating or packaging the circuitry, or implemented in software, hardware, or firmware, or in any suitable combination of any of these three implementation methods. Alternatively, at least one of the data generation module 310, semantic alignment module 320, model generation module 330, and service generation module 340 may be implemented at least partially as a computer program module, which can perform corresponding functions when the computer program module is run.

[0084] Figure 4 A block diagram schematically illustrates an electronic device suitable for implementing a user service recommendation method according to an embodiment of this application.

[0085] like Figure 4 As shown, an electronic device according to an embodiment of this application includes a processor 401, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 402 or a program loaded from a storage portion 408 into a random access memory (RAM) 403. The processor 401 may include, for example, a general-purpose microprocessor (e.g., a CPU), an instruction set processor and / or an associated chipset and / or a special-purpose microprocessor (e.g., an application-specific integrated circuit (ASIC)), etc. The processor 401 may also include onboard memory for caching purposes. The processor 401 may include a single processing unit or multiple processing units for performing different actions of the method flow according to an embodiment of this application.

[0086] RAM 403 stores various programs and data required for the operation of the electronic device. Processor 401, ROM 402, and RAM 403 are interconnected via bus 404. Processor 401 executes various operations of the method flow according to embodiments of this application by executing programs in ROM 402 and / or RAM 403. It should be noted that programs may also be stored in one or more memories other than ROM 402 and RAM 403. Processor 401 may also execute various operations of the method flow according to embodiments of this application by executing programs stored in said one or more memories.

[0087] According to embodiments of this application, the electronic device may further include an input / output (I / O) interface 405, which is also connected to a bus 404. The electronic device may also include one or more of the following components connected to the input / output (I / O) interface 405: an input section 406 including a keyboard, mouse, etc.; an output section 407 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and a speaker, etc.; a storage section 408 including a hard disk, etc.; and a communication section 409 including a network interface card such as a LAN card, modem, etc. The communication section 409 performs communication processing via a network such as the Internet. A drive 410 is also connected to the input / output (I / O) interface 405 as needed. A removable medium 411, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on the drive 410 as needed so that computer programs read from it can be installed into the storage section 408 as needed.

[0088] This application also provides a computer-readable storage medium, which may be included in the device / apparatus / system described in the above embodiments; or it may exist independently and not assembled into the device / apparatus / system. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to the embodiments of this application.

[0089] According to embodiments of this application, the computer-readable storage medium can be a non-volatile computer-readable storage medium, such as including but not limited to: portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, the computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, according to embodiments of this application, the computer-readable storage medium may include ROM 402 and / or RAM 403 and / or one or more memories other than ROM 402 and RAM 403 described above.

[0090] Embodiments of this application also include a computer program product comprising a computer program containing program code for performing the methods shown in the flowchart. When the computer program product is run on a computer system, the program code is used to enable the computer system to implement the user service recommendation method provided in the embodiments of this application.

[0091] When the computer program is executed by the processor 401, it performs the functions defined in the system / apparatus of this application embodiment. According to the embodiments of this application, the systems, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0092] In one embodiment, the computer program may rely on a tangible storage medium such as an optical storage device or a magnetic storage device. In another embodiment, the computer program may also be transmitted and distributed in the form of signals over a network medium, and downloaded and installed via communication section 409, and / or installed from removable medium 411. The program code contained in the computer program can be transmitted using any suitable network medium, including but not limited to: wireless, wired, etc., or any suitable combination thereof.

[0093] In such an embodiment, the computer program can be downloaded and installed from a network via communication section 409, and / or installed from removable medium 411. When the computer program is executed by processor 401, it performs the functions defined in the system of this application embodiment. According to embodiments of this application, the systems, devices, apparatuses, modules, units, etc., described above can be implemented by computer program modules.

