Online community knowledge sharing method and system based on deep learning

Abstract

The invention discloses an online community knowledge sharing method and system based on deep learning. The method comprises the steps of obtaining user attribute information of a target object and a user behavior record of the target object in a preset time period; performing feature coding on the user attribute information and the user behavior record to obtain a static feature vector and a dynamic feature vector, and splicing the static feature vector and the dynamic feature vector into a mixed feature vector; matching target multi-modal knowledge features corresponding to the mixed feature vectors from a community knowledge base, and determining multi-modal target knowledge data corresponding to the target multi-modal knowledge features; and determining knowledge learning paths of the multiple pieces of target knowledge data based on the community knowledge graph, and recommending the target knowledge data according to the knowledge learning paths. According to the method and the device, the technical problems of inaccurate knowledge pushing of the employees, low individuation degree and unclear learning path in the aspect of improving the knowledge quality of the employees are solved.

Description

Technical Field

[0001] This application relates to the technical field of financial services. Specifically, it relates to an online community knowledge sharing method and system based on deep learning. Background Art

[0002] In the context of the digital wave sweeping the globe, the financial industry is facing unprecedented challenges and opportunities. With the increasing complexity of financial operations and the diversification of customer needs, the traditional knowledge management system based on document libraries has become difficult to meet the needs of industry development. On the one hand, the forms of knowledge presentation are single and scattered, making it difficult to achieve cross-departmental knowledge sharing and integration. On the other hand, knowledge updates are lagging, unable to reflect the latest regulatory policy changes and market dynamics in real time, resulting in limited business response speed. In addition, the insufficient transformation of learning outcomes and the low efficiency of knowledge interaction and feedback make it impossible to effectively convert the training effects of employees into actual business performance, affecting the overall learning and development of the organization.

[0003] In response to the above problems, no effective solutions have been proposed yet. Summary of the Invention

[0004] Embodiments of this application provide an online community knowledge sharing method and system based on deep learning to at least solve the technical problems of inaccurate knowledge push, low personalization degree, and unclear learning path for employees in improving employees' knowledge literacy.

[0005] According to one aspect of the embodiments of this application, an online community knowledge sharing method based on deep learning is provided, including: obtaining user attribute information of a target object and user behavior records of the target object within a preset time period; respectively performing feature encoding on the user attribute information and the user behavior records to obtain a static feature vector corresponding to the user attribute information and a dynamic feature vector corresponding to the user behavior records, and splicing the static feature vector and the dynamic feature vector into a mixed feature vector; determining a target multimodal knowledge feature matching the mixed feature vector from a community knowledge base, and determining target knowledge data in multiple modalities corresponding to the target multimodal knowledge feature, where multiple multimodal knowledge features are stored in the community knowledge base, and each multimodal knowledge feature is spliced by feature vectors corresponding to the knowledge data in the corresponding multiple modalities; determining a knowledge learning path corresponding to the multiple target knowledge data according to a community knowledge graph, and recommending the target knowledge data to the target object according to the knowledge learning path, where the community knowledge graph is used to reflect the attributes and mutual relationships of the knowledge data in multiple modalities corresponding to each multimodal knowledge feature in the community knowledge base.

[0006] Optionally, the types of user attribute information include at least one of the following: gender, age, educational background, position, working years, skill qualifications, learning preferences; the types of user behavior records include at least one of the following: system operation records, customer interaction records, business processing records, historical learning records, and each user behavior record carries a corresponding timestamp.

[0007] Optionally, perform feature encoding on user attribute information and user behavior records respectively to obtain a static feature vector corresponding to the user attribute information and a dynamic feature vector corresponding to the user behavior records, including: encoding the user attribute information using a feature encoding network to obtain a static feature vector, where the type of the feature encoding network includes: Transformer encoding layer; sorting multiple user behavior records according to the corresponding timestamps to obtain a user behavior sequence; encoding the user behavior sequence using a feature encoding network to obtain a first feature vector; adding position encoding to the first feature vector to obtain a second feature vector; performing weight analysis on the second feature vector using a self-attention network and performing weighted processing on the second feature vector according to the weight analysis result to obtain a dynamic feature vector.

[0008] Optionally, the process of constructing the community knowledge base includes: obtaining multiple knowledge data of multiple modalities from multiple data sources, where the modalities of the knowledge data include at least one of the following: text, video, audio, image; classifying the multiple knowledge data according to the hierarchical structure of business domain - knowledge topic to obtain multiple knowledge data sets; for each knowledge data set, respectively analyzing the knowledge data of the corresponding modality in the knowledge data set using multiple modality feature extraction networks to obtain multiple knowledge feature vectors, performing weight analysis on each knowledge feature vector using a cross-modal attention network, and performing weighted processing on each knowledge feature vector according to the weight analysis result to obtain the multi-modal knowledge features corresponding to the knowledge data set; storing each knowledge data set, each multi-modal knowledge feature, the mapping relationship between each knowledge data set and each multi-modal knowledge feature, and each business domain and each knowledge topic in the community knowledge base.

[0009] Optionally, determining the target multi-modal knowledge feature matching the hybrid feature vector from the community knowledge base includes: mapping the hybrid feature vector to the same semantic space as the multi-modal knowledge features in the community knowledge base to obtain a target feature vector; calculating the similarity between each multi-modal knowledge feature in the community knowledge base and the target feature vector respectively; determining the multi-modal knowledge feature with a similarity greater than the preset threshold to the target feature vector as the target multi-modal knowledge feature.

[0010] Optionally, the construction process of the community knowledge graph includes: obtaining the learning priority of each knowledge data, where the learning priority is determined based on expert experience; determining the association relationship between each knowledge data based on historical business data and expert experience, where the association relationship at least includes: the dependency relationship between the basic knowledge data and the extended knowledge data; extracting knowledge entities from each knowledge data as the nodes of the community knowledge graph; determining the node attributes of each node, where the node attributes at least include: the business domain, knowledge theme, and learning priority of the knowledge data corresponding to the knowledge entity; determining the relationship between each node based on the association relationship between each knowledge data, and generating corresponding edges.

