Information identification method and device based on text information, equipment and storage medium

Abstract

The invention relates to the field of intelligent decision making, is applied to the field of financial science and technology businesses, and discloses an information recognition method, device and equipment based on text information.The method comprises the steps that local edge feature information of a heterogeneous graph is updated based on commodity embedding information, user embedding information and comment embedding information, and the local edge feature information of the heterogeneous graph is obtained; local information is determined based on the local edge feature information; constructing a global information graph based on the total comment information, and determining global information through the global information graph; and identifying target comment information based on the local information and the global information. Through the above mode, the system can more accurately capture the interaction mode between the user and the commodity by updating the local edge features, and the system can display the content better meeting the interest and demand of the user by identifying the target comment information, so that the user experience is improved. Through the construction of the global information graph, the system can quickly retrieve valuable information from a large number of comments, and the accuracy of identifying junk information in the financial field is improved.

Description

Technical Field

[0001] The present application relates to the field of intelligent decision-making and is applied to the field of financial technology business, and in particular to an information recognition method, device, equipment and storage medium based on text information. Background Art

[0002] For a comprehensive financial services company, while selling financial services, it is often harassed by spam comments and information, so it needs to effectively manage spam comments. Spam comments generally refer to the risk of "junk ads" appearing in the messages, such as part-time jobs, fake orders, selling diet pills, etc., and they are very confrontational. The real advertisement is not in the text itself, but on the user's homepage.

[0003] Traditional recognition methods mainly use statistical models to identify comments, from judging spam by keywords at the beginning, to using shallow models such as Naive Bayes, and then to deep neural networks such as TextCNN, using multiple algorithms to determine whether comments are spam from the text level. For example, comments are collected, manually labeled, and then the classification model is trained through the NLP (Natural Language Processing) method for identification, and the identified comments are deleted or the account is banned. However, in actual operation, it is found that spam comments change very quickly. Once a comment is deleted, a new comment strategy is immediately used, and the data must be re-labeled and a new model must be trained. Therefore, how to improve the accuracy of identifying spam in the financial field has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] The present application provides a text information-based information identification method, apparatus, device and storage medium to improve the accuracy of identifying spam information in the financial field.

[0005] In a first aspect, the present application provides an information identification method based on text information, the method comprising:

[0006] Based on the product embedding information, the user embedding information and the comment embedding information, the local edge feature information of the preset heterogeneous graph is updated, and the local information is determined based on the local edge feature information;

[0007] Constructing a global information graph based on all the comment information, and determining global information through the global information graph;

[0008] Based on the local information and the global information, target comment information in the total comment information is identified.

[0009] In a second aspect, the present application further provides an information identification device based on text information, the device comprising:

[0010] A local information determination module, used to update local edge feature information of a preset heterogeneous graph based on product embedding information, user embedding information and comment embedding information, and determine local information based on the local edge feature information;

[0011] A global information determination module, used to construct a global information graph based on the full amount of comment information, and determine the global information through the global information graph;

[0012] The target comment information identification module is used to identify the target comment information in the total comment information based on the local information and the global information.

[0013] In a third aspect, the present application also provides a computer device, comprising a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program and implement the above-mentioned text information-based information recognition method when executing the computer program.

[0014] In a fourth aspect, the present application further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the information recognition method based on text information as described above.

[0015] The present application discloses an information identification method, device, equipment and storage medium based on text information, wherein the information identification method based on text information includes updating local edge feature information of a preset heterogeneous graph based on product embedded information, user embedded information and comment embedded information, and determining local information based on the local edge feature information; constructing a global information graph based on the full amount of comment information, and determining global information through the global information graph; identifying target comment information in the full amount of comment information based on the local information and the global information. Through the above-mentioned manner, the present application can more accurately capture the interaction pattern between users and products by updating local edge features, and can display content that better meets user interests and needs by identifying target comment information. The construction of the global information graph enables the system to quickly retrieve valuable information from a large number of comments, thereby improving the accuracy of identifying spam information in the financial field. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0017] Figure 1is a schematic flow chart of an information identification method based on text information provided by the first embodiment of the present application;

