Student research interest mining method and system based on text content and academic network
By pre-trained language models and graph convolutional neural networks modeling on dynamic academic networks, the problem of insufficient research interest description and limited accuracy of prediction models is solved, and more accurate interest prediction and academic trend capture are achieved.
Patent Information
- Application Number
- CN202411442211.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-10-16
- Publication Date
- 2024-11-15
- Estimated Expiration
- 2044-10-16
AI Technical Summary
In the mining of scholars' research interests, there are fuzzy definitions of scholars' interest, insufficient text content embedding methods, and lack of considerations of heterostructure information and temporal dynamics of academic networks in the mining of scholars' research interests, resulting in insufficient accuracy in interest descriptions and limited accuracy of prediction models.
The pre-trained language model is used to generate node embeddings, build dynamic academic heterogeneous graphs, and use graph convolutional neural network DRGCN model to model scholars and domain features on dynamic academic networks, generate node vectors with time characteristics, and determine them through vector internal product. The prediction results of scholars study interests and make full use of the heterogeneous information and time dimensions of academic networks.
It improves the accuracy of scholars' research interest prediction, generates enhanced text semantic features, can more accurately capture the development trends of the academic community, and provides more accurate interest labels and prediction results.
Smart Images

Figure CN118964607A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of deep learning and data mining, and in particular to a method and system for mining scholars' research interests based on text content and academic networks. Background Art
[0002] At present, scholars’ research interests are mostly mined in an unsupervised manner, that is, from scholars’ personal profile information or the text of their academic achievements. Keywords are identified and extracted from the relevant text content of scholars as the direction of their research interests. A further step is to use topic models to model scholars’ research interests. Compared with word-level interest mining, it further extracts semantic information, making the description of scholars’ interests more accurate. With the rise of graph networks, related studies have begun to consider using scholars’ co-authorship networks and citation networks to model the changes in scholars’ research interests based on deep learning, but they do not fully utilize the rich information of heterogeneous academic networks.
[0003] Specifically, the existing technology has the following three problems: first, the definition of scholars' interests is vague, and accurate interest tags cannot be generated. Previous interests are obtained from word frequency statistics or keyword recognition in scholars' achievements, and the description of scholars' research interests is not accurate enough; second, there are deficiencies in the text content embedding method, which leads to limited accuracy of the prediction model; finally, the prediction model lacks information on the heterogeneous structure of the academic network and consideration of temporal dynamics.
[0004] Therefore, the prior art still needs to be improved and enhanced. Summary of the invention
[0005] The technical problem to be solved by the present invention is that, in view of the above-mentioned defects of the prior art, a method and system for mining scholars' research interests based on text content and academic networks is provided, aiming to solve the problems that the prior art is not accurate enough in describing scholars' research interests, the accuracy of the prediction model is limited, and there is a lack of information on the heterogeneous structure of the academic network and consideration of time dynamics.
[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is as follows: In a first aspect, the present invention provides a method for mining scholars' research interests based on text content and academic networks, wherein the method comprises: Based on the paper data, a pre-trained language model is used to generate node embeddings, including technical word node embeddings, paper node embeddings, scholar node embeddings, and field node embeddings; Construct an academic heterogeneous graph based on academic data, and use the node embedding to construct a dynamic academic network, model scholar and field characteristics on the dynamic academic network, and generate scholar node vectors and field node vectors with time characteristics; The inner product of the scholar node vector and the field node vector is performed to obtain the score between the scholar and the field, and the prediction result reflecting the scholar's research interest is determined based on the score.
[0007] In one implementation, the method of generating node embeddings based on paper data using a pre-trained language model includes: Based on the paper data, the text content of the scholar's paper title and abstract is spliced, and the named entity recognition method based on the Bert model is used to identify technical words from the text content, and the word2vec model is used to generate technical word node embedding; The SimCSE model is used to connect two fully connected layers and trained using the paper multi-label classification task to generate paper node embeddings; For scholar nodes and domain nodes, the node embeddings of their neighbor nodes are mapped to the same semantic space and then aggregated to obtain scholar node embedding and domain node embedding.