[0011] Optionally, determining the knowledge learning paths corresponding to multiple target knowledge data based on the community knowledge graph includes: determining the attributes and mutual relationships of each target knowledge data based on the community knowledge graph; combining the target knowledge data with a dependency relationship into a target knowledge data set, and determining the learning priority of the basic knowledge data in the target knowledge data set as the learning priority of the target knowledge data set, where in the target knowledge data set, the basic knowledge data is at the front, and the extended knowledge data is sorted according to the corresponding learning priority; sorting each target knowledge data set and the uncombined single target knowledge data according to the learning priority to obtain the knowledge learning path.

[0012] Optionally, the method further includes: periodically obtaining incremental knowledge data of multiple modalities from multiple data sources; incrementally updating the community knowledge base and the community knowledge graph according to the incremental knowledge data.

[0013] Optionally, the method further includes: determining the target type of the learning conversion index corresponding to each target knowledge data from a preset index mapping relationship library, where the index mapping relationship library stores the mapping relationship between different knowledge data and different learning conversion index types, and the mapping relationship is determined based on the relationship between historical learning data and historical business data and expert experience; periodically obtaining the learning conversion index data of the target type; determining that the learning state of the target object is good when all the learning conversion index data meet the preset requirements; and sending a prompt message to the target object when any learning conversion index data does not meet the preset requirements, where the prompt message is used to prompt the target object to relearn the target knowledge data corresponding to the learning conversion index data.

[0014] According to another aspect of the embodiments of the present application, there is also provided an online community knowledge sharing device based on deep learning, including: an acquisition module, configured to acquire user attribute information of a target object and user behavior records of the target object within a preset time period; a feature encoding module, configured to perform feature encoding on the user attribute information and the user behavior records respectively to obtain a static feature vector corresponding to the user attribute information and a dynamic feature vector corresponding to the user behavior records, and splice the static feature vector and the dynamic feature vector into a hybrid feature vector; a matching module, configured to determine a target multimodal knowledge feature matching the hybrid feature vector from a community knowledge base, and determine multimodal target knowledge data corresponding to the target multimodal knowledge feature, wherein multiple multimodal knowledge features are stored in the community knowledge base, and each multimodal knowledge feature is spliced by feature vectors corresponding to the corresponding multimodal knowledge data; a recommendation module, configured to determine a knowledge learning path corresponding to multiple target knowledge data according to a community knowledge graph, and recommend the target knowledge data to the target object according to the knowledge learning path, wherein the community knowledge graph is used to reflect the attributes and mutual relationships of the multimodal knowledge data corresponding to each multimodal knowledge feature in the community knowledge base.

[0015] According to another aspect of the embodiments of the present application, there is also provided a computer program product, which includes: a computer program, wherein when the computer program is executed by a processor, it implements the above-mentioned online community knowledge sharing method based on deep learning.

[0016] According to another aspect of the embodiments of the present application, there is also provided an electronic device, which includes: a memory and a processor, wherein a computer program is stored in the memory, and the processor is configured to execute the above-mentioned online community knowledge sharing method through the computer program.

[0017] In the embodiments of the present application, by obtaining the user attribute information and user behavior records of the target object and performing feature encoding, a hybrid feature vector containing static and dynamic information is generated, laying a solid foundation for subsequent accurate knowledge recommendation. With the multi-modal knowledge features stored in the community knowledge base, it is possible to quickly identify the target multi-modal knowledge features that match the user's hybrid feature vector, achieving accurate knowledge positioning. Further, by using the community knowledge graph, the knowledge learning path associated with the target knowledge data is determined. This is not just a simple knowledge recommendation list, but an ordered and coherent guidance scheme, thus guiding users to learn according to the logical relationships and priorities among the knowledge, significantly improving the learning efficiency and depth. The community knowledge graph reflects the attributes of the knowledge data and their mutual relationships, which helps to seamlessly connect knowledge with business scenarios, promote the application and transformation of knowledge in actual work, enable employees to obtain the right knowledge at the right time, better solve problems, and enhance their business capabilities, thereby solving the technical problems of inaccurate knowledge push, low personalization degree, and unclear learning path for employees in improving their knowledge literacy. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0019] Figure 1 is a schematic flowchart of an optional online community knowledge sharing method based on deep learning according to an embodiment of the present application;

[0020] Figure 2 is a schematic structural diagram of an optional online community knowledge sharing system based on deep learning according to an embodiment of the present application;

[0021] Figure 3 is a schematic structural diagram of an optional electronic device according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0022] In order to enable those skilled in the art of the present technology to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0023] It should be noted that the terms "first", "second", etc. in the description, claims and drawings of this application are used to distinguish similar users, and do not necessarily need to describe a specific order or sequence. It should be understood that such data used can be interchanged under appropriate circumstances so that the embodiments of this application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0024] In the embodiments of this application, the information collected is information and data authorized by the user or fully authorized by all parties. Moreover, for the processing of relevant data such as collection, storage, use, processing, transmission, provision, disclosure and application, etc., all comply with the relevant laws, regulations and standards of the relevant countries and regions, necessary confidentiality measures are taken, it does not violate public order and good customs, and a corresponding operation entry is provided for the user to choose to authorize or refuse.

[0025] Embodiment 1

[0026] According to the embodiments of this application, an online community knowledge sharing method based on deep learning is provided. It should be noted that the steps shown in the flowchart of the drawings can be executed in a computer system 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.

[0027] Figure 1 is a schematic flowchart of an online community knowledge sharing method based on deep learning provided according to the embodiments of this application. As Figure 1 shown, the method includes the following steps:

[0028] Step S102, obtaining the user attribute information of the target object and the user behavior records of the target object within a preset time period;

[0029] Step S104, respectively performing feature encoding on the user attribute information and the user behavior records to obtain a static feature vector corresponding to the user attribute information and a dynamic feature vector corresponding to the user behavior records, and splicing the static feature vector and the dynamic feature vector into a hybrid feature vector;

[0030] Step S106: Determine the target multimodal knowledge features that match the hybrid feature vector from the community knowledge repository, and determine the multimodal target knowledge data corresponding to the target multimodal knowledge features. Herein, multiple multimodal knowledge features are stored in the community knowledge repository, and each multimodal knowledge feature is formed by splicing the feature vectors corresponding to the corresponding multimodal knowledge data.

