Social capital index determination method, computer equipment and storage medium

Abstract

The invention relates to the technical field of data processing, and discloses a social capital index determination method, computer equipment and a storage medium. The method comprises the following steps: performing feature extraction on original data to obtain a relationship and a cognitive feature set; determining a key feature set according to the relationship, the cognitive feature set and the dimension value of each text under each language dimension; constructing an interactive network and a resource network according to the relationship and the cognitive feature set; constructing a relation network according to the key feature set, the interactive network and the resource network; inputting a corpus data set determined according to the original data into a triple knowledge extraction model to obtain an entity relationship extraction result; constructing a concept network according to an entity relationship extraction result; and performing index extraction on the four networks based on an evaluation index algorithm determined by a value capital dimension to obtain an index extraction result. According to the method, the social capital indexes in the open network community can be comprehensively extracted, and then the social capital is dynamically and comprehensively measured.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method for determining social capital indicators, a computer device, and a storage medium. Background Art

[0002] Social capital describes the resources and advantages obtained by individuals or groups through social relationship networks. Social capital is usually regarded as a resource that can promote cooperation and efficiency, and it includes elements such as trust, networks, and norms among individuals. With the development of the Internet, social interactions in the network also affect the changes in the social capital of individuals or groups. In open network communities, social capital is regarded as a resource embedded in the network. Currently, the use of scale measurement and simple network indicators for measurement is not comprehensive enough and has low accuracy. Summary of the Invention

[0003] Therefore, the technical problem to be solved by the present invention is to overcome the problem that the use of scale measurement and simple network indicators in related technologies for measurement is not comprehensive enough and has low accuracy.

[0004] To solve the above technical problem, a method for determining social capital indicators provided by the present invention is applied to a constructed network structure, and the network structure includes: an interaction network, a resource network, a relationship network, and a concept network. The method for determining social capital indicators includes:

[0005] Obtain the original data of the target community; the original data includes multiple texts; the data categories corresponding to the original data include: community interaction content, community interaction behavior, and community resource content;

[0006] Extract features from the original data to obtain a relationship and cognitive feature set;

[0007] Analyze the original data to determine the dimension value corresponding to each language style dimension of each text in a preset number of language style dimensions;

[0008] Determine a key feature set according to the relationship and cognitive feature set and the dimension value corresponding to each language style dimension of each text in a preset number of language style dimensions;

[0009] Construct an interaction network and a resource network according to the relationship and cognitive feature set;

[0010] Construct a relationship network according to the key feature set, the interaction network, and the resource network;

[0011] Determine a corpus dataset according to the original data;

[0012] Input the corpus dataset into a pre-constructed triple knowledge extraction model to extract knowledge entities and relationships, obtaining an entity relationship extraction result; the relationship represents the relationship between knowledge entities.

[0013] Construct a concept network based on the entity relationship extraction result.

[0014] Extract indicators from the interaction network, the resource network, the relationship network, and the concept network based on the evaluation index algorithm determined by the value capital dimension, obtaining an indicator extraction result; the value capital dimension includes: information acquisition dimension, social support dimension, knowledge production dimension, and social identity dimension.

[0015] In an alternative implementation, the obtaining of the original data of the target community includes:

[0016] Obtain the interaction messages in the target community.

[0017] Divide the interaction messages into community interaction content, community interaction behavior, and community resource content to obtain the original data.

[0018] In an alternative implementation, the extracting of features from the original data to obtain a relationship and cognitive feature set includes:

[0019] Analyze the original data by means of cognitive network analysis to obtain interaction texts, perform relationship coding and cognitive coding on the interaction texts to obtain a coding result, and perform feature extraction on the coding result to obtain a relationship and cognitive feature set; the relationship and cognitive feature set includes relationship features and cognitive features; the relationship coding is determined according to the relationship type to which the relationship features belong; the cognitive coding is determined according to the cognitive type to which the cognitive features belong.

[0020] In an alternative implementation, the analyzing of the original data to determine the dimension value corresponding to each text in each of a preset number of stylistic dimensions includes:

[0021] Pre-construct, by means of multi-dimensional analysis, a language feature for identifying a preset number of stylistic dimensions and the corresponding language feature for each stylistic dimension; the language feature is a language feature for representing relationship intimacy and social cognitive level.

[0022] Label the original data according to the language feature to obtain a labeling result; count the frequency of occurrence of the language feature corresponding to each stylistic dimension in each text according to the labeling result.

[0023] Determine the dimension value corresponding to each text in each of a preset number of stylistic dimensions according to the frequency of occurrence of the language feature corresponding to each stylistic dimension in each text.

[0024] Among them, when the preset quantity is 3, the stylistic dimensions include: the first stylistic dimension: the subjective stance dimension or the objective stance dimension; the second stylistic dimension: the general narrative dimension or the professional description dimension; the third stylistic dimension: the prominent emotional attitude dimension or the prominent logical relationship dimension.

[0025] In an alternative implementation, before inputting the corpus data set into a pre-constructed triple knowledge extraction model to extract knowledge entities and relationships and obtaining the entity relationship extraction result, it includes:

[0026] Annotating the historical interaction data according to the predefined knowledge entities and relationships to obtain historical sample data;

[0027] Training a pre-trained model according to the historical sample data, and obtaining a triple knowledge extraction model by iteratively optimizing the entity recognition and relationship extraction rules.

[0028] In an alternative implementation, constructing a concept network according to the entity relationship extraction result includes:

[0029] Taking the knowledge entities in the entity relationship extraction result as network nodes;

[0030] Taking the relationships between the knowledge entities in the entity relationship extraction result as connecting edges;

[0031] Determining the weights of the corresponding connecting edges according to the frequency of occurrence of the relationships between the knowledge entities in the entity relationship extraction result.

[0032] In an alternative implementation, after obtaining the index extraction result, it further includes:

[0033] Performing criterion validity verification and optimization on the index extraction result based on the measurement result of the social capital scale;

[0034] Performing secondary optimization on the index extraction result after criterion validity verification and optimization by means of exploratory factor analysis to obtain multiple target indexes;

[0035] Determining the information entropy and weights corresponding to each target index according to the entropy weight method;

[0036] Determining the comprehensive evaluation result of social capital according to the information entropy and weights corresponding to each target index.