[0018] Figure 2 This is a target comment mining model architecture diagram of an information identification method based on text information provided by the first embodiment of the present application;

[0019] Figure 3 is a schematic flow chart of an information identification method based on text information provided in the second embodiment of the present application;

[0020] Figure 4 is a schematic flow chart of an information identification method based on text information provided in the third embodiment of the present application;

[0021] Figure 5 A schematic block diagram of an information recognition device based on text information provided in an embodiment of the present application;

[0022] Figure 6 A schematic block diagram of the structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0023] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0024] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / 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 application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification 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 also be understood that the term “and / or” used in the specification and appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0027] The embodiments of the present application provide a method, device, equipment and storage medium for information identification based on text information. The method for information identification based on text information can be applied to a server. By updating local edge features, the system can more accurately capture the interaction pattern between users and commodities. By identifying target comment information, the system can display content that better meets user interests and needs. The construction of a global information graph enables the system to quickly retrieve valuable information from a large number of comments, thereby improving the accuracy of identifying spam information in the financial field. The server can be an independent server or a server cluster.

[0028] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0029] See also Figure 1 , Figure 1 This is a schematic flow chart of a text information-based information identification method provided by the first embodiment of the present application. The text information-based information identification method can be applied to a server, and is used to update local edge features so that the system can more accurately capture the interaction pattern between users and commodities. By identifying target comment information, the system can display content that is more in line with user interests and needs. The construction of a global information graph enables the system to quickly retrieve valuable information from a large number of comments, thereby improving the accuracy of identifying spam information in the financial field.

[0030] like Figure 1 As shown, the information recognition method based on text information specifically includes steps S10 to S30.

[0031] Step S10: based on the product embedding information, the user embedding information and the comment embedding information, updating the local edge feature information of the preset heterogeneous graph, and determining the local information based on the local edge feature information;

[0032] Step S20: construct a global information graph based on all the comment information, and determine the global information through the global information graph;

[0033] Step S30: identifying target comment information in the total comment information based on the local information and the global information.

[0034] In one embodiment, a heterogeneous graph is created, which contains product nodes, user nodes, and comment nodes. Each node contains its embedding information. For each edge in the graph, the feature information of the edge is updated according to the type of connected nodes (such as product-user, user-comment, etc.) combined with the embedding information of the node. This may involve calculating the relationship strength, similarity or other relevant metrics between two nodes. All comment information is integrated into a global information graph, which not only contains the comment content, but also may include the timestamp, user ID, product ID and other information of the comment, which is used to capture the global relationship between comments.

[0035] Based on the updated local edge feature information, analyze the structure and features of each local graph to identify local patterns or anomalies. By analyzing the global information graph, identify global patterns between comments, such as group opinion tendencies, hot topics, etc. Combine local information with global information, apply machine learning or statistical methods, and identify target comment information that meets specific criteria. For example, identify highly relevant comments, comments with specific sentiment tendencies, or suspected false comments.

[0036] like Figure 2 As shown, Figure 2 This is a diagram of the target review mining model architecture of a text information-based information identification method provided by the first embodiment of the present application. The model architecture is mainly divided into two parts. The right side of the model is a heterogeneous graph consisting of three nodes: user-product-review. It is mainly used to obtain local information. The embedding (embedding vector) information based on the product, user and current review is obtained through the GCN algorithm. The corresponding ones are (V i ,V u ,V e )

[0037] For the update of the edge, the above product V is integrated i 、User V u and current comments V e The characteristics of the edge are obtained as follows:

[0038]

[0039] For node feature updates, the update method for users and nodes is:

[0040]