[0031] Step S108: Determine the knowledge learning paths corresponding to the multiple target knowledge data according to the community knowledge graph, and recommend the target knowledge data to the target object according to the knowledge learning paths. Herein, the community knowledge graph is used to reflect the attributes and mutual relationships of the multimodal knowledge data corresponding to each multimodal knowledge feature in the community knowledge repository.

[0032] The following describes each step of the online community knowledge sharing method based on deep learning in combination with a specific implementation process.

[0033] Obtain the user attribute information of the target object and the user behavior records of the target object within a preset time period.

[0034] As an optional implementation manner, the types of the user attribute information include at least one of the following: gender, age, educational background, position, working years, skill qualifications, learning preferences; the types of the user behavior records include at least one of the following: system operation records, customer interaction records, business processing records, historical learning records, and each user behavior record carries a corresponding timestamp.

[0035] For example, the user attribute information can be obtained from multiple channels, including the human resources system, employee files, employee self-declaration data, etc. These information cover key attributes such as gender, age, educational background, position, working years, skill qualifications, and learning preferences. Collect the user behavior records of the target object from various business systems, including system operation records (such as login frequency, operation duration), customer interaction records (such as phone consultations, email exchanges), business processing records (such as the number of loans processed, review speed), and historical learning records (such as course completion status, knowledge point mastery level). Each record carries a corresponding timestamp, ensuring the integrity of the time series. Sort the collected user behavior records according to the timestamp to construct a user behavior sequence. This sequence is the key for the system to understand the user's dynamic learning needs and business scenarios, and is also the basis for generating the dynamic feature vector.

[0036] For rapidly changing behavioral data, such as real-time system operation records and customer interaction records, real-time data stream processing technologies like Kafka can be adopted to ensure real-time data collection and update, so as to immediately reflect the latest needs of users. For relatively stable batch data such as historical learning records and business processing records, they can be periodically collected and integrated through distributed data processing frameworks like Spark to ensure the efficiency and effectiveness of data processing.

[0037] After obtaining the user attribute information and user behavior records, perform feature encoding on the user attribute information and user behavior records respectively to obtain the static feature vector corresponding to the user attribute information and the dynamic feature vector corresponding to the user behavior record, and concatenate the static feature vector and the dynamic feature vector into a hybrid feature vector.

[0038] As an optional implementation method, to perform feature encoding on the user attribute information and user behavior records respectively to obtain the static feature vector corresponding to the user attribute information and the dynamic feature vector corresponding to the user behavior record, the following steps can be taken: use a feature encoding network to encode the user attribute information to obtain a static feature vector, where the types of the feature encoding network include: Transformer encoding layer; sort multiple user behavior records according to the corresponding timestamps to obtain a user behavior sequence; use a feature encoding network to encode the user behavior sequence to obtain a first feature vector; add position encoding to the first feature vector to obtain a second feature vector; use a self-attention network to perform weight analysis on the second feature vector and perform weighted processing on the second feature vector according to the weight analysis result to obtain a dynamic feature vector.

[0039] For example, an encoding layer with the Transformer architecture can be used to encode the static attribute information of users, such as gender, age, educational background, job position, working years, skill qualifications, and learning preferences, etc. By converting these attributes into numerical vectors, it is ensured that they can be understood by subsequent machine learning models. For categorical attributes (such as gender, job position), an embedding layer is used to convert the attributes into low-dimensional continuous vectors; for numerical attributes (such as age, working years), they are directly normalized to convert into standardized numerical vectors. During the encoding process, the numerical attributes are standardized to eliminate the influence of dimensions and ensure the comparability of different attributes during encoding.

[0040] Sort according to the timestamp of each user behavior record to construct a user behavior sequence that reflects the user behavior time series. Preprocess the behavior sequence, including operations such as removing noise data and filling in missing timestamps, to ensure the integrity and accuracy of the sequence. The Transformer encoding layer with the multi-head self-attention mechanism can be used to encode the user behavior sequence to generate the first feature vector. This step can capture the long-term dependencies between user behaviors and improve the pertinence of recommendations. Based on the first feature vector, add dynamic positional encoding to enhance the time series information and generate the second feature vector. The dynamic positional encoding can reflect the relative positions of user behaviors on the time axis and help the model understand the temporal patterns of behaviors. Use the self-attention network to weight the second feature vector. The self-attention network calculates the importance weights of each behavior record for the overall user behavior, and weight-fuse the second feature vector with the calculated weights to generate the dynamic feature vector. This process ensures that the dynamic feature vector can reflect the recent behavior patterns of users and the focus of learning needs.

[0041] Concatenate the static feature vector and the dynamic feature vector in the feature space to form a hybrid feature vector that comprehensively reflects user attributes and behaviors. Considering that the hybrid feature vector may contain a large number of features, dimensionality reduction techniques such as principal component analysis or t-distributed stochastic neighbor embedding can be applied to reduce the computational cost while maintaining the core information of the vector.

[0042] When encoding the user behavior sequence, specific adjustments can be made to the Transformer architecture, such as adding additional attention heads to more finely capture the interactions between user behaviors. When adding positional encoding, various strategies may be adopted, including fixed positional encoding and learnable positional encoding, to adapt to the characteristics of user behavior sequences in different scenarios. The weight analysis results of the self-attention network will be dynamically adjusted according to user feedback to optimize the generation of the dynamic feature vector and more accurately reflect user needs.