[0037] In an alternative implementation, the method further includes:

[0038] Constructing a social capital dynamic evaluation model based on a temporal attention convolutional neural network; the social capital dynamic evaluation model is used to predict social capital.

[0039] In a second aspect, the present invention provides a computer device, comprising: a memory and a processor, the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for determining social capital indicators of the first aspect or any corresponding embodiment thereof by executing the computer instructions.

[0040] In a third aspect, the present invention provides a computer-readable storage medium, on which a single computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for determining social capital indicators of the first aspect or any corresponding embodiment thereof.

[0041] In a fourth aspect, the present invention provides a computer program product, comprising computer instructions for causing a computer to execute the method for determining social capital indicators of the first aspect or any corresponding embodiment thereof.

[0042] The technical solution provided by the present invention has the following technical effects:

[0043] The technical solution of the embodiment of the present invention determines the dimension value corresponding to each text in a preset number of language dimensions by analyzing the original data, and further screens the relationship and cognitive feature sets according to the dimension values ​​to obtain the key feature set, thereby improving the accuracy of the relationship feature and cognitive feature extraction.

[0044] By analyzing multiple texts in the original data in different language dimensions and extracting relational and cognitive features, we can comprehensively and deeply explore the social capital information contained in the community, avoid the limitations of single-dimensional analysis, and make the understanding of social capital more comprehensive and accurate.

[0045] Based on the extracted feature sets, interaction networks, resource networks and relationship networks are constructed, and the relationship networks are further improved in combination with key feature sets, which can clearly show the complex relationship structure between individuals and resources, and between individuals in the community.

[0046] Using the triple knowledge extraction model to extract knowledge entities and relationships from the corpus dataset to construct a conceptual network can quickly sort out the knowledge system and cognitive architecture within the community, providing strong support for further analysis of social capital related to knowledge production and dissemination.

[0047] The evaluation index algorithm determined based on the value capital dimension extracts indicators from each network, covering key dimensions such as information acquisition, social support, knowledge production and social identity, and can conduct a comprehensive quantitative assessment of social capital from multiple important angles. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in related technologies, the following will briefly introduce the drawings required for use in the description of the specific embodiments or related technologies. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0049] Figure 1 It is a schematic flowchart of the method for determining the social capital index in an embodiment of the present invention;

[0050] Figure 2 It is a schematic diagram of extracting relationship or cognitive features in interactive text based on TextCNN in an embodiment of the present invention;

[0051] Figure 3 It is a schematic diagram of the triple knowledge extraction method based on BERT pre-training in an embodiment of the present invention;

[0052] Figure 4 It is a schematic diagram of the concept network similarity calculation scheme in an embodiment of the present invention;

[0053] Figure 5 It is a schematic diagram of the STGACN model framework for dynamically predicting social capital in an embodiment of the present invention;

[0054] Figure 6 It is a schematic diagram of the technical architecture for dynamically calculating social capital indicators in an embodiment of the present invention;

[0055] Figure 7 It is a schematic diagram of the hardware structure of a computer device in an embodiment of the present invention. Specific Embodiments

[0056] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0057] Social capital describes the resources and advantages obtained by individuals or groups through social relationship networks. Social capital is usually regarded as a resource that can promote cooperation and efficiency, and it includes elements such as trust, networks, and norms among individuals.

[0058] With the development of the Internet, social interactions in the network also affect the changes in the social capital of individuals or groups. In the communities of open networks, social capital is regarded as a resource embedded in the network. Calculating social capital indicators and dynamically evaluating them in open network communities has always been a difficult problem. Most related solutions use scale measurement and indicators of simple networks for measurement. Although the scale measurement method can measure multiple dimensions of social capital from multiple dimensions, it cannot achieve automated dynamic prediction. While using simple network indicators can achieve dynamic automated evaluation, it cannot comprehensively estimate the social capital of learners.

[0059] In view of this, the present invention proposes a social capital indicator calculation and dynamic evaluation technology, aiming to dynamically and comprehensively measure the social capital in open network communities.

[0060] Embodiments of the present invention provide a method for determining social capital indicators, a computer device, and a storage medium to solve the problems in related technologies.

[0061] According to an embodiment of the present invention, an embodiment of a method for determining social capital indicators is provided. It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer device such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0062] Figure 1 It is a flowchart of the method for determining social capital indicators according to an embodiment of the present invention.

[0063] As Figure 1 shown, an embodiment of the present invention provides a method for determining social capital indicators, which is applied to the constructed network structure. The network structure includes: an interaction network, a resource network, a relationship network, and a concept network. The personal network structure and the community network structure can be constructed according to the determination methods of the interaction network, the resource network, the relationship network, and the concept network in the following social capital indicator determination solution. The constructed network structure includes a personal network structure and a community network structure. The community network structure includes the personal network structure.

[0064] The method for determining the social capital indicators includes:

[0065] S101: Obtain the original data of the target community.

[0066] In this embodiment, the original data includes multiple texts, and the relevant content of a post can be used as one text. The target community is a community of an open network. The data categories corresponding to the original data include: community interaction content, community interaction behavior, and community resource content.

[0067] In this embodiment, the specific steps of S101 for obtaining the original data of the target community are as follows:

[0068] S1011: Obtain the interaction messages in the target community. In an open learning online community such as a video network, there are interaction behaviors during the learning process of learners. For example, watching videos, browsing posts, posting comments, etc. will generate interaction messages.

[0069] S1012: Classify the interaction messages into community interaction content, community interaction behaviors, and community resource content to obtain the original data. Among them, community interaction content includes, but is not limited to, posts and comments posted by learners, and community interaction behaviors include, but are not limited to, likes and follows of learners. Community resource content includes, but is not limited to, blog posts, articles learned and browsed by learners, and downloaded file resources.

[0070] S102: Extract features from the original data to obtain a set of relationship and cognitive features.

[0071] In this embodiment, S102 is the definition and extraction of social capital features. The specific steps of extracting features from the original data in S102 to obtain a set of relationship and cognitive features are as follows:

[0072] Analyze the original data by means of cognitive network analysis to obtain interaction texts, perform relationship coding and cognitive coding on the interaction texts to obtain coding results, and extract features from the coding results to obtain a set of relationship and cognitive features.