[0041] The present embodiment discloses an information identification method, apparatus, device and storage medium based on text information, wherein the information identification method based on text information includes updating local edge feature information of a preset heterogeneous graph based on product embedded information, user embedded information and comment embedded information, and determining local information based on the local edge feature information; constructing a global information graph based on the full amount of comment information, and determining global information through the global information graph; identifying target comment information in the full amount of comment information based on the local information and the global information. Through the above-mentioned manner, the present application can more accurately capture the interaction pattern between users and products by updating local edge features, and can display content that better meets user interests and needs by identifying target comment information. The construction of the global information graph enables the system to quickly retrieve valuable information from a large number of comments, thereby improving the accuracy of identifying spam information in the financial field.

[0042] based on Figure 1 In the embodiment shown, in this embodiment, before step S10, the following steps are included:

[0043] Respectively acquiring the product embedded information about the product, the user embedded information about the user, and the comment embedded information about the comment;

[0044] The preset heterogeneous graph is generated by using a preset machine learning model, the product embedding information, the user embedding information and the comment embedding information, and the local edge feature information is determined.

[0045] In one embodiment, data about products, users, and comments are collected from a database or data source. The data may include product information, user information, and comment content. The collected data is preprocessed, including cleaning the data, processing missing values, removing outliers, etc., to ensure data quality.

[0046] Extract features from product, user, and comment data, which will be used to generate embedded information. For example, for products, you can extract features such as price, category, and brand; for users, you can extract features such as purchase history and browsing history; for comments, you can extract features such as text content and sentiment.

[0047] Use machine learning models (such as neural networks, random forests, etc.) to train the features of products, users, and comments to generate product embedding information, user embedding information, and comment embedding information. These embedding information can represent the high-dimensional features of products, users, and comments. Based on the product embedding information, user embedding information, and comment embedding information, a heterogeneous graph is constructed. In the graph, products, users, and comments are different types of nodes, and the relationships between them are edges. You can use a graph database (such as Neo4j) or a graph analysis tool (such as NetworkX) to create and manage heterogeneous graphs.

[0048] In a heterogeneous graph, for each edge, the feature information of the edge is determined based on the type of connected nodes (such as product-user, user-comment, etc.) combined with the embedding information of the nodes. This may involve calculating the relationship strength, similarity or other related metrics between two nodes.

[0049] The present embodiment discloses a method, device, equipment and storage medium for information identification based on text information, wherein the method for information identification based on text information includes respectively obtaining the product embedding information about the product, the user embedding information about the user and the comment embedding information about the comment; generating the preset heterogeneous graph through the preset machine learning model, the product embedding information, the user embedding information and the comment embedding information, and determining the local edge feature information; updating the local edge feature information of the preset heterogeneous graph based on the product embedding information, the user embedding information and the comment embedding information, and determining the local information based on the local edge feature information; constructing a global information graph based on the full amount of comment information, and determining the global information through the global information graph; identifying the target comment information in the full amount of comment information based on the local information and the global information. In the above manner, the present application can more accurately capture the interaction mode between the user and the product by updating the local edge features, and by identifying the target comment information, the system can display content that is more in line with the user's interests and needs. The construction of the global information graph enables the system to quickly retrieve valuable information from a large number of comments, thereby improving the accuracy of identifying spam information in the financial field.

[0050] See also Figure 3 , Figure 3 This is a schematic flow chart of a text information-based information identification method provided in the second embodiment of the present application. The text information-based information identification method can be applied to a server, and is used to update local edge features so that the system can more accurately capture the interaction pattern between users and commodities. By identifying target comment information, the system can display content that is more in line with user interests and needs. The construction of a global information graph enables the system to quickly retrieve valuable information from a large number of comments, thereby improving the accuracy of identifying spam information in the financial field.

[0051] based on Figure 1 The embodiment shown, this embodiment Figure 3 As shown, step S30 includes steps S301 to S303.