[0008] In one implementation, the academic heterogeneous graph includes four types of nodes and six types of relationship information, wherein the four types of nodes include: technical word nodes, paper nodes, scholar nodes, and field nodes; The six types of relationship information include: relationship information between scholars and papers, relationship information between scholars, relationship information between scholars and fields, relationship information between papers, relationship information between papers and technical terms, and relationship information between papers and fields.
[0009] In one implementation, the use of the node embedding to construct a dynamic academic network, modeling scholar and field characteristics on the dynamic academic network, and generating scholar node vectors and field node vectors with time characteristics include: Based on node embedding, the relationship information in the academic heterogeneous graph is modeled, and a dynamic academic network is constructed by slicing the paper collection into five-year time slices; Delete the edges between papers and fields, and between scholars and fields on the academic heterogeneous graph of the last time slice of the dynamic academic network, generate vector representations of each node under the static graph on the dynamic academic network, and uniformly input them into the Transformer model to learn the node temporal features; The node time series features of neighbor nodes on the academic heterogeneous graph are aggregated to generate scholar node vectors and field node vectors with time characteristics.
[0010] In one implementation, determining a prediction result based on the score to reflect the scholar's research interests includes: Sort the calculated scores of scholars and fields to obtain a score sequence; The scholars and fields corresponding to the scores at the front preset positions in the score sequence are used as the prediction results.
[0011] In one implementation, the dynamic academic network is a graph convolutional neural network DRGCN model.
[0012] In one implementation, the method further includes: The graph convolutional neural network DRGCN model is tuned using contrastive learning, and Margin Loss is used as the loss function for training.
[0013] In a second aspect, an embodiment of the present invention further provides a system for mining scholars' research interests based on text content and academic networks, wherein the system comprises: A node embedding generation module, which is used to generate node embeddings based on paper data using a pre-trained language model, wherein the node embeddings include technical word node embeddings, paper node embeddings, scholar node embeddings, and field node embeddings; A scholar interest feature extraction module is used to construct an academic heterogeneous graph based on academic data, and use the node embedding to construct a dynamic academic network, model scholar and field features on the dynamic academic network, and generate scholar node vectors and field node vectors with time characteristics; The scholar research interest prediction module is used to perform vector inner product on the scholar node vector and the field node vector to obtain the score between the scholar and the field, and determine the prediction result reflecting the scholar's research interest based on the score.
[0014] In a third aspect, an embodiment of the present invention further provides a computer device, wherein the computer device includes a memory, a processor, and a program for mining scholars' research interests based on text content and academic networks, which is stored in the memory and can be run on the processor. When the processor executes the program for mining scholars' research interests based on text content and academic networks, the steps of the method for mining scholars' research interests based on text content and academic networks in any one of the above-mentioned schemes are implemented.
[0015] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein a program for mining scholars' research interests based on text content and academic networks is stored on the computer-readable storage medium. When the program for mining scholars' research interests based on text content and academic networks is executed by a processor, the steps of the method for mining scholars' research interests based on text content and academic networks described in any one of the above-mentioned schemes are implemented.
[0016] Beneficial effects: Compared with the prior art, the present invention provides a method for mining the research interests of scholars based on text content and academic networks. The present invention first generates node embeddings based on paper data using a pre-trained language model, and the node embeddings include technical word node embeddings, paper node embeddings, scholar node embeddings, and field node embeddings. Then, an academic heterogeneous graph is constructed based on academic data, and a dynamic academic network is constructed using the node embeddings. Scholar and field features are modeled on the dynamic academic network to generate scholar node vectors and field node vectors with time characteristics. Finally, the scholar node vector and the field node vector are vector inner products to obtain the scores between scholars and fields, and the prediction results for reflecting the research interests of scholars are determined based on the scores. The present invention fully considers the rich heterogeneous information of the academic network and uses it for the feature extraction of scholars, and considers the dynamic nature of entities such as scholars, fields, and technical words in the time dimension to more accurately capture the development trend of the academic community, thereby generating enhanced text semantic features and improving the accuracy of the prediction results of scholars' interests. BRIEF DESCRIPTION OF THE DRAWINGS
[0017] Figure 1 A flowchart of a preferred embodiment of a method for mining scholars' research interests based on text content and academic networks provided in an embodiment of the present invention.