[0073] In this embodiment, the set of relationship and cognitive features includes relationship features and cognitive features. Relationship coding is determined according to the relationship type to which the relationship features belong. Cognitive coding is determined according to the cognitive type to which the cognitive features belong. The relationship types include four types: directional selection, surface contact, intimate reciprocity, and stable compatibility. The cognitive types are divided into five types in the order of gradually increasing knowledge construction level: sharing and comparison, divergence and exploration, negotiation and co-construction, inspection and correction, and consensus and application. Relationship features include language and behavior, and cognitive features include language and keywords.

[0074] As an example, the method of cognitive network analysis is specifically virtual ethnography.

[0075] In this embodiment, the method of virtual ethnography can be used to extract the relationship and cognitive features of different value-class social capitals.

[0076] The specific scheme for analyzing the original data based on virtual ethnography includes:

[0077] Use means such as interviews and participatory observations to deeply analyze the original data, so as to obtain rich interaction texts.

[0078] Based on the obtained interactive texts, the interactive behaviors and contents in the community are classified and coded. In this process, the relevant behaviors, languages ​​and keyword features are accurately determined through continuous comparison and refinement.

[0079] In the relationship coding stage, the interpersonal relationship development stages are first divided, and then the framework is optimized with the help of qualitative research. Specifically, the relationship types are subdivided into four levels: directional selection, surface contact, intimate reciprocity, and stable compatibility, taking into account factors such as the degree of emotional involvement of the interaction content and the frequency of the interaction behavior. For example, in an entrepreneurial incubation community, new members may initially be in the directional selection stage and establish initial contact with mentors by participating in specific entrepreneurial training courses. As communication increases, they enter the surface contact stage, such as occasional interaction in some entrepreneurial exchange activities. If the cooperation project goes smoothly, the two parties will deepen resource sharing and experience exchange, and then develop to the intimate reciprocity stage. After long-term and stable cooperation, the relationship reaches a stable and compatible level, and supports each other in multiple entrepreneurial projects.

[0080] Cognitive coding is divided into five types according to the order of knowledge construction level from low to high: sharing and comparison, disagreement and exploration, consultation and co-construction, verification and revision, and consensus and application. For example, in an academic research community, members share their research results at a seminar, which belongs to the sharing and comparison stage. Proposing different opinions on a certain research point of view and launching discussions is the disagreement and exploration stage. Jointly formulating research plans and dividing labor and cooperating are in the consultation and co-construction stage. Evaluating and improving the interim results during the research process belongs to the verification and revision stage. Finally reaching a consensus on the research conclusions and promoting and applying them within the community is the consensus and application stage.

[0081] Through the above-mentioned precise definition and feature extraction of relationship and cognitive characteristics, a data foundation is provided for the subsequent construction of interaction networks, resource networks, relationship networks and concept networks. In particular, it greatly improves the accuracy of relationship networks in revealing interpersonal relationships and concept networks in reflecting cognition, which is conducive to the comprehensive extraction of indicators.

[0082] S103: Analyze the original data to determine the dimension value corresponding to each text in each of the preset number of language dimensions.

[0083] In this embodiment, the above S103 analyzes the original data to determine the dimension value corresponding to each text in each of the preset number of language dimensions, specifically including:

[0084] S1031: Pre-constructing, by means of multi-dimensional analysis, a method for identifying a preset number of language dimensions and language features corresponding to each language dimension.

[0085] In this embodiment, the language features are language features used to represent relationship intimacy and social cognition level.

[0086] In this embodiment, in order to explore the linguistic features in the interactive text that reveal the relationship intimacy and the level of social cognitive construction, the present invention introduces a multi-dimensional analysis method, or a multi-feature analysis method, in cognitive network analysis (such as virtual ethnography analysis) to construct the linguistic features that identify multiple stylistic dimensions. In this embodiment, the linguistic features have been verified through 17 styles. According to the research context, as an example, when the preset quantity is 3, as shown in Table 1, the stylistic dimensions include: the first stylistic dimension: the subjective stance dimension or the objective stance dimension ("interaction highlighting personal stance VS objective and accurate information"), the second stylistic dimension: the general narrative dimension or the professional description dimension ("general narrative VS professional description"), and the third stylistic dimension: the dimension highlighting emotional attitude or the dimension highlighting logical relationship ("highlighting emotional attitude VS highlighting logical relationship").

[0087] Table 1 Example of multi-dimensional stylistic features for mining relationship and cognitive capital features

[0088]

[0089] S1032: Annotate the original data according to the linguistic features to obtain the annotation result. According to the annotation result, count the frequency of occurrence of the linguistic features corresponding to each stylistic dimension in each text.

[0090] As an example, corresponding labels can be set for each feature in the linguistic features to distinguish different situations under the stylistic dimension. For example, the label for the adverb corresponding to the subjective stance dimension is A1, and the label for the adverb corresponding to the objective stance dimension is A2. The number of labels of the linguistic features corresponding to the subjective stance dimension in the first stylistic dimension can be counted as the frequency. For example, in a text A, the label of the linguistic feature corresponding to the subjective stance dimension in the first stylistic dimension is only the adverb label A1, then the number of label A1 is used as the frequency of occurrence of the linguistic features corresponding to the subjective stance dimension in the first stylistic dimension in text A.

[0091] As another example, different labels can be set for different dimension situations in each stylistic dimension. For example, in the first stylistic dimension, for the subjective stance dimension: when there are adverbs: M, first-person pronouns, and directional verbs: V, etc., they are uniformly labeled as label C, and the linguistic features corresponding to the subjective stance dimension are uniformly labeled as C. The number of label C in the text is counted as the frequency of occurrence of the linguistic features corresponding to the subjective stance dimension in the first stylistic dimension in the text.

[0092] S1033: Determine the dimension value corresponding to each stylistic dimension of each text in each of the preset number of stylistic dimensions according to the frequency of occurrence of the linguistic features corresponding to each stylistic dimension in each text.

[0093] In this embodiment, the frequency of occurrence of the language features corresponding to each language dimension in each text is standardized in units of thousands of words. For example, the standardized frequency of "I" is (50 / 5000)×1000=10 times / thousand words, and the standardized frequency of "professional term W" is (30 / 5000)×1000=6 times / thousand words. Through this standardization operation, the language features in texts of different lengths are comparable.

[0094] The Z value of each language dimension can be calculated according to the frequency of occurrence of the language features corresponding to each language dimension in each text, and the dimension value corresponding to each language dimension in a preset number of language dimensions of each text can be calculated based on the Z value through dimensionality reduction.