[0052] Step S301, extracting similarity edge feature information from the global information;

[0053] Step S302: Fusing the similarity edge feature information with the local edge feature information to generate fused feature information;

[0054] Step S303: classify the fused feature information through a pre-trained machine learning model to determine the target comment information.

[0055] In one embodiment, similarity edge feature information is extracted from the global information graph. This may involve calculating the relationship strength, similarity or other related metrics between two nodes. For example, methods such as cosine similarity, Jaccard similarity or Pearson correlation coefficient can be used to calculate the similarity between nodes.

[0056] The similarity edge feature information is fused with the local edge feature information to generate fused feature information. This can be achieved through simple concatenation or more complex fusion techniques (such as attention mechanism, convolutional neural network, etc.). The fused feature information is classified using a pre-trained machine learning model. This model can be a support vector machine (SVM), random forest, deep neural network, etc. The goal of the model is to identify target comment information that meets specific criteria, such as highly relevant comments, comments with specific emotional tendencies, or suspected false comments.

[0057] During the training process, you can use techniques such as cross-validation and grid search to evaluate and optimize the performance of the model. Depending on the performance of the model, you may need to adjust the model's parameters, feature selection, or training algorithm.

[0058] The present embodiment discloses a method, device, equipment and storage medium for information recognition based on text information, wherein the method for information recognition based on text information includes updating local edge feature information of a preset heterogeneous graph based on product embedded information, user embedded information and comment embedded information, and determining local information based on the local edge feature information; constructing a global information graph based on the full amount of comment information, and determining global information through the global information graph; extracting similarity edge feature information in the global information; fusing the similarity edge feature information with the local edge feature information to generate fused feature information; and classifying and processing the fused feature information through a pre-trained machine learning model to determine the target comment information. In the above manner, the present application can more accurately capture the interaction pattern between users and products by updating local edge features, and can display content that is more in line with user interests and needs by identifying target comment information. The construction of a global information graph enables the system to quickly retrieve valuable information from a large number of comments, thereby improving the accuracy of identifying spam information in the financial field.

[0059] See also Figure 4 , Figure 4This is a schematic flow chart of a text information-based information identification method provided by the third embodiment of the present application. The text information-based information identification method can be applied to a server, and is used to update local edge features so that the system can more accurately capture the interaction pattern between users and commodities. By identifying target comment information, the system can display content that is more in line with user interests and needs. The construction of a global information graph enables the system to quickly retrieve valuable information from a large number of comments, thereby improving the accuracy of identifying spam information in the financial field.

[0060] based on Figure 1 The embodiment shown, this embodiment Figure 4 As shown, step S10 includes step S101 to step S102.

[0061] Step S101, establishing a product node, a user node and a comment node according to the product embedding information, the user embedding information and the comment embedding information respectively;

[0062] Step S102: based on the association relationship between the product nodes, the user nodes and / or the comment nodes, the product nodes, the user nodes and / or the comment nodes are spliced ​​to generate current local edge feature information, and the current local edge feature information is used as the local edge feature information of the preset heterogeneous graph.

[0063] In one embodiment, product nodes, user nodes and comment nodes are created respectively according to product embedding information, user embedding information and comment embedding information. Each node contains its embedding information, which may include feature vectors, attribute values ​​and the like.

[0064] Based on the associations between product nodes, user nodes, and comment nodes, determine how they are connected. For example, product nodes and user nodes may be connected through purchase behavior, and user nodes and comment nodes may be connected through comment posting behavior. Splice product nodes, user nodes, and / or comment nodes to generate current local edge feature information. This may involve calculating the relationship strength, similarity, or other related metrics between two nodes. For example, methods such as cosine similarity, Jaccard similarity, or Pearson correlation coefficient can be used to calculate the similarity between nodes.

[0065] The generated current local edge feature information is used as the local edge feature information of the preset heterogeneous graph. This may involve updating the edge information in the graph database, or using graph analysis tools to manage and query this information.