[0018] Figure 2 An algorithm flow chart of a method for mining scholars' research interests based on text content and academic networks provided in an embodiment of the present invention.
[0019] Figure 3 An academic heterogeneous graph in a method for mining scholars' research interests based on text content and academic networks provided in an embodiment of the present invention.
[0020] Figure 4 A structural diagram of the graph convolutional neural network DRGCN model in the scholar research interest mining method based on text content and academic networks provided in an embodiment of the present invention.
[0021] Figure 5 A structural diagram of a system for mining scholars' research interests based on text content and academic networks provided in an embodiment of the present invention.
[0022] Figure 6 A functional block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0023] In order to make the purpose, technical solution and effect of the present invention clearer and more specific, the present invention is further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.
[0024] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations or steps, nor must they be executed in the order described. For example, some operations or steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.
[0025] It should be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.
[0026] It should be understood that, in order to clearly describe the technical solutions of the embodiments of the present invention, in the embodiments of the present invention, words such as "first" and "second" are used to distinguish between identical or similar items with substantially identical functions and effects. For example, the first control information and the second control information are only used to distinguish different control information, and their order is not limited.
[0027] Those skilled in the art can understand that the words "first", "second", etc. do not limit the quantity and execution order, and the words "first", "second", etc. do not necessarily limit the differences.
[0028] It should be further understood that the term “and / or” used in the present specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.
[0029] In order to solve the problems of the prior art, this embodiment provides a method for predicting the research interests of scholars based on text content and academic networks. Specifically, it can be extracted from the three-level field labels of the Microsoft academic data set to obtain more accurate interest tags, solving the fuzzy problem caused by directly using word frequency to extract interest tags; for text content embedding, a pre-trained language model is used to train with text multi-classification tasks to extract the text semantic features of scholars' paper results, which can improve the accuracy of scholars' research interest prediction; for the rich heterogeneous information of the academic network, a dynamic heterogeneous graph neural network is used to transfer and aggregate the node features on the heterogeneous graph, and capture the temporal features of the nodes. The contrastive learning method is used to calculate the interval loss, which improves the prediction effect of scholars' research interests. In practical application, this embodiment first generates node embedding based on paper data using a pre-trained language model, and the node embedding includes technical word node embedding, paper node embedding, scholar node embedding and field node embedding. Then, an academic heterogeneous graph is constructed based on academic data, and the node embedding is used to construct a dynamic academic network, and the scholar and field features are modeled on the dynamic academic network to generate scholar node vectors and field node vectors with time characteristics. Finally, the inner product of the scholar node vector and the field node vector is performed to obtain the score between the scholar and the field, and the prediction result reflecting the scholar's research interest is determined based on the score.
[0030] like Figure 1 As shown, this embodiment provides a method for mining scholars' research interests based on text content and academic networks. The method is executed by a server or a terminal, or can be executed collaboratively by a server and a terminal. The terminal can be an intelligent product device such as a computer. This embodiment takes the method executed by a terminal as an example for explanation, and includes the following steps: Step S100: Based on the paper data, a pre-trained language model is used to generate node embeddings, wherein the node embeddings include technical word node embeddings, paper node embeddings, scholar node embeddings, and field node embeddings.
[0031] Combination Figure 2 As shown, this embodiment can build an academic heterogeneous graph based on academic data. The academic heterogeneous graph includes four types of nodes and six types of relationship information, as shown in FIG. Figure 3 As shown in , the four types of nodes include: technical word nodes, paper nodes, scholar nodes and field nodes; the six types of relationship information include: the relationship information between scholars and papers, the relationship information between scholars, the relationship information between scholars and fields, the relationship information between papers, the relationship information between papers and technical words, and the relationship information between papers and fields.