[0095] For example, if the frequency of a language feature is a, its mean is b, and its standard deviation is c, then Z = (ab) / c.

[0096] In this embodiment, for the first type of language dimension: the dimension value corresponding to the subjective stance dimension or the objective stance dimension ("interaction that highlights personal stance VS objective and accurate information"), the dimension value reflects the extent to which the text tends to highlight the interactive expression of personal stance or the transmission of objective and accurate information. If the frequency of occurrence of relevant language features such as adverbs and first-person pronouns in the text is high and after standardization and Z-value calculation, the dimension value on this dimension tends to highlight the end of interaction of personal stance. On the contrary, if nouns, distinguishing words and other objective and accurate information-related features are dominant, the dimension value is close to the objective and accurate information end.

[0097] For the second language dimension: the dimension value corresponding to the general narrative dimension or the professional description dimension ("general narrative VS professional description"), this dimension value reflects whether the narrative style of the text is closer to general narrative or professional description. When there are more tense auxiliary words such as "着" and "了" and state words, the dimension value tends to be general narrative. When features such as professional terms, idioms and longer average word length are prominent, the dimension value tends to be professional description, which intuitively presents the position of the text in terms of knowledge depth and expression style.

[0098] For the third language dimension: the dimension value corresponding to the dimension of highlighting the emotional attitude dimension or the dimension of highlighting the logical relationship ("highlighting emotional attitude VS highlighting logical relationship"), this dimension value indicates whether the text focuses on expressing emotional attitude or building logical relationships. The frequent appearance of emotional-related language features such as modal particles and second-person pronouns will make the dimension value close to the emotional attitude end. The presence of many logical association features such as conjunctions and interrogative pronouns will make the dimension value tend to the logical relationship end, thereby accurately measuring the characteristics of the text in this dimension.

[0099] S104: Determine a key feature set according to the relationship and cognitive feature set and the dimension value corresponding to each language dimension of a preset number of language dimensions for each text.

[0100] In this embodiment, the key feature set can be determined based on the TextCNN algorithm according to the relationship and cognitive feature set and the dimension values corresponding to each text in a preset number of language style dimensions.

[0101] In this embodiment, the intervention of the TextCNN algorithm on the basis of virtual ethnography or cognitive network analysis method further strengthens the extraction of cognitive features in interactive texts.

[0102] In this embodiment, to support the automatic extraction of language features and keyword features related to relationships and cognition in interactive texts, and to realize the automatic construction and calculation of relevant networks and indicators, the present invention intends to extract relationship types or cognitive features from learners' interactive texts by using the TextCNN algorithm on the basis of defining relevant features in the cognitive network analysis method. These features will support the empowerment in the relationship network and the calculation of relevant capital indicators. The extraction process of relationship and cognitive capital features based on the TextCNN algorithm is as Figure 2 shown.

[0103] Determining the key feature set based on the TextCNN algorithm according to the relationship and cognitive feature set and the language style dimension values of each text has many important technical effects:

[0104] The TextCNN algorithm itself has strong and fast ability to extract shallow features of texts, has obvious advantages in processing a large number of interactive texts, is widely used in the field of short text classification, and is suitable for the feature extraction requirements of short interactive texts in the present invention.

[0105] Combining the relationship and cognitive feature set and the language style dimension values can more accurately focus on the features that are important for social capital analysis. For example, in the texts of academic communication communities, relationship features such as "common research direction" and "citing the same literature" that are closely related to research cooperation, and cognitive features such as "discussion on theoretical innovation" and "thinking on method improvement" that reflect academic depth can be quickly identified, avoiding interference from irrelevant information, greatly improving the accuracy and efficiency of feature extraction, and providing a high-quality data basis for subsequent analysis.

[0106] The determined key feature set can be directly applied to the empowerment of the relationship network and the calculation of relevant capital indicators. In the relationship network, the connection weights between nodes can be reasonably determined through key features. For example, in a technical community, if key features such as the core knowledge contribution of users in a certain technical field and frequent technical communication and interaction are extracted, the connection weights with relevant users can be enhanced, more accurately reflecting the user relationship. In terms of indicator calculation, such as the knowledge production indicator, the key features can be used to accurately measure the knowledge sharing, innovation, etc. of users, making the network construction and indicator calculation more scientific and reasonable, and accurately reflecting the actual situation of social capital.

[0107] Determining the key feature set by integrating various aspects of information helps to deeply explore the hidden relationships and cognitive information in interactive texts. In the text of social media emotional communication, the deep - level social support relationships and cognitive attitude change characteristics behind emotions can be identified, further enriching the research dimensions of social capital, making the research results more comprehensive and in - depth, providing strong support for comprehensively evaluating social capital, and making up for the deficiencies of traditional methods in mining the potential value of texts.

[0108] S105: Construct an interaction network and a resource network according to the relationship and cognitive feature set.

[0109] In this embodiment, an interaction network and a resource network can be constructed based on the multi - layer network construction method according to the relationship and cognitive feature set.

[0110] S106: Construct a relationship network according to the key feature set, the interaction network and the resource network.

[0111] In this embodiment, the edges and weights of the interaction network can be adjusted according to the key feature set and the resource network, and finally a relationship network is formed.

[0112] Based on the automatic feature extraction, an interaction network and a resource network are constructed using the multi - layer network construction method according to the learner's own characteristics and their explicit and implicit interaction behaviors. Further, by integrating the resource access relationships in the resource network and the relationship characteristics revealed in the interaction content, the edges and weights of the interaction network are adjusted, and finally a relationship network is formed.

[0113] S107: Determine the corpus dataset according to the original data.

[0114] S108: Input the corpus dataset into a pre - constructed triple knowledge extraction model to extract knowledge entities and relationships, and obtain the entity - relationship extraction result.

[0115] In this embodiment, a relationship represents the relationship between knowledge entities.

[0116] In this embodiment, before inputting the corpus dataset into a pre - constructed triple knowledge extraction model to extract knowledge entities and relationships and obtain the entity - relationship extraction result, it is necessary to annotate the historical interaction data according to the predefined knowledge entities and relationships to obtain historical sample data (annotated historical interaction data). The pre - trained model is trained according to the historical sample data, and the triple knowledge extraction model is obtained by iteratively optimizing the entity recognition and relationship extraction rules.