[0043] For example, Zhang, a credit specialist at a bank, has recently been handling a technology loan business involving specialized, sophisticated, distinctive, and innovative enterprises. Zhang's static feature vector may include: gender (male), age (32 years old), educational background (Master of Finance), position (Senior Manager of the Credit Department), work experience (8 years), skill qualification (CFA holder), and learning preference (preferring case studies and video tutorials). Zhang's dynamic feature vector is based on his recent behavior records, such as repeatedly referring to the risk assessment guidelines for technology loans, participating in an online seminar on the financing policies for specialized, sophisticated, distinctive, and innovative enterprises, and frequently operating the module related to intellectual property assessment in the loan approval system. Using a deep learning model (such as a multi-modal Transformer), Zhang's hybrid feature vector is mapped to the semantic space of the multi-modal knowledge feature vector in the community knowledge base for subsequent similarity calculation. Traverse the community knowledge base, which contains a large number of multi-modal knowledge features regarding credit policies, risk assessments, and financing cases of technology enterprises. For each multi-modal knowledge feature vector, the system calculates its cosine similarity with Zhang's hybrid feature vector. Set a preset similarity threshold, such as 0.8. The system filters out all knowledge features with a similarity higher than this threshold. Based on the filtering results, the system recommends a series of multi-modal knowledge contents to Zhang, including interpretation videos on the financing policies for specialized, sophisticated, distinctive, and innovative enterprises, document materials containing intellectual property assessment methods, and successful case analyses related to technology loans.

[0044] After obtaining the hybrid feature vector, determine the target multi-modal knowledge feature that matches the hybrid feature vector from the community knowledge base, and determine the multi-modal target knowledge data corresponding to the target multi-modal knowledge feature. Among them, multiple multi-modal knowledge features are stored in the community knowledge base, and each multi-modal knowledge feature is formed by splicing the feature vectors corresponding to the knowledge data of the corresponding modality.

[0045] As an optional implementation manner, the construction process of the community knowledge base can be carried out in the following steps: obtain multiple knowledge data of multiple modalities from multiple data sources, where the modalities of the knowledge data include at least one of the following: text, video, audio, image; classify the multiple knowledge data according to the hierarchical structure of business domain - knowledge topic to obtain multiple knowledge data sets; for each knowledge data set, respectively use feature extraction networks of multiple modalities to analyze the knowledge data of the corresponding modality in the knowledge data set to obtain multiple knowledge feature vectors, use a cross-modal attention network to perform weight analysis on each knowledge feature vector, and perform weighted processing on each knowledge feature vector according to the weight analysis results to obtain the multi-modal knowledge feature corresponding to the knowledge data set; store each knowledge data set, each multi-modal knowledge feature, the mapping relationship between each knowledge data set and each multi-modal knowledge feature, and each business domain and each knowledge topic in the community knowledge base.

[0046] For example, obtain knowledge data in different modalities such as text, video, audio, and images from various sources within the bank's internal department systems, external regulatory announcements, industry research reports, and learning platforms. Use modality conversion technologies, such as automatic speech recognition to convert audio to text and optical character recognition to convert text information in images to a readable format, to ensure that all modality data can be effectively processed and utilized. For each knowledge data set, use a dedicated feature extraction network to analyze the knowledge data in different modalities such as text, video, and images, generate knowledge feature vectors corresponding to the modalities, adopt a cross-modal attention network to perform weight analysis on the knowledge feature vectors of each modality, and perform weighted processing on the knowledge feature vectors according to the analysis results to integrate information from different modalities. Finally, generate a multi-modal knowledge feature vector, and store each knowledge data set, the corresponding multi-modal knowledge features, and the mapping relationship between the business domain and the knowledge topic in the community knowledge base to form a structured knowledge graph.

[0047] As an alternative implementation, to determine the target multi-modal knowledge features that match the hybrid feature vector from the community knowledge base, the following steps can be taken: map the hybrid feature vector to the same semantic space as the multi-modal knowledge features in the community knowledge base to obtain the target feature vector; calculate the similarity between each multi-modal knowledge feature in the community knowledge base and the target feature vector respectively; determine the multi-modal knowledge features with a similarity greater than the preset threshold to the target feature vector as the target multi-modal knowledge features.

[0048] For example, mapping the hybrid feature vector to the same semantic space as the multi-modal knowledge features in the community knowledge base. This mapping process is usually achieved through a deep learning model (such as a multi-modal Transformer), which can convert the user feature vector into a form that can be compared with the knowledge feature vector. During the mapping process, the model takes into account the comprehensive influence of user attribute information and behavior records to ensure that the target feature vector can accurately reflect the user's current learning needs and business scenarios. After the hybrid feature vector is mapped to the semantic space, the similarity between each multi-modal knowledge feature in the community knowledge base and the target feature vector will be calculated respectively. The methods for calculating similarity are usually cosine similarity, Euclidean distance, or Jaccard similarity, depending on the type of vector representation and the attributes of the feature space. For example, for semantic vectors, cosine similarity is commonly used because it can reflect the angular relationship between vectors rather than just focusing on the distance.

[0049] After obtaining the target multi-modal knowledge features, determine the knowledge learning paths corresponding to multiple target knowledge data according to the community knowledge graph, and recommend the target knowledge data to the target object according to the knowledge learning paths. The community knowledge graph is used to reflect the attributes and mutual relationships of the multi-modal knowledge data corresponding to each multi-modal knowledge feature in the community knowledge base.

[0050] As an optional implementation manner, the construction process of the community knowledge graph can take the following steps: obtain the learning priority of each knowledge data, where the learning priority is determined based on expert experience; determine the association relationship between each knowledge data according to historical business data and expert experience, where the association relationship at least includes: the dependency relationship between the basic knowledge data and the extended knowledge data; extract knowledge entities from each knowledge data as the nodes of the community knowledge graph; determine the node attributes of each node, where the node attributes at least include: the business field, knowledge theme, and learning priority of the knowledge data corresponding to the knowledge entity; determine the relationship between each node according to the association relationship between each knowledge data, and generate the corresponding edges.