[0066] The present embodiment discloses a method, device, equipment and storage medium for information identification based on text information. The method for information identification based on text information includes establishing a product node, a user node and a comment node according to the product embedded information, the user embedded information and the comment embedded information respectively; splicing the product node, the user node and / or the comment node based on the association relationship between the product node, the user node and / or the comment node to generate the current local edge feature information, and using the current local edge feature information as the local edge feature information of the preset heterogeneous graph; constructing a global information graph based on the full amount of comment information, and determining the global information through the global information graph; identifying the target comment information in the full amount of comment information based on the local information and the global information. Through the above-mentioned method, the present application can more accurately capture the interaction mode between the user and the product by updating the local edge features, and by identifying the target comment information, the system can display content that is more in line with the user's interests and needs. The construction of the global information graph enables the system to quickly retrieve valuable information from a large number of comments, thereby improving the accuracy of identifying spam information in the financial field.

[0067] based on Figure 1 Embodiment, in this embodiment, step S20 includes:

[0068] De-duplication processing is performed on all initial comment information to determine the full amount of comment information;

[0069] Generate sentence vector information by using a preset unsupervised algorithm and the full amount of comment information;

[0070] By using an approximate neighbor algorithm and each of the sentence vector information, the associated information in the full amount of comment information is removed to generate the global information graph.

[0071] In one embodiment, all initial comment information is deduplicated to determine the full amount of comment information. This may involve using hash functions, data structures (such as dictionaries), or database indexes to identify and remove duplicate comments. Sentence vector information is generated by using a preset unsupervised algorithm and the full amount of comment information. This may involve using natural language processing techniques (such as word embedding, sequence models, etc.) to extract semantic features of comments and convert the comments into numerical vectors. By using an approximate neighbor algorithm and each sentence vector information, the associated information in the full amount of comment information is removed to generate a global information graph. This may involve using a graph database (such as Neo4j), a graph analysis tool (such as NetworkX), or a machine learning model (such as a clustering algorithm) to build and manage a global information graph.

[0072] In the process of building the global information graph, an approximate neighbor algorithm (such as K-nearest neighbor algorithm, locality sensitive hashing, etc.) is used to identify and remove related information between reviews. This may involve calculating similarity, distance or other related metrics between vectors to identify and remove duplicate or similar reviews.

[0073] The present embodiment discloses a method, device, equipment and storage medium for information identification based on text information, wherein the method includes updating local edge feature information of a preset heterogeneous graph based on product embedding information, user embedding information and comment embedding information, and determining local information based on the local edge feature information; performing deduplication processing on all initial comment information to determine the full amount of comment information; generating sentence vector information through a preset unsupervised algorithm and each of the full amount of comment information; removing the associated information in the full amount of comment information through an approximate neighbor algorithm and each of the sentence vector information to generate the global information graph, and determining global information through the global information graph; identifying the target comment information in the full amount of comment information based on the local information and the global information. In the above manner, the present application can more accurately capture the interaction mode between users and products by updating local edge features, and can display content that is more in line with user interests and needs by identifying target comment information. The construction of the global information graph enables the system to quickly retrieve valuable information from a large number of comments, thereby improving the accuracy of identifying spam information in the financial field.

[0074] based on Figure 1 Embodiment, in this embodiment, step S20 includes:

[0075] Determine similar comment information groups in the global information graph by using a preset community detection algorithm, and determine similar information features based on the similar comment information groups;

[0076] Analyzing the similar information features to determine target information features among the similar information features;

[0077] The target information feature is converted into the global information.

[0078] In one embodiment, similar comment information groups in the global information graph are determined by a preset community detection algorithm. This may involve using graph analysis techniques (such as modularity optimization, hierarchical clustering, etc.) to identify and divide similarities or correlations between comments. Based on similar comment information groups, similar information features are determined. This may involve extracting common features of comments within the group, such as word frequency, topic model, sentiment tendency, etc.

