Method and apparatus for analyzing authorship of a document based on hypergraph modeling, and electronic device

Abstract

The application relates to the technical field of document analysis, in particular to a document author identity analysis method and device based on hypergraph modeling and electronic equipment, wherein the method comprises the following steps: acquiring a document data set, and constructing a hypergraph of each document file in the document data set; extracting a structural feature of the hypergraph, reasoning an author label of the document file according to the structural feature, and establishing a training data set according to the document data set, the structural feature and the author label of the document file, wherein the author label is used for identifying the author identity of the document file; training a pre-constructed analysis model by using the training data set, and analyzing the author identity of any document file by using the trained analysis model. Therefore, the method in the related art has the problems of low adaptability to different analysis scenes, poor capturing ability for the overall structure and dependency relationship of a document text, low reliability of the author identity analysis result, and inability to meet actual needs.

Description

Technical Field

[0001] This application relates to the field of document analysis technology, and in particular to a method, apparatus and electronic device for document author identity analysis based on hypergraph modeling. Background Technology

[0002] Authorship analysis, also known as author identification, identifies the author of a specific text or set of texts based on assumptions about different writing styles, preferences, vocabulary choices, grammatical patterns, and other identifiable characteristics, thus answering the question, "Who wrote this?"

[0003] In related technologies, author identification techniques can include two main steps: feature extraction and classification or modeling. Feature extraction can extract various features from the text content of a document to capture the unique characteristics and writing style of each author; classification or modeling can classify documents by author based on features, learn patterns from labeled datasets of known authors, and use them to predict the authorship of unknown texts.

[0004] However, the methods in related technologies are challenging in some scenarios (such as when the author intentionally conceals their writing style or when the amount of data is limited), and the methods focus on individual features or patterns in the text, ignoring the capture of the overall structure and dependencies, resulting in low reliability of identity analysis results and failing to meet practical needs. Summary of the Invention

[0005] This application provides a document author identity analysis method, device, electronic device, and storage medium based on hypergraph modeling to solve the problems of low adaptability of related technologies to different analysis scenarios, poor ability to capture the overall structure and dependencies of document text, resulting in low reliability of author identity analysis results and inability to meet practical needs.

[0006] The first aspect of this application provides a document author identity analysis method based on hypergraph modeling, comprising the following steps: acquiring a document dataset and constructing a hypergraph for each document file in the document dataset; extracting structural features of the hypergraph and inferring author tags for the document files based on the structural features; establishing a training dataset based on the document dataset, the structural features of the document files, and the author tags, wherein the author tags are used to identify the author identity of the document files; training a pre-constructed analysis model using the training dataset; and analyzing the author identity of any document file using the trained analysis model.

[0007] Optionally, multiple text elements are extracted from the document file; each text element is represented as a node of the hypergraph; the hyperedges of the hypergraph are determined based on the relationships and dependencies between the text elements; a hypergraph association matrix is ​​established based on the relationship between nodes and hyperedges, and the hyperedge degree of each hyperedge is determined; the hypergraph is constructed based on the nodes, the hyperedges, the hypergraph association matrix, and the hyperedge degree.

[0008] Optionally, the text elements include one or more elements among words, phrases, sentences, and topics.

[0009] Optionally, the extraction of structural features of the hypergraph includes: extracting node features of the hypergraph based on node feature capture and information related to a single node in the hypergraph; extracting hyperedge features of the hypergraph based on hyperedge feature mining and information related to a single hyperedge in the hypergraph; and extracting graph-level metrics of the hypergraph based on the overall structure and connectivity of the hypergraph.

[0010] Optionally, inferring the author tag of the document file based on the structural features includes: classifying the structural features; determining the author identity of the document file based on the similarity between the classified structural features and the target features; and generating the author tag of the document file based on the author identity.