[0051] For example, by consulting experts in each business field, summarize the learning priorities of different knowledge data, which is the first step in constructing the knowledge graph. For example, legal rules may be the knowledge data with the highest priority for employees in the compliance department. Based on historical business data, analyze the dependency relationship between each knowledge data, such as the loan application process depends on basic knowledge data such as credit assessment criteria. Combining expert experience judgment, supplement potential associations that may not be fully reflected in the business data to ensure the comprehensiveness of the relationship between knowledge. Automatically extract knowledge entities from multi-modal knowledge data, such as small and micro enterprise loans, financial derivatives, etc., as the nodes of the community knowledge graph, and assign attributes such as business field, knowledge theme, and learning priority to each node. These attributes will be used for subsequent knowledge path construction and recommendation. According to the association relationship between knowledge data, determine the dependency edges between nodes, such as the pre-and post-relationship between the personal financial planning node and the portfolio analysis node, and assign attributes to each edge, such as dependency strength, correlation score, etc. These attributes will affect the construction of the knowledge path during recommendation.

[0052] As an alternative implementation, to determine the knowledge learning paths corresponding to multiple target knowledge data based on the community knowledge graph, the following steps can be taken: Determine the attributes and interrelationships of each target knowledge data based on the community knowledge graph; Combine the target knowledge data with dependency relationships into a target knowledge data set, and determine the learning priority of the basic knowledge data in the target knowledge data set as the learning priority of the target knowledge data set. Among them, in the target knowledge data set, the basic knowledge data is at the forefront, and the extended knowledge data is sorted according to the corresponding learning priorities; Sort each target knowledge data set and the uncombined single target knowledge data according to the learning priorities to obtain the knowledge learning paths.

[0053] For example, based on the community knowledge graph, reconfirm the attributes of each target knowledge data (such as business domain, knowledge topic, learning priority) and the relationships with other knowledge data, analyze the knowledge graph, identify the target knowledge data that is the learning basis for other knowledge data, that is, the basic knowledge data. For the extended knowledge data related to the basic knowledge data, sort them according to their respective learning priorities to form a target knowledge data set. Set the learning priority of the basic knowledge data as the highest priority of the set, and other extended knowledge data are sorted according to their positions (such as depth, path length) and correlation scores in the knowledge graph. According to the learning priorities, sort the target knowledge data sets and the uncombined single target knowledge data to form one or more knowledge learning paths. Considering factors such as the user's learning progress and the mastered knowledge points, dynamically adjust the knowledge learning paths to avoid repeated learning or skipping necessary knowledge.

[0054] As an alternative implementation, the method further includes: Periodically obtaining multi-modal incremental knowledge data from multiple data sources; Incrementally updating the community knowledge base and the community knowledge graph based on the incremental knowledge data.

[0055] As an alternative implementation, the method further includes: Determining the target types of learning conversion indicators corresponding to each target knowledge data from a preset index mapping relationship library, where the index mapping relationship library stores the mapping relationships between different knowledge data and different learning conversion indicator types, and the mapping relationships are determined based on the relationships between historical learning data and historical business data and expert experience; Periodically obtaining the learning conversion indicator data of the target type; When all the learning conversion indicator data meet the preset requirements, determining that the learning status of the target object is good; When any learning conversion indicator data does not meet the preset requirements, sending a prompt message to the target object, where the prompt message is used to prompt to relearn the target knowledge data corresponding to the learning conversion indicator data.

[0056] For example, increment knowledge data from multiple data sources (including internal business systems, external regulatory announcements, industry research reports, etc.) is periodically monitored and captured, covering various modalities such as text, video, audio, and images. For incremental data, real-time stream processing technologies such as Kafka can be used for rapid capture and preliminary analysis. At the same time, for larger amounts of data, distributed batch processing frameworks such as Spark are used for in-depth processing and integration. Technologies such as natural language processing and image recognition are used to parse the incremental knowledge data, extract key information and knowledge entities, and based on graph neural networks, the nodes and edges in the knowledge graph are updated in real time to ensure that the knowledge graph can reflect the latest knowledge data structure and relationships. The incremental knowledge data is integrated into the community knowledge base, including classifying, annotating, and storing the knowledge data, as well as updating the metadata of the knowledge data.

[0057] Based on the association between historical learning data and historical business data, as well as expert experience, an index mapping relationship library is constructed to clarify the mapping relationships between different knowledge data and business metrics (such as work efficiency improvement, business compliance rate, customer satisfaction, etc.). From the mapping relationship library, the target types of learning conversion metrics related to each target knowledge data are identified for subsequent learning effectiveness evaluation. Learning conversion metric data is periodically captured from relevant business systems and data sources, including but not limited to employees' business processing efficiency, compliance attainment rate, customer feedback scores, etc. The captured metric data is matched and merged with the user's learning records to form a learning effectiveness dataset for each user, and at the same time, data cleaning is performed to remove invalid or outlier values.

[0058] The learning conversion metric data can be evaluated according to preset learning effectiveness criteria to determine whether it meets the requirements. If all learning conversion metric data meet the preset requirements, it is judged that the user's learning status is good, which may trigger further recommendations for advanced knowledge content or reward mechanisms. If any learning conversion metric data fails to meet the standard, the system will automatically send a prompt message to the user, suggesting to re-learn the knowledge data corresponding to the unqualified metric to improve the learning effectiveness.

[0059] In the embodiments of the present application, by obtaining the user attribute information and user behavior records of the target object and performing feature encoding, a hybrid feature vector containing static and dynamic information is generated, laying a solid foundation for subsequent accurate knowledge recommendation. With the multi-modal knowledge features stored in the community knowledge base, it is possible to quickly identify the target multi-modal knowledge features that match the user's hybrid feature vector, achieving accurate knowledge positioning. Further, using the community knowledge graph, the knowledge learning path associated with the target knowledge data is determined. This is not just a simple knowledge recommendation list, but an ordered and coherent guidance scheme, thus guiding users to learn according to the logical relationships and priorities between knowledge, significantly improving the learning efficiency and depth. The community knowledge graph reflects the attributes of knowledge data and their mutual relationships, which helps to seamlessly connect knowledge with business scenarios, promote the application and transformation of knowledge in actual work, enabling employees to obtain the right knowledge at the right time, better solve problems, and enhance their business capabilities. Furthermore, it solves the technical problems of inaccurate knowledge push, low personalization degree, and unclear learning path for employees in improving their knowledge literacy.