[0117] In this embodiment, knowledge entities and relationships can be predefined manually.

[0118] After obtaining the predefined knowledge entities and relationships, annotate the historical interaction data, and repeatedly train the BERT pre-trained model with the annotated historical interaction data to iteratively optimize the entity recognition and relationship extraction rules, and finally obtain a triple knowledge extraction model with accuracy meeting the preset conditions. After constructing a concept network based on this extraction model, the present invention will assign weights to the network edges according to the frequency of occurrence of relevant triples (entity-relationship-entity) in the corpus data set under study. The corpus samples here are the texts of individuals' original and interactive content in the community, which are classified and text-processed (such as word segmentation) based on the context (such as articles, article comments, forum posts) to form corpus documents based on the context. Thus, the triple knowledge extraction model pre-trained by BERT can realize the automatic construction of the concept network of individuals / groups and one / group of corpus documents in the community.

[0119] As an example, during the annotation process, the annotator accurately marks various possible knowledge entities such as professional terms, people, organizations, events, etc. in the text according to the predefined knowledge entity and relationship system, and clearly marks the semantic relationships existing between them. For example, in academic research texts, "artificial intelligence algorithms" and "machine learning models" are marked as knowledge entities, and the "applied to" relationship between them is noted.

[0120] Next, repeatedly train the BERT pre-trained model with the annotated historical interaction data. At the initial stage of training, set reasonable parameters such as the learning rate and the number of iterations, and use a large amount of annotated data to let the model initially learn the feature patterns of knowledge entities and relationships. As the training progresses, continuously adjust the parameters according to the performance of the model on the validation set. For example, when it is found that the model has a low recognition accuracy for certain complex relationships, reduce the learning rate and increase the sample proportion of this type of relationship in the training data, and continuously iteratively optimize the entity recognition and relationship extraction rules.

[0121] During the training process, use various evaluation metrics such as accuracy, recall rate, F1 value, etc. to comprehensively evaluate the performance of the model. When the indicators of the model on the test set reach the preset accuracy conditions, such as the F1 value is higher than 0.85, it is determined that a triple knowledge extraction model meeting the requirements has been obtained.

[0122] S109: Construct a concept network according to the entity relationship extraction result.

[0123] In this embodiment, the above S109 constructing a concept network according to the entity relationship extraction result specifically includes:

[0124] S1091: Use the knowledge entities in the entity relationship extraction result as network nodes.

[0125] S1092: Use the relationships between the knowledge entities in the entity relationship extraction result as edges.

[0126] S1093: Determine the weight of the corresponding edge according to the frequency of occurrence of the relationships between knowledge entities in the entity relationship extraction result.

[0127] In this embodiment, as an example, when constructing a concept network based on the extraction model, strictly follow step S1091 to accurately use the knowledge entities in the extraction result as network nodes. In S1092, precisely transform the relationships between knowledge entities into edges. For S1093, by statistically analyzing the frequency of occurrence of each triple in a large-scale corpus dataset and using a scientific normalization method, such as dividing the frequency by the total number of triples in the corpus dataset, determine the weight of the corresponding edge, thereby constructing a concept network with a clear structure and reasonable weights, providing a solid support for the knowledge structure analysis in social capital research, and realizing the effective mapping and in-depth mining of the community knowledge system.

[0128] This invention adopts a triple knowledge extraction scheme based on BERT pre-training to extract knowledge entities and relationships from the corpus dataset constructed based on the original data, and construct a concept network. As Figure 3 shown, this method can more accurately reflect the knowledge and cognition of individuals and groups compared with the concept network constructed based on concept co-occurrence or simple concept similarity calculation, and is more suitable for the evaluation requirements of the invention in terms of individual and group knowledge and cognition in knowledge production and social identity.

[0129] S1010: Extract indicators from the interaction network, resource network, relationship network, and concept network based on the evaluation index algorithm determined by the value capital dimension to obtain the index extraction result.

[0130] In this embodiment, the value capital dimension includes: information acquisition dimension, social support dimension, knowledge production dimension, and social identity dimension. The index extraction result can be used to represent an individual's social capital in the target community, that is, the influence of an individual in the target community.

[0131] As an example, when determining the target learner, the evaluation index algorithm determined by the value capital dimension can be used to extract indicators from the interaction network, resource network, relationship network, and concept network of the personal network structure corresponding to the target learner and the interaction network, resource network, relationship network, and concept network in the community network structure to obtain the index extraction result.

[0132] As an example, the network scale and information flow metrics in the dimension of information acquisition can be determined based on the number of connections and connection strength of nodes in the interactive network, where the network scale is the scale of the individual relationship network, and the information flow is measured by the network entropy index in the individual relationship network. The relational diversity metric is determined based on the identity background differences between nodes, that is, the heterogeneity of the node identity backgrounds in the individual relationship network.

[0133] As an example, the amount of shared resources and resource update metrics in the dimension of knowledge production can be determined based on the number of nodes and resource update status in the resource network. The amount of shared resources is the number of nodes in the individual resource network, and the resource update is the average value of the new increment of resource network nodes per time unit. The knowledge quality metric is determined based on the average number of likes of the resources by community members, that is, the resource recognition degree.

[0134] As an example, the availability and network reciprocity metrics in the dimension of social support can be determined based on the density and reciprocity of the relationship network. The availability is the density of the individual in the relationship network, and the network reciprocity reflects the reciprocity of the individual relationship network. The relational support metric is determined by the network distance between the individual and the community opinion leaders, that is, the important relationships. The sense of belonging metric is measured by the clustering coefficient of the individual relationship network, that is, the network cohesion.

[0135] As an example, the social status metric in the dimension of social identity can be determined based on the centrality of the individual in the community network. The social influence metric is determined according to the ratio of the scale of the individual relationship network to the scale of the community network, that is, the relationship network scale. The important relational person metric is determined based on the proportion of community opinion leaders among the individual's relational people. The shared cognition metric is determined by the similarity between the individual concept network and the community concept network. The knowledge contribution metric is determined by the product of the amount of resources published by the individual and the average resource popularity, where the resource popularity is the ratio of the resource view count to the average view count of community resources. The social cognition input metric is determined by the total value of the social cognition level of the individual's participation in dialogue interactions.