[0032] This embodiment can use the field of academic data to screen and extract the three-level fields of paper data, and use a pre-trained language model to vectorize the nodes to generate node embeddings, so as to obtain the research interests of scholars and solve the problem of fuzzy interests. In the specific implementation, this embodiment splices the text content of the title and abstract of the scholar's paper, and uses the method of named entity recognition based on the Bert model. The Bert model is a language model that is conducive to identifying technical words from text content, and at the same time uses the word2vec model to generate technical word node embeddings. The word2vec model is a tool for calculating word vectors and a language algorithm model for calculating the vector representation of technical words to generate technical word node embeddings. This embodiment uses the SimCSE model to connect two fully connected layers, and uses the paper multi-label classification task training to generate paper node embeddings. The SimCSE model is a sentence vector model used to calculate the vector representation of the paper to generate the paper node embedding. For scholar nodes and field nodes, this embodiment maps the node embeddings of its neighboring nodes to the same semantic space and aggregates them to obtain scholar node embeddings and field node embeddings. After obtaining the above node embedding, this embodiment can use the node embedding to construct a dynamic learning network in subsequent steps to obtain text-level semantic information for predicting scholars' research interests.
[0033] Step S200, constructing an academic heterogeneous graph based on academic data, and using the node embedding to construct a dynamic academic network, modeling scholar and field characteristics on the dynamic academic network, and generating scholar node vectors and field node vectors with time characteristics.
[0034] After building an academic heterogeneous graph, this embodiment uses the node embedding to construct a dynamic academic network, embeds the node features of each node into a fusion of technical word distribution as the initial features, models scholar and field features on the dynamic academic network, optimizes through the connection prediction task on the graph neural network, and explores the research interests of scholars.
[0035] Specifically, the changes in scholars' research interests are closely related to the papers they publish, the fields to which the papers belong, and the co-authorship relationship between scholars. Based on node embedding, this embodiment models the relationship information in the academic heterogeneous graph, constructs a dynamic academic network with a collection of papers as a five-year time slice, samples the node embeddings on the dynamic academic graph network according to the edge type (i.e., the edge between scholars and fields), generates negative samples, and judges the connection relationship between scholars and research interests (i.e., fields) through connection prediction tasks, thereby mining scholars' research interests.
[0036] Specifically, in order to model a dynamic academic network, this embodiment uses a graph convolutional neural network DRGCN model as a prediction model, and constructs a dynamic academic network of length 4 by slicing the paper collection into five-year time slices to achieve information transmission on the network structure. The structure of the graph convolutional neural network DRGCN model in this embodiment is as follows: Figure 4 As shown in the figure, the training process of the graph convolutional neural network DRGCN model is as follows.
[0037] Enter: Dynamic Academic Network , T is time, node set V, edge set E, node type set N, edge type set R, initial node embedding x, feature conversion function f Output: Scholar field label L Initialize the model for epoch=1 to n for in : for r=< , , > in R, is the scholar node embedding in the dynamic academic network, u is the scholar node, is the domain node embedding in the dynamic academic network, and e is the domain node: Constructing a relationship subgraph , calculated under each relationship subgraph for v in V Node input features are transformed to the same space: z v =f(x v ) for e=<u, v is greater than in E: Message passing, feature aggregation, update node feature z v =agg(z u ,z v ) The generated dynamic node feature sequence z is extracted through the Transformer model to generate the scholar node embedding A and the domain node embedding F, and the scholar domain label L Negative sampling of scholars-fields, calculating positive and negative sample scores Calculate loss function, back propagate, and update model parameters reture L from Figure 4It can be seen that the encoder of the graph convolutional neural network DRGCN model can model the relationship information in the academic heterogeneous graph. In this embodiment, the edges between papers and fields, and the edges between scholars and fields on the academic heterogeneous graph of the last time slice of the dynamic academic network can be deleted, and vector representations of nodes under each static graph are generated on the dynamic academic network. The vector representations are uniformly input into the Transformer model, and the node temporal characteristics are learned through the multi-head attention mechanism of the Transformer model. In this embodiment, 4 attention heads are used to enhance the acquired node semantic features, and the node temporal features of neighbor nodes on the academic heterogeneous graph are aggregated to generate scholar node vectors and field node vectors with time characteristics.