[0011] Optionally, training the pre-built analysis model using the training dataset includes: pre-building the analysis model based on a selected machine learning model; training the analysis model using the training dataset until the performance of the analysis model reaches the desired performance, and then stopping the training.

[0012] A second aspect of this application provides a document author identity analysis device based on hypergraph modeling, comprising: an acquisition module for acquiring a document dataset and constructing a hypergraph for each document file in the document dataset; an establishment module for extracting structural features of the hypergraph, inferring author tags for the document files based on the structural features, and establishing a training dataset based on the document dataset, the structural features of the document files, and the author tags, wherein the author tags are used to identify the author identity of the document files; and a training module for training a pre-constructed analysis model using the training dataset and analyzing the author identity of any document file using the trained analysis model.

[0013] Optionally, the acquisition module is further configured to: extract multiple text elements from the document file; represent each of the multiple text elements as a node of a hypergraph; determine the hyperedges of the hypergraph based on the relationships and dependencies between the text elements; establish a hypergraph association matrix based on the relationships between nodes and hyperedges, and determine the hyperedge degree of each hyperedge; and construct the hypergraph based on the nodes, the hyperedges, the hypergraph association matrix, and the hyperedge degrees.

[0014] Optionally, the text elements include one or more elements among words, phrases, sentences, and topics.

[0015] Optionally, the establishment module is further configured to: extract node features of the hypergraph based on node feature capture and information related to a single node in the hypergraph; extract hyperedge features of the hypergraph based on hyperedge feature mining and information related to a single hyperedge in the hypergraph; and extract graph-level metrics of the hypergraph based on the overall structure and connectivity of the hypergraph.

[0016] Optionally, the establishment module is further configured to: classify the structural features; determine the author identity of the document file based on the similarity between the classified structural features and the target features; and generate author tags for the document file based on the author identity.

[0017] Optionally, the training module is further configured to: pre-build an analysis model based on a selected machine learning model; train the analysis model using the training dataset until the performance of the analysis model reaches the desired performance, and then stop training.

[0018] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the document author identification method based on hypergraph modeling as described in the above embodiments.

[0019] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the document author identification method based on hypergraph modeling as described in the above embodiments.

[0020] Therefore, this application has at least the following beneficial effects:

[0021] This application embodiment can establish a training dataset based on the document file dataset, the features of the hypergraph representation, and the author tags, to gain a more comprehensive understanding of the document structure, enrich the dataset content, and improve the comprehensiveness of the dataset. Furthermore, this application embodiment can use the training set to train an analysis model, and use the trained analysis model to analyze the author identity of any document file. It learns and summarizes from a wider range of document files, and uses mathematical expressions and training algorithms to improve prediction capabilities, thereby improving the accuracy of author identity analysis and the reliability of the analysis results, and meeting practical needs.

[0022] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0023] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0024] Figure 1 A flowchart illustrating a document author identity analysis method for hypergraph modeling according to an embodiment of this application;

[0025] Figure 2 A schematic diagram illustrating the framework of the document author identity analysis method for hypergraph modeling in this application embodiment;

[0026] Figure 3 Example diagram of a document author identity analysis device for hypergraph modeling according to embodiments of this application;

[0027] Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0028] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0029] Author identity analysis, also known as author identification, is a field of document analysis that uses computational and statistical techniques to determine the unique writing style, patterns, or characteristics of individual authors, thereby identifying the author of a specific text or a set of texts.

[0030] Specifically, author identification can be based on the assumption that authors possess different writing styles, preferences, vocabulary choices, grammatical patterns, and other identifiable characteristics. The goal of author identification is to answer the question "Who wrote this?" by analyzing the text content of one or more documents. Related technologies for author identification typically include two main steps: feature extraction and classification (modeling), as detailed below:

[0031] (1) Feature extraction: This step can extract various features from the text content of the document to capture the unique characteristics and writing style of each author. Features may include statistical measures (such as word frequency, character N-grams, syntactic patterns, punctuation usage, vocabulary richness, etc.), stylistic features (such as sentence length, use of specific phrases, etc.), and discourse annotation.