[0136] The evaluation metric algorithms specifically include:

[0137] Based on the above-mentioned feature automatic recognition technology and network construction technology, the present invention will conduct algorithm research and development on the metrics related to the four types of value-oriented social capital feature elements. The conceptions of the metric algorithms for the related feature elements are shown in Table 2. Specifically, the expert method can be used to determine the key metrics in Table 2.

[0138] Table 2 Four Types of Value Capital Evaluation Elements, Key Metrics and Their Descriptions

[0139]

[0140]

[0141] As can be seen from Table 2, a considerable proportion of the indicators are related to the network structure characteristics and node location characteristics of the individual or community network constructed by the present invention, such as indicators such as the network scale, heterogeneity, and network density of the relationship network, the degree centrality and indirect centrality of individual nodes, etc. The present invention selects relevant indicator algorithms or further statistically defines algorithms for relevant indicators. To examine the entire community resources and capabilities that an individual and their related persons can utilize, relevant indicators also involve the calculation of relevant network characteristics of the community. In addition, some indicators can be directly obtained through statistical analysis of behavior frequencies or the number of nodes. For example, the resource recognition degree is obtained by statistically calculating the average value of the total number of likes for resources in the community.

[0142] The indicator algorithms that the present invention needs to particularly break through are the indicator algorithms related to the calculation of relationship and cognitive capital characteristics. Such algorithms need to be based on the results of the aforementioned extraction of interactive text characteristics. For example, based on the extraction of relationship characteristics and cognitive characteristics in the interactive text, social cognitive input indicators such as collaborative production factors and knowledge contribution factors can be calculated. By extracting the stylistic characteristics of the interactive text, emotional expression indicators of the trust degree factor can be calculated.

[0143] In addition, the similarity of the concept network is closely related to the calculation of some indicators related to cognitive capital characteristics. For example, the cognitive difference indicator in the knowledge diversity characteristics in knowledge production, and the community cognitive consistency indicator in shared cognition in social identity. For the calculation of the similarity of the concept network, the present invention intends to adopt graph embedding technology (such as TransH). First, the concept networks to be compared are quantified to obtain the vectors of each node of the two concept networks respectively. Then, the node vectors of the two networks are aggregated through methods such as averaging to obtain the vector of the network. Finally, the similarity score of the two concept networks is obtained by calculating the cosine similarity. The specific process is as Figure 4 shown.

[0144] In an alternative embodiment, after obtaining the indicator extraction results, the technical solution of the present invention further includes:

[0145] Performing criterion validation and optimization on the indicator extraction results based on the measurement results of the social capital scale.

[0146] Performing secondary optimization on the indicator extraction results after criterion validation and optimization by means of exploratory factor analysis to obtain multiple target indicators.

[0147] Determining the information entropy and weights corresponding to each target indicator according to the entropy weight method.

[0148] Determine the comprehensive evaluation result of social capital based on the information entropy and weight corresponding to each target index. The information entropy and weight are in an inverse relationship. The entropy weight method determines the weight by calculating the information entropy of each index of social capital. The index with a smaller information entropy has a greater impact on the comprehensive evaluation of social capital, so its weight is also larger. On the contrary, the index with a larger information entropy has a smaller role in the comprehensive evaluation, and its weight is also smaller.

[0149] In this embodiment, the measurement results of the social capital scale developed based on the three-dimensional structure of social capital (covering six dimensions such as social relationships) can be used for correlation analysis with the index extraction results. In this way, the effectiveness of the indexes is investigated, and the indexes that meet the requirements are screened out to ensure that the indexes can effectively reflect the situation of social capital. The social capital scale includes six dimensions: social relationship, common language, common vision, trust, reciprocity criterion, and identity recognition.

[0150] For the indexes optimized by criterion test, the exploratory factor analysis method is used to verify the index structure, and the indexes with low loadings are removed to further improve the index system and make the index system more scientific and reasonable.

[0151] The entropy weight method is a relatively commonly used objective weighting method for determining the weights of various different factors. The entropy weight method mainly determines the weights based on the information entropy principle and the maximum entropy principle. It measures the dispersion degree of indexes through information entropy, treats the degree of disorder as the weight, and realizes the transformation from subjective weighting to objective weighting. The entropy weight method first normalizes different types of indexes and then standardizes the original data. Then, it calculates the information entropy of each index according to the definition, and further determines the entropy weight. The weight is corrected by calculating the information redundancy, and finally, the weights of the overall social capital, four types of value capital dimensions, element characteristics, and indexes are realized, so that the overall social capital and the levels of various types of value capital can be calculated based on these indexes.

[0152] As an example, the specific calculation process of the entropy weight method includes:

[0153] (1) Index normalization: Convert different types of indexes (such as the type with the larger the better, the type with the smaller the better, and the type with the best at a certain point) into the type with the larger the better for subsequent processing.

[0154] (2) Data standardization: Normalize the original data to eliminate the influence of dimensions.

[0155] (3) Calculate the entropy of the index: According to the definition of information entropy, calculate the information entropy of each index. The calculation formula of information entropy:

[0156] where E represents information entropy, n represents the number of samples, and p ij represents the proportion of the i-th sample in the j-th index.

[0157] (4) Determine the entropy weight: Define the entropy weight of the index, that is, the entropy E of the i-th index j and its weight w j The relationship between them. The calculation formula of the entropy weight is as follows:

[0158] where k represents the number of indicators, j represents the number of index items, and the j-th index.

[0159] (5) Calculate the weight: Modify the weight of the index by calculating the information redundancy to obtain the weights of each index, and the difference coefficient D j and the weight w j The calculation formula of the weight of the index is as follows:

[0160] D j = 1 - E j

[0161] where m represents the total number of indicators.

[0162] (6) Finally, calculate the comprehensive score S i of each scheme. The calculation formula of the comprehensive score is as follows:

[0163] where, x ij represents the value of the i-th sample on the j-th index, and w j represents the weight corresponding to the j-th index.

[0164] In an alternative embodiment, the technical solution of the present invention further includes:

[0165] Construct a dynamic evaluation model of social capital based on a moment attention convolutional neural network. The dynamic evaluation model of social capital is used to predict social capital.

[0166] In an alternative embodiment, the technical solution of the present invention further includes: constructing a dynamic evaluation model of social capital based on the interaction time series characteristics of the relationship network.