[0038] Step S300: perform vector inner product on the scholar node vector and the field node vector to obtain the score between the scholar and the field, and determine the prediction result reflecting the scholar's research interest based on the score.
[0039] This embodiment makes connection predictions on the static graph under the last time slice of the dynamic academic network, deletes the edges between papers and fields, and the edges between scholars and fields on the academic heterogeneous graph, samples 100 negative samples for training on a positive sample edge, and performs vector inner product on the scholar node vector and the field node vector generated on the dynamic network to obtain the scores between scholars and fields. Then, the calculated scores of scholars and fields are sorted to obtain a score sequence. Then, this embodiment uses the scholars and fields corresponding to the scores of the front preset positions in the score sequence as the prediction results, which can reflect the fields of research interests of scholars. The graph convolutional neural network DRGCN model is tuned by contrastive learning, and Margin Loss is used as the loss function for training. After back propagation, the model parameters are updated. The updated graph convolutional neural network DRGCN model can more accurately analyze the research interests of scholars.
[0040] In summary, the present invention fully considers the rich heterogeneous information of the academic network and uses it for feature extraction of scholars. It also considers the dynamics of entities such as scholars, fields, and technical terms in the time dimension to more accurately capture the development trends of the academic community, thereby generating enhanced text semantic features and improving the accuracy of scholar interest prediction results.
[0041] Based on the above embodiments, the present invention also provides a system for mining scholars' research interests based on text content and academic networks. Figure 5As shown, the system includes: a node embedding generation module 10, a scholar interest feature extraction module 20 and a scholar research interest prediction module 30. Specifically, the node embedding generation module 10 is used to generate node embeddings based on paper data using a pre-trained language model, and the node embeddings include technical word node embeddings, paper node embeddings, scholar node embeddings and field node embeddings. The scholar interest feature extraction module 20 is used to construct an academic heterogeneous graph based on academic data, and use the node embeddings to construct a dynamic academic network, model scholar and field features on the dynamic academic network, and generate scholar node vectors and field node vectors with time characteristics. The scholar research interest prediction module 30 is used to perform vector inner product on the scholar node vector and the field node vector to obtain the score between the scholar and the field, and determine the prediction result reflecting the scholar's research interest based on the score.
[0042] The working principles of each module in the system for mining scholars' research interests based on text content and academic networks in this embodiment are the same as the principles of each step in the above method embodiment, and will not be repeated here.
[0043] Each module in the above-mentioned system for mining scholars' research interests based on text content and academic networks can be implemented in whole or in part through software, hardware and their combination. Each of the above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the corresponding operations of each of the above modules.
[0044] In some embodiments, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 6As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a method for mining the research interests of scholars based on text content and academic networks is implemented. The display unit of the computer device is used to form a visually visible image, and can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen. The input device of the computer device can be a touch layer covered on the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse, etc.
[0045] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0046] In some embodiments, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and when the processor executes the computer program, the following steps are implemented: Based on the paper data, a pre-trained language model is used to generate node embeddings, including technical word node embeddings, paper node embeddings, scholar node embeddings, and field node embeddings; Construct an academic heterogeneous graph based on academic data, and use the node embedding to construct a dynamic academic network, model scholar and field characteristics on the dynamic academic network, and generate scholar node vectors and field node vectors with time characteristics; The inner product of the scholar node vector and the field node vector is performed to obtain the score between the scholar and the field, and the prediction result reflecting the scholar's research interest is determined based on the score.
[0047] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but are not limited to this.
[0048] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0049] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A method for mining scholars' research interests based on text content and academic networks, characterized in that: The method comprises: Based on the paper data, a pre-trained language model is used to generate node embeddings, including technical word node embeddings, paper node embeddings, scholar node embeddings, and field node embeddings; Construct an academic heterogeneous graph based on academic data, and use the node embedding to construct a dynamic academic network, model scholar and field characteristics on the dynamic academic network, and generate scholar node vectors and field node vectors with time characteristics; The inner product of the scholar node vector and the field node vector is performed to obtain the score between the scholar and the field, and the prediction result reflecting the scholar's research interest is determined based on the score.