[0032] (2) Classification or Modeling: This step can be performed by using machine learning algorithms or statistical models to classify the authors of documents based on the features extracted. This allows for the learning of patterns from labeled datasets with known author identities, and the use of these patterns to predict the authorship of unknown texts. Relevant classification or modeling methods can include SVM (Support Vector Machine), Naive Bayes, random forests, or neural networks.

[0033] Author identity analysis has applications in various fields, such as forensic linguistics, plagiarism detection, literary analysis, historical research, and legal investigations. It can help identify anonymous authors, assign the identity of disputed or anonymous texts to specific authors, detect unauthorized use of someone's work, and provide insights into an author's writing style, influence, or evolution over time.

[0034] However, author identity analysis is a challenging task, especially when authors intentionally conceal their writing style or when data is limited; methods in related technologies focus on individual features or patterns in the text while neglecting to capture the overall structure and dependencies.

[0035] To address the problems mentioned in the background section, this application provides a method for analyzing document author identity in hypergraph modeling. The following description, with reference to the accompanying drawings, outlines the method, apparatus, electronic device, and storage medium for analyzing document author identity in hypergraph modeling according to embodiments of this application.

[0036] Specifically, Figure 1 This is a flowchart illustrating a document author identity analysis method for hypergraph modeling provided in an embodiment of this application.

[0037] like Figure 1As shown, the document author identity analysis method of this hypergraph modeling includes the following steps:

[0038] In step S101, the document dataset is obtained, and a hypergraph is constructed for each document file in the document dataset.

[0039] It is understandable that, such as Figure 2 As shown, in this embodiment of the application, a document dataset can be obtained first to construct a hypergraph. The document dataset may include text elements of each document file, etc. In this embodiment of the application, at least one method can be used to construct the hypergraph of each document file, without specific limitations.

[0040] As one possible approach, a hypergraph is constructed for each document file in the document dataset, including: extracting multiple text elements from the document file; representing each text element as a node of the hypergraph; determining the hyperedges of the hypergraph based on the relationships and dependencies between the text elements; establishing a hypergraph association matrix based on the relationships between nodes and hyperedges, and determining the hyperedge degree of each hyperedge; and constructing the hypergraph based on the nodes, hyperedges, hypergraph association matrix, and hyperedge degrees.

[0041] The text elements include one or more of the following: words, phrases, sentences, and topics; each text element can be represented as a node in a hypergraph, and a hypergraph can connect multiple nodes simultaneously; a hyperedge represents a higher-order relation, indicating that the connected nodes have a common characteristic or property; the hyperedge degree can represent the number of nodes connected by a hyperedge, representing the size of the hyperedge.

[0042] It is understood that, in this embodiment of the application, text elements in a document file can be used as nodes of a hypergraph to capture the relationships and dependencies between elements, and the hyperedges of the hypergraph can be determined using the relationships and dependencies between text elements; after obtaining the hyperedges of the hypergraph, a hypergraph association matrix is ​​established using the relationship between nodes and hyperedges, and the hyperedge degree is obtained at the same time; thus, this embodiment of the application can construct a hypergraph using nodes, hyperedges, hypergraph association matrix and hyperedge degree.

[0043] Specifically, in the hypergraph construction, each document, word, and topic is represented as a node. Nodes are the basic entities that capture the elements of a collection of documents in the analysis; for example, a document represents a unit of text, a word represents a vocabulary, and a topic represents a basic topical concept.

[0044] In this application embodiment, nodes can represent single words or higher-level units, such as phrases or sentences; unlike edges that connect nodes in pairs, hypergraphs can connect multiple nodes simultaneously, and this application embodiment can capture the relationships and dependencies between text elements.