[0094] According to embodiments of this application, program code for executing the computer programs provided in the embodiments of this application can be written in any combination of one or more programming languages. Specifically, these computational programs can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. Programming languages ​​include, but are not limited to, languages ​​such as Java, C++, Python, "C", or similar programming languages. The program code can be executed entirely on the user's computing device, partially on the user's device, partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0095] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0096] Those skilled in the art will understand that the features described in the various embodiments of this application can be combined and / or combined in various ways, even if such combinations or combinations are not explicitly described in this application. In particular, the features described in the various embodiments of this application can be combined and / or combined in various ways without departing from the spirit and teachings of this application. All such combinations and / or combinations fall within the scope of this application.

Claims

1. A user service recommendation method characterized by comprising: The method comprises: parsing raw data from multiple heterogeneous educational data sources to generate a standardized intermediate data stream; semantically aligning entities in the intermediate data stream with entities in a pre-built educational knowledge graph to generate an aligned knowledge graph; fine-tuning a pre-trained language model using an educational domain corpus to generate a domain-enhanced model; wherein the fine-tuning process uses the aligned knowledge graph as a constraint; in response to a service request from a current user, generating personalized educational intelligent services for the current user based on the service request, the aligned knowledge graph, and the domain-enhanced model.

2. The method of claim 1, wherein, The semantic alignment of entities in the intermediate data stream with entities in the pre-built educational knowledge graph comprises: constructing a data schema graph corresponding to the intermediate data stream; based on a graph neural network, calculating the semantic similarity between nodes in the data schema graph and nodes in the educational knowledge graph; according to the semantic similarity, establishing a mapping relationship between entities in the intermediate data stream and entities in the educational knowledge graph.

3. The method of claim 2, wherein, The graph neural network-based calculation of the semantic similarity between nodes in the data schema graph and nodes in the educational knowledge graph comprises: using a heterogeneous graph attention network to encode the data schema graph and the educational knowledge graph respectively to obtain node feature vectors; based on the node feature vectors, calculating the semantic similarity between nodes through a cross-graph attention mechanism.

4. The method of claim 2, wherein, The method further comprises: transferring the mapping relationship learned from one data source to another data source through feature space projection; wherein the bottom-layer parameters of the heterogeneous graph attention network are frozen during the transfer, and the top-layer classifier parameters are fine-tuned.

5. The method of claim 1, wherein, The use of the aligned knowledge graph as a constraint comprises: introducing a regularization term based on triples in the aligned knowledge graph into the loss function for fine-tuning the pre-trained language model, which is used to minimize the distance between the vector representations of the head entity, the relationship, and the tail entity.

6. The method of claim 1, wherein, The fine-tuning process further comprises: constructing professional terms and synonymous colloquial expressions as positive sample pairs; constructing professional terms and unrelated expressions as negative sample pairs; by optimizing the contrastive learning loss function, increasing the semantic similarity of the positive sample pairs and reducing the semantic similarity of the negative sample pairs.

7. A user service recommendation apparatus characterized by comprising: The method comprises: a data generation module for parsing raw data from multiple heterogeneous educational data sources to generate a standardized intermediate data stream; a semantic alignment module for semantically aligning entities in the intermediate data stream with entities in a pre-built educational knowledge graph to generate an aligned knowledge graph; a model generation module for fine-tuning a pre-trained language model using an educational domain corpus to generate a domain-enhanced model; wherein the fine-tuning process uses the aligned knowledge graph as a constraint; a service generation module for generating personalized educational intelligent services for a current user in response to a service request from the current user based on the service request, the aligned knowledge graph, and the domain-enhanced model.

8. An electronic device comprising: one or more processors; A memory for storing one or more computer programs, wherein the one or more processors execute the one or more computer programs to implement the steps of the method according to any one of claims 1-6.

9. A computer readable storage medium having stored thereon a computer program or instructions, characterized in that, The computer program or instructions, when executed by a processor, implement the steps of the method according to any one of claims 1-6.

10. A computer program product comprising computer programs or instructions, characterized in that, The computer program or instructions, when executed by a processor, implement the steps of the method according to any one of claims 1-6.

Images