[0079] Parse similar information features and identify target information features among similar information features. This may involve using natural language processing techniques (such as word embedding, sequence models, etc.) to extract and identify key features of reviews, such as keywords, entities, sentiment, etc.

[0080] Convert target information features into global information. This may involve converting features into numerical vectors, graph structures, or other forms for easier management and querying.

[0081] The present embodiment discloses an information identification method, apparatus, device and storage medium based on text information. The information identification method based on text information includes updating local edge feature information of a preset heterogeneous graph based on product embedding information, user embedding information and comment embedding information, and determining local information based on the local edge feature information; performing deduplication processing on all initial comment information to determine the full amount of comment information; generating sentence vector information through a preset unsupervised algorithm and each of the full amount of comment information; removing related information in the full amount of comment information through an approximate neighbor algorithm and each of the sentence vector information to generate the global information graph, determining similar comment information groups in the global information graph through a preset community detection algorithm, and determining similar information features based on the similar comment information groups; parsing the similar information features to determine target information features in the similar information features; converting the target information features into the global information; and identifying the target comment information in the full amount of comment information based on the local information and the global information. Through the above method, by updating local edge features, the system of this application can more accurately capture the interaction pattern between users and products. By identifying target comment information, the system can display content that is more in line with user interests and needs. The construction of a global information graph enables the system to quickly retrieve valuable information from a large number of comments, thereby improving the accuracy of identifying spam information in the financial field.

[0082] In a more preferred embodiment, all initial comment information is deduplicated, and before determining the full amount of comment information, the following steps are included:

[0083] Based on the current number of comments of all the initial comment information and a preset comment number threshold, invalid comment information in all the initial comment information is filtered.

[0084] See also Figure 5 , Figure 5 The embodiment of the present application provides a schematic block diagram of an information identification device based on text information, and the information identification device based on text information is used to execute the aforementioned information identification method based on text information. The information identification device based on text information can be configured on a server.

[0085] like Figure 5As shown, the information identification device 400 based on text information includes:

[0086] A local information determination module 410, configured to update local edge feature information of a preset heterogeneous graph based on product embedding information, user embedding information, and comment embedding information, and determine local information based on the local edge feature information;

[0087] A global information determination module 420, configured to construct a global information graph based on all the comment information, and determine global information through the global information graph;

[0088] The target comment information identification module 430 is used to identify the target comment information in the total comment information based on the local information and the global information.

[0089] Furthermore, the text information-based information identification device 400 includes:

[0090] An embedded information acquisition module, used to respectively acquire the product embedded information about the product, the user embedded information about the user, and the comment embedded information about the comment;

[0091] The local edge feature information determination module is used to generate the preset heterogeneous graph and determine the local edge feature information through a preset machine learning model, the product embedding information, the user embedding information and the comment embedding information.

[0092] Furthermore, the target comment information identification module 430 includes:

[0093] A similarity edge feature information extraction unit, used to extract similarity edge feature information from the global information;

[0094] a fused feature information generating unit, configured to fuse the similarity edge feature information with the local edge feature information to generate fused feature information;

[0095] The target comment information determination unit is used to classify the fused feature information through a pre-trained machine learning model to determine the target comment information.

[0096] Furthermore, the local information determination module 410 includes:

[0097] A node establishment unit, used to establish a product node, a user node and a comment node according to the product embedding information, the user embedding information and the comment embedding information respectively;

[0098] A local edge feature information determination unit is used to splice the product nodes, the user nodes and / or the comment nodes based on the association relationship between the product nodes, the user nodes and / or the comment nodes, generate current local edge feature information, and use the current local edge feature information as the local edge feature information of the preset heterogeneous graph.

[0099] Furthermore, the global information determination module 420 includes:

[0100] A deduplication processing unit, used to perform deduplication processing on all initial comment information to determine the full amount of comment information;

[0101] A sentence vector information generating unit, used to generate sentence vector information by using a preset unsupervised algorithm and the full amount of comment information;

[0102] A global information graph generating unit is used to remove the associated information in the total amount of comment information through an approximate neighbor algorithm and each of the sentence vector information to generate the global information graph.