[0045] To formally represent the relationship between nodes and hyperedges, embodiments of this application may use a hypergraph incidence matrix; the hypergraph incidence matrix provides a compact representation of the hypergraph structure, representing the relationship between nodes and hyperedges.

[0046] It should be noted that in this embodiment, hyperedge degree refers to the number of nodes connected by a hyperedge, representing the size of the hyperedge. A hyperedge with a higher degree may indicate a strong connection or dependency between multiple text elements. Throughout the hypergraph, the degree of hyperedges can vary to reflect different levels of relationships in the text data. Hypergraph connectivity refers to the degree of connection between nodes in the hypergraph, characterizing the extent to which nodes are interconnected through the hypergraph. Higher connectivity indicates a denser, more interdependent structure, with nodes having more relationships with other nodes. Therefore, analyzing hypergraph connectivity can provide deeper insights into the relationships and dependencies in text data.

[0047] By constructing a hypergraph from document files and capturing the relationships and dependencies between text elements, this application embodiment can comprehensively understand the file structure; hypergraph representation is beneficial for advanced analysis techniques and inference algorithms, improving the accuracy and efficiency of author identity analysis; this application embodiment can use mathematical language to represent the above hypergraph construction steps, as follows:

[0048] In this embodiment, the set of nodes can be represented as V, and the set of superedges as E. Therefore, we have: V = {v1, v2, ..., v...} n} represents the set of nodes in the hypergraph, where n is the total number of nodes; E = {e1, e2, ..., e...} m} represents the set of hyperboundaries in the hypergraph, where m is the total number of hyperboundaries.

[0049] The hypergraph incidence matrix H can be defined as an n×m binary matrix, if node v i With Hypergraph e j If there is a correlation, then H[i,j] = 1; otherwise, H[i,j] = 0.

[0050] The degree c(e) of the hypergraph j ), indicating the hypergraph e j The number of nodes included in the hypergraph; the connectivity κ(v) of the hypergraph. i ), representing node v i The number of associated hypergraphs quantifies the connectivity of a node in the hypergraph; where node v i The connectivity of a hypergraph can be represented as:

[0051] κ(v i )=Σ e H[i,j],

[0052] Where, Σ eThis indicates that for all superedges e j The summation, where H[i,j] represents the summation of node v. i and super edge e j The elements of the corresponding hypergraph incidence matrix.

[0053] In step S102, the structural features of the hypergraph are extracted, and the author tags of the document files are inferred based on the structural features. A training dataset is established based on the document dataset, the structural features of the document files, and the author tags, where the author tags are used to identify the author of the document files.

[0054] The features include point-based features, hyperedge-based features, and graph-based features.

[0055] It is understood that the embodiments of this application can extract meaningful features or characteristics from the hypergraph structure and text content to obtain relevant information for author identity analysis and infer author tags; and use the document dataset, extracted features and author tags to build a training dataset.

[0056] In this embodiment, extracting the structural features of the hypergraph includes: extracting node features of the hypergraph based on node feature capture and information related to a single node in the hypergraph; extracting hyperedge features of the hypergraph based on hyperedge feature mining and information related to a single hyperedge in the hypergraph; and extracting graph-level metrics of the hypergraph based on the overall structure and connectivity of the hypergraph.

[0057] It is understood that the extraction of hypergraph structural features in this application embodiment may include: node-based feature capture, and extraction of node features of the hypergraph based on information related to a single node in the hypergraph; hyperedge-based feature mining, and extraction of hyperedge features of the hypergraph based on information related to a single hyperedge in the hypergraph; extraction of graph-level metrics of the hypergraph based on the overall structure and connectivity of the hypergraph; wherein, this application embodiment may use at least one method to implement feature extraction, and may use various mathematical and computational techniques; for example, it may use computational centrality metrics, analyze graph connectivity, perform statistical analysis on text attributes, or apply natural language processing algorithms, etc. This application embodiment may select the specific method of feature extraction according to the characteristics of the text data and the goal of the author identity analysis task, and does not make specific limitations in this regard.