[0167] The present invention intends to adopt a neural network, for example, a moment attention convolutional neural network to construct a dynamic evaluation model for four types of value capital dimensions and overall social capital. This model defines interaction features based on the historical data of each sample and constructs its relationship network. Among them, the interaction features are calculated based on the corresponding feature indicators defined by five types of features selected through literature research, including five aspects: interaction object (such as community status, homogeneity), interaction method (such as attention, access, like, reply), interaction input (such as interaction intensity, interaction regularity, interaction word count), interaction language (such as interactivity, narrativeness, emotionality), and interaction quality (such as the number of likes of a post, reply rate).

[0168] The interaction features and the relational network contain spatio-temporal sequence features. The social capital prediction task can be defined as predicting the social capital sequence S of nodes within the future time range H given the sequence of relational networks of all nodes in the past time T h , where the social capital sequence S of nodes within the future time range H is predicted f . In a scenario with N sample nodes, the input sequence data of the model is: T h ={V1, V2, V3,..., V t ,..., V T}, where where represents the interaction features at each moment of the sample, including interaction mode features, interaction object features, interaction input features, interaction language features, and interaction quality features, and r t n represents the relational network graph structure G(v, e, w) of the nth sample at the t-th moment, where v is the node, e is the edge, and w is the weight represents the four types of value capital, namely information acquisition, knowledge production, social support, and social recognition, and the overall social capital obtained by the nth sample at the t-th moment through the social capital evaluation model

[0169] where represents the social capital prediction result of N samples at the moment T + t, that is

[0170] where represents the social capital prediction result of the nth sample at the future moment T + t

[0171] The overall framework of the STGACN model is as shown in Figure 5 the framework of the STGACN model for dynamically predicting social capital. The model consists of a historical data embedding module, a spatio-temporal convolution module group, and a social capital prediction module. First, the historical data embedding module obtains the historical data of each sample. The historical data includes the sample interaction text data, semantic sentiment data, resource jump data, and interaction behavior data at each moment. Subsequently, the interaction features of each moment are calculated and a relational network is constructed. The relational network sequence is input into the spatio-temporal convolution module group, and through several spatio-temporal convolution blocks, the feature extraction and aggregation of the graph structure are completed. The features extracted by the spatio-temporal convolution module group are concatenated with the interaction features to form the final feature data, which is finally input into the social capital prediction module. In the social capital prediction module, the feature data is encoded and decoded and then enters the fully connected layer to complete the prediction of social capital

[0172] The spatio-temporal convolution block consists of two gated temporal convolution layers and a graph attention network layer. The gated convolution layer uses the gated temporal convolution method based on CNN for temporal feature extraction. The input of the gated temporal convolution layer can be expressed as Where M is the sequence length, C i is the number of input channels. The gated time convolution formula is as follows:

[0173]

[0174] Convolution Kernel ⊙ represents the Hadamard product. Here, tanh and sigmoid activation functions are used to enhance the feature extraction capability of time series data. 2C0 represents the number of output channels. P and Q represent the inputs of each gate of the gated linear unit. K t Indicates the time dimension size of the convolution kernel.

[0175] The graph attention layer combines spatial attention with graph convolution, where the attention matrix is ​​calculated as follows:

[0176] W=Oσ(Z l-1 U1U2(U3Z l-1 ) T +b), where Z l-1 is the input of the l-1th spatiotemporal block and also the input of the lth spatiotemporal block. U1, U2 and U3 are all parameter matrices to be optimized, b represents the bias term, and O represents the learnable parameter matrix.

[0177] The social capital prediction layer uses a gated recurrent unit (GRU) neural network, and combines it with the encoder and decoder to extract important spatiotemporal sequence features, and then connects to the fully connected layer to complete the prediction of social capital.

[0178] In order to realize the dynamic calculation of social capital indicators and the construction of dynamic evaluation models for recalculation, this patent needs to solve the problem of automatic extraction of social capital characteristics, network construction and indicator development based on feature extraction, indicator empowerment and calculation of various types of value capital, and dynamic prediction technology of social capital based on dynamic interaction data. The interaction data contains important information closely related to social capital relations and cognitive structure. This patent intends to break through the limitations of existing social capital evaluation that ignores the relationships and cognitive information implied by interaction data, and integrates qualitative virtual ethnography methods, language analysis, literature research and other methods to define relevant features, and uses machine learning methods to realize automatic identification of relevant features and dynamic construction of relevant networks, thereby supporting the dynamic calculation of indicators and the construction of evaluation models based on this. The overall architecture concept is as follows: Figure 6 shown.

[0179] It should be noted that the contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

[0180] The embodiment of the present invention also provides a computer device, see Figure 7, Figure 7 is a schematic diagram of the hardware structure of the computer device according to an embodiment of the present invention. As Figure 7 shown, the computer device includes: one or more processors 10, a memory 20, and interfaces for connecting various components, including a high-speed interface and a low-speed interface. Each component communicates with each other using different buses and can be installed on a common motherboard or installed in other ways as needed. The processor can process instructions executed within the computer device, including instructions stored in the memory or on the memory to display graphical information of the GUI on an external input / output device (such as a display device coupled to the interface). In an alternative embodiment, if necessary, multiple processors and / or multiple buses can be used together with multiple memories and multiple memories. Similarly, multiple computer devices can be connected, and each device provides some necessary operations (for example, as a server array, a set of blade servers, or a multi-processor device). Figure 7 In

[0181]

[0182] The memory 20 stores instructions executable by at least one processor 10, so that at least one processor 10 executes the method shown in the above embodiment.

[0183] The memory 20 may include a program storage area and a data storage area. Among them, the program storage area may store an operating device and application programs required for at least one function. The data storage area may store data created according to the use of the computer device. In addition, the memory 20 may include a high-speed random access memory, and may also include a non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state storage devices. In an alternative embodiment, the memory 20 may optionally include a memory remotely provided with respect to the processor 10, and these remote memories may be connected to the computer device through a network. Examples of the above network include but are not limited to the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.

[0184] The memory 20 may include a volatile memory, such as a random access memory. The memory may also include a non-volatile memory, such as a flash memory, a hard disk, or a solid-state drive. The memory 20 may also include a combination of the above types of memories.

[0185] The computer device further includes a communication interface 30 for the computer device to communicate with other devices or a communication network.