[0103] Furthermore, the global information determination module 420 includes:

[0104] A similar information feature determination unit, configured to determine similar comment information groups in the global information graph by using a preset community detection algorithm, and determine similar information features based on the similar comment information groups;

[0105] a target information feature determination unit, configured to analyze the similar information features and determine the target information features among the similar information features;

[0106] The global information professional unit is used to convert the target information characteristics into the global information.

[0107] Furthermore, the text information-based information identification device 400 includes:

[0108] The invalid comment information filtering module is used to filter invalid comment information in the entire initial comment information based on the current comment quantity of the entire initial comment information and a preset comment quantity threshold.

[0109] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described device and each module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0110] The above-mentioned device can be implemented in the form of a computer program. Figure 6 Runs on the computer device shown.

[0111] See also Figure 6 , Figure 6 1 is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device may be a server.

[0112] See also Figure 6 The computer device includes a processor, a memory and a network interface connected through a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0113] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any information recognition method based on text information.

[0114] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0115] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any information recognition method based on text information.

[0116] The network interface is used for network communication, such as sending assigned tasks, etc. 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.

[0117] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0118] In one embodiment, the processor is used to run a computer program stored in the memory to implement the following steps:

[0119] Based on the product embedding information, the user embedding information and the comment embedding information, the local edge feature information of the preset heterogeneous graph is updated, and the local information is determined based on the local edge feature information;

[0120] Constructing a global information graph based on all the comment information, and determining global information through the global information graph;

[0121] Based on the local information and the global information, target comment information in the total comment information is identified.

[0122] In one embodiment, based on the product embedding information, the user embedding information and the comment embedding information, the local edge feature information of the preset heterogeneous graph is updated, and before the local information is determined based on the local edge feature information, it is used to implement:

[0123] Respectively acquiring the product embedded information about the product, the user embedded information about the user, and the comment embedded information about the comment;

[0124] The preset heterogeneous graph is generated by using a preset machine learning model, the product embedding information, the user embedding information and the comment embedding information, and the local edge feature information is determined.

[0125] In one embodiment, based on the local information and the global information, target review information in the full review information is identified to achieve:

[0126] Extracting similarity edge feature information from the global information;

[0127] Fusing the similarity edge feature information with the local edge feature information to generate fused feature information;

[0128] The fused feature information is classified and processed through a pre-trained machine learning model to determine the target comment information.

[0129] In one embodiment, based on the product embedding information, the user embedding information and the comment embedding information, the local edge feature information of the preset heterogeneous graph is updated to achieve:

[0130] Establishing a product node, a user node and a comment node according to the product embedding information, the user embedding information and the comment embedding information respectively;

[0131] Based on the association relationship between the product nodes, the user nodes and / or the comment nodes, the product nodes, the user nodes and / or the comment nodes are spliced ​​to generate current local edge feature information, and the current local edge feature information is used as the local edge feature information of the preset heterogeneous graph.

[0132] In one embodiment, a global information graph based on all comment information is constructed to achieve:

[0133] De-duplication processing is performed on all initial comment information to determine the full amount of comment information;

[0134] Generate sentence vector information by using a preset unsupervised algorithm and the full amount of comment information;

[0135] By using an approximate neighbor algorithm and each of the sentence vector information, the associated information in the full amount of comment information is removed to generate the global information graph.

[0136] In one embodiment, the global information is determined by the global information graph to achieve:

[0137] Determine similar comment information groups in the global information graph by using a preset community detection algorithm, and determine similar information features based on the similar comment information groups;

[0138] Analyzing the similar information features to determine target information features among the similar information features;

[0139] The target information feature is converted into the global information.