[0058] Specifically, (1) based on node feature capture, information related to a single node in the hypergraph is extracted to reveal the node features of the hypergraph:

[0059] Features of this application embodiment may include centrality and node attributes from text content; wherein, centrality may include node degree, proximity centrality or gap centrality, which quantifies the importance or influence of a node in the hypergraph structure; node attributes from text content may include word frequency, partial speech tags or sentiment scores.

[0060] In this embodiment, F_node = {f_node_1, f_node_2, ..., f_node_k} can be set to represent a node-based feature set; where k is the total number of node-based features. Each node v i Both are associated with an eigenvector x i =[x i1 ,x i2 ,...,x i k] is related, where x i j represents node v i The j-th node-based feature value.

[0061] (2) Based on the feature mining of hyperedges, extract the hyperedge features of the hypergraph by relating them to information related to individual hyperedges in the hypergraph:

[0062] Features in this application embodiment may include hyperedge weights, representing the strength or importance of the relationship between hyperedges; other hyperlink-based features can be obtained from the text content related to the hyperedge, such as average sentiment score, topic distribution, or syntactic patterns.

[0063] In this embodiment, F_graph = {f_graph_1, f_graph_2, ..., f_graph_p} can be set to represent a graph-based feature set, where p is the total number of graph-based features. The hypergraph is associated with a feature vector z = [z1, z2, ..., zp], where zm represents the value of the m-th graph-based feature.

[0064] (3) Extracting graph-level metrics from the hypergraph based on its overall structure and connectivity:

[0065] Features extracted from nodes, hyperedges, and the entire hypergraph can be combined into a comprehensive feature vector to represent the authorship analysis for each document; this feature vector derives relevant information from the hypergraph structure and text content, providing a rich representation for further analysis and modeling.

[0066] In this embodiment of the application, X can be set to [x1, x2, ..., x...]. n Let Y = [y1, y2, ..., y] denote the node-based feature matrix. m Let F = [X,Y,Z] represent the feature matrix based on hypergraph modeling, and let Z represent the graph-based feature vector; thus, the final feature vector of a document is constructed as F = [X,Y,Z].

[0067] Therefore, the embodiments of this application can extract features such as structural relationships, node attributes, and text content-related information in the hypergraph, providing rich representations of text data and making author identity analysis more accurate and comprehensive.

[0068] In this embodiment of the application, inferring the author tag of a document file based on structural features includes: classifying the structural features; determining the author identity of the document file based on the similarity between the classified structural features and the target features; and generating the author tag of the document file based on the author identity.

[0069] It is understood that the embodiments of this application can use the hypergraph structure and extracted features to perform reasoning and predict the author identity of the document, thereby realizing author tag inference based on hypergraph modeling; wherein, the embodiments of this application can first classify the structural features, and then use the classified structural features to compare the similarity between target features to determine the author identity of the document file, thereby generating author tags.

[0070] Specifically, embodiments of this application can use a hypergraph Laplacian matrix to capture the relationship between nodes and hyperboundaries; the hypergraph Laplacian matrix is ​​represented as L and defined as:

[0071] L = DH * W * H^T

[0072] Where D is a diagonal matrix representing the degree of nodes, H is the hypergraph incidence matrix, and W is the hyperedge weight matrix.

[0073] In this application embodiment, the representation learning technique based on hypergraph modeling aims to learn low-dimensional representations or embeddings of nodes in the hypergraph; these representations can capture the structural and semantic relationships between nodes and can enhance the author identity analysis process. For example, techniques such as hypergraph random walk, hypergraph convolutional neural network or hypergraph embedding algorithm can be used to learn meaningful node embeddings.

[0074] In this embodiment of the application, X can be set to [x1, x2, ..., x...]. n [x] represents the node-based feature matrix obtained from feature extraction; representation learning techniques based on hypergraph modeling transform X into a low-dimensional matrix Y = [y1, y2, ..., y]. n ], where each y i Representative node v i Learning representation.