2. The method for mining scholars' research interests based on text content and academic networks according to claim 1 is characterized in that: Based on the paper data, the pre-trained language model is used to generate node embeddings, including: Based on the paper data, the text content of the scholar's paper title and abstract is spliced, and the named entity recognition method based on the Bert model is used to identify technical words from the text content. At the same time, the word2vec model is used to generate technical word node embedding; The SimCSE model is used to connect two fully connected layers and trained using the paper multi-label classification task to generate paper node embeddings; For scholar nodes and domain nodes, the node embeddings of their neighbor nodes are mapped to the same semantic space and then aggregated to obtain scholar node embedding and domain node embedding.
3. The method for mining scholars' research interests based on text content and academic networks according to claim 1 is characterized in that: The academic heterogeneous graph includes four types of nodes and six types of relationship information, wherein the four types of nodes include: technical word nodes, paper nodes, scholar nodes and field nodes; The six types of relationship information include: relationship information between scholars and papers, relationship information between scholars, relationship information between scholars and fields, relationship information between papers, relationship information between papers and technical terms, and relationship information between papers and fields.
4. The method for mining scholars' research interests based on text content and academic networks according to claim 3 is characterized in that: The method of using the node embedding to construct a dynamic academic network, modeling scholar and field characteristics on the dynamic academic network, and generating scholar node vectors and field node vectors with time characteristics includes: Based on node embedding, the relationship information in the academic heterogeneous graph is modeled, and a dynamic academic network is constructed by slicing the paper collection into five-year time slices; Delete the edges between papers and fields, and between scholars and fields on the academic heterogeneous graph of the last time slice of the dynamic academic network, generate vector representations of each node under the static graph on the dynamic academic network, and uniformly input them into the Transformer model to learn the node temporal features; The node time series features of neighbor nodes on the academic heterogeneous graph are aggregated to generate scholar node vectors and field node vectors with time characteristics.
5. The method for mining scholars' research interests based on text content and academic networks according to claim 4 is characterized in that: The predicted results based on the scores are used to reflect the research interests of scholars, including: Sort the calculated scores of scholars and fields to obtain a score sequence; The scholars and fields corresponding to the scores at the front preset positions in the score sequence are used as the prediction results.
6. The method for mining scholars' research interests based on text content and academic networks according to claim 5 is characterized in that: The dynamic academic network is a graph convolutional neural network DRGCN model.
7. The method for mining scholars' research interests based on text content and academic networks according to claim 6 is characterized in that: The method further comprises: The graph convolutional neural network DRGCN model is tuned using contrastive learning, and Margin Loss is used as the loss function for training.
8. A system for mining scholars' research interests based on text content and academic networks, characterized in that: The system comprises: A node embedding generation module, which is used to generate node embeddings based on paper data using a pre-trained language model, wherein the node embeddings include technical word node embeddings, paper node embeddings, scholar node embeddings, and field node embeddings; A scholar interest feature extraction module is used to construct an academic heterogeneous graph based on academic data, and use the node embedding to construct a dynamic academic network, model scholar and field features on the dynamic academic network, and generate scholar node vectors and field node vectors with time characteristics; The scholar research interest prediction module is used to perform vector inner product on the scholar node vector and the field node vector to obtain the score between the scholar and the field, and determine the prediction result reflecting the scholar's research interest based on the score.
9. A computer device, characterized in that: The computer device includes a memory, a processor, and a program for mining scholars' research interests based on text content and academic networks that is stored in the memory and can be run on the processor. When the processor executes the program for mining scholars' research interests based on text content and academic networks, it implements the steps of the method for mining scholars' research interests based on text content and academic networks as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a program for mining scholars' research interests based on text content and academic networks. When the program for mining scholars' research interests based on text content and academic networks is executed by a processor, the steps of the method for mining scholars' research interests based on text content and academic networks as described in any one of claims 1 to 7 are implemented.
Citation Information
Patent Citations
Scholar cooperation relationship prediction method based on interest evolution
CN111325390A
Scholar recommendation method and system based on scholar research interest knowledge graph, and medium
CN114547275A
Method, device and equipment for clustering research interest graphs of literature accomplishers and medium
CN116561605A