[0060] Embodiment 2

[0061] According to the embodiments of the present application, there is also provided a deep learning-based online community knowledge sharing system for implementing the deep learning-based online community knowledge sharing method in Embodiment 1, as Figure 2 shown. The deep learning-based online community knowledge sharing system at least includes: an acquisition module 21, a feature encoding module 22, a matching module 23, and a recommendation module 24, where:

[0062] The acquisition module 21 is configured to obtain the user attribute information of the target object and the user behavior records of the target object within a preset time period;

[0063] The feature encoding module 22 is configured to perform feature encoding on the user attribute information and the user behavior records respectively to obtain a static feature vector corresponding to the user attribute information and a dynamic feature vector corresponding to the user behavior records, and splice the static feature vector and the dynamic feature vector into a hybrid feature vector;

[0064] The matching module 23 is configured to determine, from the community knowledge base, target multi-modal knowledge features that match the hybrid feature vector, and determine multi-modal target knowledge data corresponding to the target multi-modal knowledge features. Among them, multiple multi-modal knowledge features are stored in the community knowledge base, and each multi-modal knowledge feature is spliced by feature vectors corresponding to the corresponding multi-modal knowledge data;

[0065] A recommendation module 24 is used to determine knowledge learning paths corresponding to multiple target knowledge data based on a community knowledge graph, and recommend the target knowledge data to a target object according to the knowledge learning paths. The community knowledge graph is used to reflect the attributes and mutual relationships of multimodal knowledge data corresponding to various multimodal knowledge features in a community knowledge base.

[0066] The functions of each module of the online community knowledge sharing system based on deep learning will be described below in combination with specific implementation processes.

[0067] An acquisition module acquires user attribute information of a target object and user behavior records of the target object within a preset time period.

[0068] As an optional implementation manner, the types of user attribute information include at least one of the following: gender, age, educational background, position, working years, skill qualifications, learning preferences; the types of user behavior records include at least one of the following: system operation records, customer interaction records, business processing records, historical learning records, and each user behavior record carries a corresponding timestamp.

[0069] A feature encoding module respectively performs feature encoding on the user attribute information and the user behavior records to obtain a static feature vector corresponding to the user attribute information and a dynamic feature vector corresponding to the user behavior records, and splices the static feature vector and the dynamic feature vector into a mixed feature vector.

[0070] As an optional implementation manner, respectively performing feature encoding on the user attribute information and the user behavior records to obtain a static feature vector corresponding to the user attribute information and a dynamic feature vector corresponding to the user behavior records may adopt the following steps: encoding the user attribute information by using a feature encoding network to obtain a static feature vector, where the type of the feature encoding network includes: a Transformer encoding layer; sorting multiple user behavior records according to the corresponding timestamps to obtain a user behavior sequence; encoding the user behavior sequence by using a feature encoding network to obtain a first feature vector; adding a position encoding to the first feature vector to obtain a second feature vector; performing weight analysis on the second feature vector by using a self-attention network, and performing weighted processing on the second feature vector according to the weight analysis result to obtain a dynamic feature vector.

[0071] A matching module determines target multimodal knowledge features matching the mixed feature vector from a community knowledge base, and determines multimodal target knowledge data corresponding to the target multimodal knowledge features. The community knowledge base stores multiple multimodal knowledge features, and each multimodal knowledge feature is spliced by feature vectors corresponding to the corresponding multimodal knowledge data.

[0072] As an alternative implementation, the construction process of the community knowledge base can follow the following steps: Obtain multiple knowledge data in multiple modalities from multiple data sources, where the modalities of the knowledge data include at least one of the following: text, video, audio, image; classify the multiple knowledge data according to the hierarchical structure of business domain - knowledge topic to obtain multiple knowledge data sets; for each knowledge data set, respectively use feature extraction networks in multiple modalities to analyze the knowledge data in the corresponding modality in the knowledge data set to obtain multiple knowledge feature vectors, use a cross-modal attention network to perform weight analysis on each knowledge feature vector, and perform weighted processing on each knowledge feature vector according to the weight analysis result to obtain the multi-modal knowledge features corresponding to the knowledge data set; store each knowledge data set, each multi-modal knowledge feature, the mapping relationship between each knowledge data set and each multi-modal knowledge feature, and between each business domain and each knowledge topic in the community knowledge base.

[0073] As an alternative implementation, to determine the target multi-modal knowledge features matching the hybrid feature vector from the community knowledge base, the following steps can be taken: Map the hybrid feature vector to the same semantic space as the multi-modal knowledge features in the community knowledge base to obtain a target feature vector; calculate the similarity between each multi-modal knowledge feature in the community knowledge base and the target feature vector respectively; determine the multi-modal knowledge features with a similarity greater than a preset threshold to the target feature vector as the target multi-modal knowledge features.

[0074] The recommendation module determines the knowledge learning paths corresponding to multiple target knowledge data based on the community knowledge graph, and recommends the target knowledge data to the target object according to the knowledge learning paths, where the community knowledge graph is used to reflect the attributes and mutual relationships of the multi-modal knowledge data corresponding to each multi-modal knowledge feature in the community knowledge base.

[0075] As an alternative implementation, the construction process of the community knowledge graph can follow the following steps: Obtain the learning priority of each knowledge data, where the learning priority is determined based on expert experience; determine the association relationships between each knowledge data based on historical business data and expert experience, where the association relationships at least include: the dependency relationship between basic knowledge data and extended knowledge data; extract knowledge entities from each knowledge data as the nodes of the community knowledge graph; determine the node attributes of each node, where the node attributes at least include: the business domain, knowledge topic, and learning priority of the knowledge data corresponding to the knowledge entity; determine the relationships between each node based on the association relationships between each knowledge data and generate corresponding edges.