[0075] This application embodiment can also measure the similarity between documents based on the hypergraph representation of the documents, that is, it can use some similarity measures to identify documents with similar authors; wherein, this application embodiment can use at least one similarity measurement method, such as cosine similarity, Jaccard similarity, or graph edit distance to quantify the similarity between hypergraphs, etc. Let Y i and Y j Let Y and Y represent the learned representations of files i and j, respectively. Calculate a similarity measure S(Y). i ,Y jThis determines the degree of similarity or dissimilarity between hypergraph representations.

[0076] Therefore, embodiments of this application can use hypergraph modeling-based reasoning techniques to perform accurate authorship analysis by employing structural relationships, extracted features, and learned representations, and utilize mathematical expressions and computational algorithms to analyze and predict the authorship of documents based on the underlying hypergraph structure and the information captured therein.

[0077] In step S103, a pre-built analysis model is trained using the training dataset, and the trained analysis model is used to analyze the author identity of any document file.

[0078] It is understood that the embodiments of this application may use a model integration approach, using the Xunlan dataset to train an analysis model, and using the trained analysis model to analyze the author identity of document folders; thereby, the use of the trained analysis model can improve the accuracy of the author identity analysis process and meet the needs of practical use.

[0079] Specifically, training a pre-built analytical model using a training dataset includes: pre-building an analytical model based on a selected machine learning model; and training the analytical model using the training dataset until the analytical model's performance reaches the desired performance, at which point training stops.

[0080] It is understood that the embodiments of this application can use a given feature matrix Y and corresponding author identity labels to train a classification model to predict the author identity of unseen documents; the trained model can be used to classify new documents based on hypergraph representation.

[0081] In this embodiment, X_train = [X_train1, X_train2, ..., X_train] can be set to [X_train1, X_train2, ..., X_train]. k ] represents the feature matrix of the training dataset, where each X_train i Let Y_train = [Y_train1, Y_train2, ..., Y_train...] represent the feature vector of document i; k ] represents the corresponding author tag for the training document.

[0082] After obtaining the hypergraph structure and node representation, at least one classification algorithm can be used for author identification analysis in this embodiment of the application. The classification algorithm in this embodiment of the application can use extracted features, learned representations and other relevant information to predict the author identity of the document. For example, support vector machines, random forests, neural networks or GNNs (Graph Neural Networks) can be used. The choice of model in this embodiment of the application depends on factors such as the nature of the data, the complexity of the relationships and the desired performance.

[0083] Let f(X) denote the chosen machine learning model that maps the feature vector X to the predicted author identity label; the model f(X) can be trained on the labeled training dataset (X_train, Y_train) to learn the relationship between features and author labels; where the model f(X) is trained by minimizing a suitable loss function, such as mean squared error or cross-entropy loss.

[0084] It should be noted that the machine learning model selected in this application embodiment is trained using a labeled training dataset, so the model parameters can be optimized to minimize the prediction error; this application embodiment can use at least one training algorithm to optimize the parameters, such as gradient descent or backpropagation; thus, this application embodiment can adjust the model parameters to minimize the difference between the predicted author identity label f(X) and the true label Y_train.

[0085] After optimizing the model parameters, the embodiments of this application can use evaluation metrics such as accuracy, precision, recall, or F1 score to evaluate the trained model and assess its performance. After the model is trained and evaluated, it can predict the author identity of unseen documents. Feature extraction and hypergraph-based inference techniques are applied to new documents, and the learned model is used to predict its author label. That is, given a new document with a feature vector X_test, the trained model f(X) predicts its author label Y_test = f(X_test), thus completing the analysis of the author identity of any document file.