[0140] In one embodiment, all initial comment information is deduplicated before determining the full amount of comment information, to achieve:

[0141] Based on the current number of comments of all the initial comment information and a preset comment number threshold, invalid comment information in all the initial comment information is filtered.

[0142] A computer-readable storage medium is also provided in an embodiment of the present application, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and the processor executes the program instructions to implement any one of the information recognition methods based on text information provided in the embodiments of the present application.

[0143] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (Smart Medi a Card, SMC), a secure digital (Secure Digital, SD) card, a flash memory card (Flash Card), etc., equipped on the computer device.

[0144] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. An information recognition method based on text information, characterized in that: include: Based on the product embedding information, the user embedding information and the comment embedding information, the local edge feature information of the preset heterogeneous graph is updated, and the local information is determined based on the local edge feature information; Constructing a global information graph based on all the comment information, and determining global information through the global information graph; Based on the local information and the global information, target comment information in the total comment information is identified.

2. The information recognition method based on text information according to claim 1, characterized in that: The updating of the local edge feature information of the preset heterogeneous graph based on the product embedding information, the user embedding information and the comment embedding information, and before determining the local information based on the local edge feature information, includes: Respectively acquiring the product embedded information about the product, the user embedded information about the user, and the comment embedded information about the comment; The preset heterogeneous graph is generated by using a preset machine learning model, the product embedding information, the user embedding information and the comment embedding information, and the local edge feature information is determined.

3. The information recognition method based on text information according to claim 1, characterized in that: The identifying target review information in the total review information based on the local information and the global information includes: Extracting similarity edge feature information from the global information; Fusing the similarity edge feature information with the local edge feature information to generate fused feature information; The fused feature information is classified and processed through a pre-trained machine learning model to determine the target comment information.

4. The information recognition method based on text information according to claim 1, characterized in that: The updating of the local edge feature information of the preset heterogeneous graph based on the product embedding information, the user embedding information and the comment embedding information includes: Establishing a product node, a user node and a comment node according to the product embedding information, the user embedding information and the comment embedding information respectively; Based on the association relationship between the product nodes, the user nodes and / or the comment nodes, the product nodes, the user nodes and / or the comment nodes are spliced ​​to generate current local edge feature information, and the current local edge feature information is used as the local edge feature information of the preset heterogeneous graph.

5. The information recognition method based on text information according to claim 1, characterized in that: The construction of a global information graph based on all the comment information includes: De-duplication processing is performed on all initial comment information to determine the full amount of comment information; Generate sentence vector information by using a preset unsupervised algorithm and the full amount of comment information; By using an approximate neighbor algorithm and each of the sentence vector information, the associated information in the full amount of comment information is removed to generate the global information graph.

6. The information recognition method based on text information according to claim 5, characterized in that: The determining of global information through the global information graph includes: Determine similar comment information groups in the global information graph by using a preset community detection algorithm, and determine similar information features based on the similar comment information groups; Analyzing the similar information features to determine target information features among the similar information features; The target information feature is converted into the global information.

7. The information recognition method based on text information according to claim 5, characterized in that: The deduplication process of all initial comment information is performed before determining the full amount of comment information, including: Based on the current number of comments of all the initial comment information and a preset comment number threshold, invalid comment information in all the initial comment information is filtered.

8. An information recognition device based on text information, characterized in that: include: A local information determination module, used to update local edge feature information of a preset heterogeneous graph based on product embedding information, user embedding information and comment embedding information, and determine local information based on the local edge feature information; A global information determination module, used to construct a global information graph based on the full amount of comment information, and determine the global information through the global information graph; A target comment information identification module is used to identify the target comment information in the total comment information based on the local information and the global information.

9. A computer device, characterized in that: The computer device includes a memory and a processor; The memory is used to store computer programs; The processor is used to execute the computer program and implement the information recognition method based on text information as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor implements the information recognition method based on text information according to any one of claims 1 to 7.

Images