[0076] As an alternative implementation, to determine the knowledge learning paths corresponding to multiple target knowledge data based on the community knowledge graph, the following steps can be taken: Determine the attributes and mutual relationships of each target knowledge data based on the community knowledge graph; Combine the target knowledge data with dependency relationships into target knowledge data sets, and determine the learning priorities of the basic knowledge data in the target knowledge data sets as the learning priorities of the target knowledge data sets. Among them, in the target knowledge data sets, the basic knowledge data is at the forefront, and the extended knowledge data is sorted according to the corresponding learning priorities; Sort each target knowledge data set and the uncombined single target knowledge data according to the learning priorities to obtain the knowledge learning paths.

[0077] As an alternative implementation, the method further includes: Periodically obtaining incremental knowledge data of multiple modalities from multiple data sources; Incrementally updating the community knowledge base and the community knowledge graph based on the incremental knowledge data.

[0078] As an alternative implementation, the method further includes: Determining the target types of learning conversion indicators corresponding to each target knowledge data from a preset indicator mapping relationship library, where the indicator mapping relationship library stores the mapping relationships between different knowledge data and different learning conversion indicator types, and the mapping relationships are determined based on the relationships between historical learning data and historical business data and expert experience; Periodically obtaining learning conversion indicator data of the target type; When all the learning conversion indicator data meet the preset requirements, determining that the learning status of the target object is good; When any learning conversion indicator data does not meet the preset requirements, sending a prompt message to the target object, where the prompt message is used to prompt the target object to relearn the target knowledge data corresponding to the learning conversion indicator data.

[0079] It should be noted that each module in the online community knowledge sharing system based on deep learning in the embodiments of the present application corresponds one by one to each implementation step of the online community knowledge sharing method based on deep learning in Embodiment 1. Since detailed descriptions have been made in Embodiment 1, some details not shown in this embodiment can be referred to Embodiment 1 and will not be elaborated here.

[0080] Embodiment 3

[0081] According to the embodiments of the present application, there is also provided a computer program product, which includes a computer program. When the computer program is executed by a processor, it implements the online community knowledge sharing method based on deep learning in Embodiment 1.

[0082] According to an embodiment of the present application, a non-volatile storage medium is further provided. The non-volatile storage medium includes a stored computer program. Wherein, the device where the non-volatile storage medium is located executes the online community knowledge sharing method based on deep learning in Embodiment 1 by running the computer program.

[0083] According to an embodiment of the present application, a processor is further provided. The processor is used to run a computer program. Wherein, when the computer program runs, it executes the online community knowledge sharing method based on deep learning in Embodiment 1.

[0084] According to an embodiment of the present application, an electronic device is further provided. The electronic device includes: a memory and a processor. Wherein, a computer program is stored in the memory, and the processor is configured to execute the online community knowledge sharing method based on deep learning in Embodiment 1 through the computer program.

[0085] Specifically, when the computer program runs, it executes the following steps: obtaining the user attribute information of the target object and the user behavior record of the target object within a preset time period; respectively performing feature encoding on the user attribute information and the user behavior record to obtain a static feature vector corresponding to the user attribute information and a dynamic feature vector corresponding to the user behavior record, and splicing the static feature vector and the dynamic feature vector into a mixed feature vector; determining a target multi-modal knowledge feature matching the mixed feature vector from the community knowledge base, and determining the multi-modal target knowledge data corresponding to the target multi-modal knowledge feature. Wherein, multiple multi-modal knowledge features are stored in the community knowledge base, and each multi-modal knowledge feature is spliced by feature vectors corresponding to the corresponding multi-modal knowledge data; determining the knowledge learning paths corresponding to the multiple target knowledge data according to the community knowledge graph, and recommending the target knowledge data to the target object according to the knowledge learning paths. Wherein, the community knowledge graph is used to reflect the attributes and mutual relationships of the multi-modal knowledge data corresponding to each multi-modal knowledge feature in the community knowledge base.

[0086] As an optional implementation manner, the above-mentioned electronic device may exist in the form of a mobile terminal, a computer terminal or a similar computing device. Figure 3 Shows a hardware structure block diagram of an electronic device for implementing the online community knowledge sharing method based on deep learning. As Figure 3As shown, the electronic device 30 may include one or more processors 302 (shown as 302a, 302b, ……, 302n in the figure) (the processor 302 may include, but is not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA), a memory 304 for storing data, and a transmission device 306 for communication functions. In addition, it may further include: a display, an input / output interface (I / O interface), a universal serial bus (USB) port (which may be included as one of the ports of the BUS bus), a network interface, a power supply, and / or a camera. Those of ordinary skill in the art can understand that Figure 3 the structure shown is only schematic and does not limit the structure of the above-mentioned electronic device. For example, the electronic device 30 may further include more or fewer components than Figure 3 shown in, or have a different configuration from Figure 3 that shown.

[0087] It should be noted that the above one or more processors 302 and / or other data processing circuits are generally referred to as "data processing circuits" herein. The data processing circuit may be embodied in software, hardware, firmware, or any combination thereof, in whole or in part. In addition, the data processing circuit may be a single independent processing module, or be incorporated in whole or in part into any one of the other elements in the electronic device 30. As involved in the embodiments of the present application, the data processing circuit is a processor control (such as the selection of a variable resistance terminal path connected to an interface).

[0088] The memory 304 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the online community knowledge sharing method based on deep learning in the embodiments of the present application. The processor 302 executes various functional applications and data processing by running the software programs and modules stored in the memory 304, that is, implements the vulnerability detection method of the above application program. The memory 304 may include a high-speed random access memory, and may further include a non-volatile memory, such as one or more magnetic storage devices, a flash memory, or other non-volatile solid-state memories. In some instances, the memory 304 may further include a memory remotely set relative to the processor 302, and these remote memories may be connected to the electronic device 30 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.

[0089] The transmission device 306 is used to receive or send data via a network. Specific examples of the above-mentioned network may include a wireless network provided by a communication provider of the electronic device 30. In one example, the transmission device 306 includes a network adapter (Network Interface Controller, NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 306 can be a Radio Frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0090] The display can be, for example, a touch-screen liquid crystal display (LCD), which enables the user to interact with the user interface of the electronic device 30.