[0086] In summary, the document author identification method based on hypergraph modeling proposed in this application can establish a training dataset based on the document file dataset, the features of the hypergraph representation, and author tags, thereby gaining a more comprehensive understanding of the document structure, enriching the dataset content, and improving the dataset's comprehensiveness. Furthermore, this application embodiment can use the training set to train an analysis model, and use the trained analysis model to analyze the author identification of any document file. It learns and summarizes from a wider range of document files, and uses mathematical expressions and training algorithms to improve prediction capabilities, thereby improving the accuracy of author identification analysis and the reliability of the analysis results, meeting practical needs.

[0087] Next, referring to the accompanying drawings, a document author identification analysis device for hypergraph modeling proposed according to an embodiment of this application is described.

[0088] Figure 3 This is a block diagram of a document author identity analysis device for hypergraph modeling according to an embodiment of this application.

[0089] like Figure 3As shown, the document author identity analysis device 10 for hypergraph modeling includes: an acquisition module 100, an establishment module 200, and a training module 300.

[0090] The acquisition module 100 is used to acquire the document dataset and construct a hypergraph for each document file in the document dataset; the construction module 200 is used to extract the structural features of the hypergraph, infer the author tags of the document files based on the structural features, and construct a training dataset based on the document dataset, the structural features of the document files, and the author tags, where the author tags are used to identify the author of the document file; the training module 300 is used to train a pre-constructed analysis model using the training dataset, and use the trained analysis model to analyze the author of any document file.

[0091] Optionally, the acquisition module 100 is further configured to: extract multiple text elements from a document file; represent each text element as a node of a hypergraph; determine the hyperedges of the hypergraph based on the relationships and dependencies between the text elements; establish a hypergraph association matrix based on the relationships between nodes and hyperedges, and determine the hyperedge degree of each hyperedge; and construct a hypergraph based on nodes, hyperedges, the hypergraph association matrix, and hyperedge degrees.

[0092] Optionally, text elements include one or more of the following: words, phrases, sentences, and themes.

[0093] Optionally, the establishment module 200 is further used for: extracting node features of the hypergraph based on node feature capture and information related to a single node in the hypergraph; extracting hyperedge features of the hypergraph based on hyperedge feature mining and information related to a single hyperedge in the hypergraph; and extracting graph-level metrics of the hypergraph based on the overall structure and connectivity of the hypergraph.

[0094] Optionally, the module 200 is further used to: classify structural features; determine the author identity of the document file based on the similarity between the classified structural features and the target features; and generate author tags for the document file based on the author identity.

[0095] Optionally, the training module 300 is further configured to: pre-build an analysis model based on a selected machine learning model; train the analysis model using a training dataset until the performance of the analysis model reaches the desired performance, and then stop training.

[0096] It should be noted that the explanation of the above-described embodiments of the document author identity analysis method for hypergraph modeling also applies to the document author identity analysis device for hypergraph modeling in this embodiment, and will not be repeated here.

[0097] The document author identity analysis device based on hypergraph modeling proposed in this application can establish a training dataset based on the document file dataset, the features of the hypergraph representation, and the author tags, thereby gaining a more comprehensive understanding of the document structure, enriching the dataset content, and improving the comprehensiveness of the dataset. Furthermore, this application embodiment can use the training set to train the analysis model, and use the trained analysis model to analyze the author identity of any document file. It learns and summarizes from a wider range of document files, and uses mathematical expressions and training algorithms to improve the prediction ability, thereby improving the accuracy of author identity analysis and the reliability of the analysis results, and meeting practical needs.

[0098] Figure 4 A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include:

[0099] The memory 401, the processor 402, and the computer program stored on the memory 401 and capable of running on the processor 402.

[0100] When the processor 402 executes the program, it implements the document author identity analysis method for hypergraph modeling provided in the above embodiments.

[0101] Furthermore, electronic devices also include:

[0102] Communication interface 403 is used for communication between memory 401 and processor 402.

[0103] The memory 401 is used to store computer programs that can run on the processor 402.