[0186] An embodiment of the present invention also provides a computer-readable storage medium. The method according to the embodiment of the present invention can be implemented in hardware, firmware, or be implemented as computer code that can be recorded on a storage medium, or be implemented as computer code originally stored in a remote storage medium or a non-transitory machine-readable storage medium and downloaded through a network and to be stored in a local storage medium, so that the method described herein can be processed by such software stored on a storage medium using a general-purpose computer, a dedicated processor, or programmable or dedicated hardware. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory, a random access memory, a flash memory, a hard disk, or a solid-state drive, etc. Further, the storage medium can also include a combination of the above-mentioned types of memories. It can be understood that a computer, a processor, a microprocessor controller, or programmable hardware includes a storage component that can store or receive software or computer code. When the software or computer code is accessed and executed by the computer, the processor, or the hardware, the method shown in the above embodiment is implemented.

[0187] A part of the present invention can be applied as a computer program product, such as computer program instructions. When executed by a computer, through the operation of the computer, the method and / or technical solution according to the present invention can be called or provided. Those skilled in the art should be able to understand that the forms of existence of computer program instructions in a computer-readable medium include, but are not limited to, source files, executable files, installation package files, etc. Correspondingly, the ways for computer program instructions to be executed by a computer include, but are not limited to: the computer directly executes the instructions, or the computer compiles the instructions and then executes the corresponding compiled program, or the computer reads and executes the instructions, or the computer reads and installs the instructions and then executes the corresponding installed program. Herein, the computer-readable medium can be any available computer-readable storage medium or communication medium accessible by the computer.

[0188] Although the embodiments of the present invention are described in conjunction with the accompanying drawings, those skilled in the art can make various modifications and variations without departing from the spirit and scope of the present invention, and such modifications and variations all fall within the scope defined by the appended claims.

Claims

1. A method for determining a social capital index is applied to a constructed network structure, wherein the network structure comprises: Interaction networks, resource networks, relationship networks and concept networks, characterized by including: Acquire the original data of the target community; the original data includes multiple texts; the data categories corresponding to the original data include: community interaction content, community interaction behavior and community resource content; Extracting features from the original data to obtain a set of relational and cognitive features; Analyze the original data to determine the dimension value corresponding to each text in each of the preset number of language dimensions; Determine a key feature set according to the relationship and cognitive feature set and the dimension value corresponding to each language dimension of a preset number of language dimensions for each text; constructing interaction networks and resource networks based on the set of relationships and cognitive features; constructing a relationship network according to the key feature set, the interaction network and the resource network; Determine a corpus data set according to the original data; Inputting the corpus data set into a pre-built triple knowledge extraction model to extract knowledge entities and relations, and obtaining entity relationship extraction results; the relations represent the relations between knowledge entities; Constructing a concept network according to the entity relationship extraction results; An evaluation index algorithm determined based on the value capital dimension extracts indexes from the interaction network, the resource network, the relationship network and the concept network to obtain index extraction results; the value capital dimension includes: information acquisition dimension, social support dimension, knowledge production dimension and social recognition dimension.

2. The method according to claim 1, characterized in that The obtaining of the original data of the target community includes: Get interactive messages in the target community; The interactive messages are divided into community interactive content, community interactive behavior and community resource content to obtain the original data.

3. The method according to claim 1, characterized in that The extracting features from the original data to obtain a set of relationship and cognitive features includes: The raw data is analyzed by cognitive network analysis to obtain interactive text, relational coding and cognitive coding are performed on the interactive text to obtain coding results, and feature extraction is performed on the coding results to obtain a relational and cognitive feature set; the relational and cognitive feature set includes relational features and cognitive features; the relational coding is determined according to the relational type to which the relational features belong; and the cognitive coding is determined according to the cognitive type to which the cognitive features belong.

4. The method according to claim 1, characterized in that: The analyzing the original data to determine the dimension value corresponding to each text in each of the preset number of language dimensions includes: Pre-constructing, by means of multi-dimensional analysis, a language feature for identifying a preset number of language dimensions and corresponding to each language dimension; the language feature is a language feature for indicating the intimacy of a relationship and the level of social cognition; The original data is annotated according to the language features to obtain an annotation result; and the frequency of occurrence of the language features corresponding to each language dimension in each text is counted according to the annotation result; Determine the dimension value corresponding to each of the preset number of language dimensions for each text according to the frequency of occurrence of the language features corresponding to each language dimension in each text; Among them, when the preset number is 3, the style dimensions include: the first style dimension: subjective stance dimension or objective stance dimension, the second style dimension: general narrative dimension or professional description dimension, the third style dimension: highlighting emotional attitude dimension or highlighting logical relationship dimension.

5. The method according to claim 1, characterized in that Before the corpus data set is input into the pre-built triple knowledge extraction model to extract knowledge entities and relationships and obtain entity relationship extraction results, the method includes: Label the historical interaction data according to pre-defined knowledge entities and relationships to obtain historical sample data; The pre-trained model is trained according to the historical sample data, and a triple knowledge extraction model is obtained by iteratively optimizing entity recognition and relationship extraction rules.

6. The method according to claim 1, characterized in that The step of constructing a concept network according to the entity relationship extraction result includes: Taking the knowledge entities in the entity relationship extraction result as network nodes; Using the relationship between knowledge entities in the entity relationship extraction result as an edge; The weight of the corresponding edge is determined according to the frequency of occurrence of the relationship between the knowledge entities in the entity relationship extraction result.

7. The method according to claim 1, characterized in that After obtaining the index extraction results, it also includes: Based on the measurement results of the social capital scale, the index extraction results are optimized by criterion validation; The exploratory factor analysis method is used to perform secondary optimization on the index extraction results after criterion verification optimization to obtain multiple target indicators. Determine the information entropy and weight corresponding to each target indicator according to the entropy weight method; The comprehensive evaluation results of social capital are determined based on the information entropy and weights corresponding to each target indicator.

8. The method according to claim 1, characterized in that The method further comprises: A social capital dynamic evaluation model is constructed based on a moment-by-moment attention convolutional neural network; the social capital dynamic evaluation model is used to predict social capital.

9. A computer device, characterized in that: include: A memory and a processor, wherein the memory and the processor are communicatively connected to each other, the memory stores computer instructions, and the processor executes the method for determining the social capital indicator according to any one of claims 1 to 8 by executing the computer instructions.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and the computer instructions are used to enable a computer to execute the method for determining the social capital indicator according to any one of claims 1 to 8.

Images