[0104] The memory 401 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0105] If the memory 401, processor 402, and communication interface 403 are implemented independently, then the communication interface 403, memory 401, and processor 402 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0106] Optionally, in a specific implementation, if the memory 401, processor 402, and communication interface 403 are integrated on a single chip, then the memory 401, processor 402, and communication interface 403 can communicate with each other through an internal interface.

[0107] Processor 402 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of this application.

[0108] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the document author identification analysis method for hypergraph modeling as described above.

[0109] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0110] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0111] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0112] It should be understood that the various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0113] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0114] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A document author identity analysis method based on hypergraph modeling, characterized in that, Includes the following steps: Obtain the document dataset and construct a hypergraph for each document file in the dataset; Extract the structural features of the hypergraph, infer the author tags of the document files based on the structural features, and establish a training dataset based on the document dataset, the structural features of the document files, and the author tags, wherein the author tags are used to identify the author of the document files; The pre-built analysis model is trained using the training dataset, and the trained analysis model is used to analyze the author identity of any document file. The construction of the hypergraph for each document file in the document dataset includes: Extract multiple text elements from the document file; Each of the multiple text elements is represented as a node of the hypergraph; The hyperedges of the hypergraph are determined based on the relationships and dependencies between the text elements in the plurality of text elements; Establish a hypergraph association matrix based on the relationship between nodes and hyperedges, and determine the hyperedge degree of each hyperedge; The hypergraph is constructed based on the nodes, the hyperedges, the hypergraph association matrix, and the hyperedge degree.

2. The document author identity analysis method based on hypergraph modeling according to claim 1, characterized in that, The text elements include one or more of the following: words, phrases, sentences, and themes.

3. The document author identity analysis method based on hypergraph modeling according to claim 1, characterized in that, The Extracting the structural features of the hypergraph includes: Based on node feature capture, information related to individual nodes in the hypergraph is extracted to identify node features of the hypergraph. Based on the feature mining of hyperedges, the hyperedge features of the hypergraph are extracted by extracting information related to individual hyperedges in the hypergraph. Measurement of the graph layer of a hypergraph is extracted based on the overall structure and connectivity of the hypergraph.

4. The document author identity analysis method based on hypergraph modeling according to claim 1, characterized in that, The Inferring the author tag of the document file based on the structural features includes: The structural features are classified; The author identity of the document file is determined based on the similarity between the classified structural features and the target features, and an author tag for the document file is generated based on the author identity.

5. The document author identity analysis method based on hypergraph modeling according to claim 1, characterized in that, The Training a pre-built analytical model using the training dataset includes: Analytical models are pre-built based on the selected machine learning model; The analysis model is trained using the training dataset until its performance reaches the desired level, at which point training stops.

6. A document author identity analysis device based on hypergraph modeling, characterized in that, include: The acquisition module is used to acquire the document dataset and construct a hypergraph for each document file in the document dataset; A module is established to extract the structural features of the hypergraph, infer the author tags of the document files based on the structural features, and establish a training dataset based on the document dataset, the structural features of the document files, and the author tags, wherein the author tags are used to identify the author of the document files; The training module is used to train a pre-built analysis model using the training dataset, and to analyze the author identity of any document file using the trained analysis model. The acquisition module is further used for: Extract multiple text elements from the document file; Each of the multiple text elements is represented as a node of the hypergraph; The hyperedges of the hypergraph are determined based on the relationships and dependencies between the text elements in the plurality of text elements; Establish a hypergraph association matrix based on the relationship between nodes and hyperedges, and determine the hyperedge degree of each hyperedge; The hypergraph is constructed based on the nodes, the hyperedges, the hypergraph association matrix, and the hyperedge degree.

7. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, the processor executing the program to implement the document author identification method based on hypergraph modeling as described in any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the document author identification method based on hypergraph modeling as described in any one of claims 1-5.

Images