[0091] The above-mentioned serial numbers of the embodiments are only for description and do not represent the advantages or disadvantages of the embodiments.

[0092] In the above embodiments of the present application, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0093] In several embodiments provided in the present application, it should be understood that the disclosed technical content can be implemented in other ways. Among them, the device embodiments described above are only illustrative. For example, the division of units can be a logical function division. In actual implementation, there can be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed coupling or direct coupling or communication connection to each other can be through some interfaces. The indirect coupling or communication connection of units or modules can be in an electrical or other form.

[0094] The units described as separate components may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place or distributed to multiple units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0095] In addition, the functional units in the various embodiments of the present application can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0096] When the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of this application. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), mobile hard disks, magnetic disks, or optical discs.

[0097] The above are only the preferred embodiments of this application. It should be noted that for those of ordinary skill in the art of this technology, without departing from the principle of this application, several improvements and refinements can still be made, and these improvements and refinements should also be regarded as the protection scope of this application.

Claims

1. An online community knowledge sharing method based on deep learning, characterized in that, Including: Obtain the user attribute information of the target object and the user behavior records of the target object within a preset time period; Respectively perform feature encoding on the user attribute information and the user behavior records to obtain a static feature vector corresponding to the user attribute information and a dynamic feature vector corresponding to the user behavior records, and splice the static feature vector and the dynamic feature vector into a hybrid feature vector; Determine the target multi-modal knowledge feature matching the hybrid feature vector from the community knowledge base, and determine the multi-modal target knowledge data corresponding to the target multi-modal knowledge feature, wherein multiple multi-modal knowledge features are stored in the community knowledge base, and each multi-modal knowledge feature is spliced by feature vectors corresponding to the corresponding multi-modal knowledge data; Determine the knowledge learning paths corresponding to the multiple target knowledge data according to the community knowledge graph, and recommend the target knowledge data to the target object according to the knowledge learning paths, wherein the community knowledge graph is used to reflect the attributes and mutual relationships of the multi-modal knowledge data corresponding to each multi-modal knowledge feature in the community knowledge base.

2. The method according to claim 1, wherein: The types of the user attribute information include at least one of the following: gender, age, educational background, position, working years, skill qualifications, learning preferences; The types of the user behavior records include at least one of the following: system operation records, customer interaction records, business processing records, historical learning records, and each user behavior record carries a corresponding timestamp.

3. The method according to claim 2, characterized in that, Respectively performing feature encoding on the user attribute information and the user behavior records to obtain a static feature vector corresponding to the user attribute information and a dynamic feature vector corresponding to the user behavior records includes: Use a feature encoding network to encode the user attribute information to obtain the static feature vector, wherein the type of the feature encoding network includes: Transformer encoding layer; Sort the multiple user behavior records according to the corresponding timestamps to obtain a user behavior sequence; Use the feature encoding network to encode the user behavior sequence to obtain a first feature vector; Add position encoding to the first feature vector to obtain a second feature vector; Use a self-attention network to perform weight analysis on the second feature vector, and perform weighted processing on the second feature vector according to the weight analysis result to obtain the dynamic feature vector.

4. The method according to claim 1, wherein The construction process of the community knowledge base includes: Obtain multiple knowledge data of multiple modalities from multiple data sources, wherein the modalities of the knowledge data include at least one of the following: text, video, audio, image; Classify the multiple knowledge data according to the hierarchical structure of business domain - knowledge topic to obtain multiple knowledge data sets; For each knowledge data set, use the feature extraction networks of multiple modalities to analyze the knowledge data of the corresponding modality in the knowledge data set, obtain multiple knowledge feature vectors, use the cross-modal attention network to perform weight analysis on each of the knowledge feature vectors, and perform weighted processing on each of the knowledge feature vectors according to the weight analysis results to obtain the multi-modal knowledge features corresponding to the knowledge data set; Store each of the knowledge data sets, each of the multi-modal knowledge features, the mapping relationships between each of the knowledge data sets and each of the multi-modal knowledge features, and between each business domain and each knowledge topic in the community knowledge base.

5. The method according to claim 1, wherein Determine the target multi-modal knowledge features matching the hybrid feature vector from the community knowledge base, including: Map the hybrid feature vector to the same semantic space as the multi-modal knowledge features in the community knowledge base to obtain a target feature vector; Calculate the similarity between each multi-modal knowledge feature in the community knowledge base and the target feature vector respectively; Determine the multi-modal knowledge features with a similarity greater than a preset threshold to the target feature vector as the target multi-modal knowledge features.

6. The method according to claim 4, wherein The construction process of the community knowledge graph includes: Obtain the learning priority of each knowledge data, where the learning priority is determined based on expert experience; Determine the association relationships between each knowledge data based on historical business data and expert experience, where the association relationships at least include: the dependency relationship between basic knowledge data and extended knowledge data; Extract knowledge entities from each knowledge data as the nodes of the community knowledge graph; Determine the node attributes of each node, where the node attributes at least include: the business domain, knowledge topic, and learning priority of the knowledge data corresponding to the knowledge entity; Determine the relationships between each node based on the association relationships between each knowledge data and generate corresponding edges.

7. The method according to claim 6, wherein Determine the knowledge learning paths corresponding to multiple target knowledge data according to the community knowledge graph, including: Determine the attributes and mutual relationships of each target knowledge data according to the community knowledge graph; Combine the target knowledge data with dependency relationships into a target knowledge data set, and determine the learning priority of the basic knowledge data in the target knowledge data set as the learning priority of the target knowledge data set. Among the target knowledge data set, the basic knowledge data is at the front, and the extended knowledge data is sorted according to the corresponding learning priority; Sort each of the target knowledge data sets and the uncombined single target knowledge data according to the learning priority to obtain the knowledge learning path.

8. The method according to claim 6, characterized in that, The method further includes: Periodically obtain incremental knowledge data of multiple modalities from the multiple data sources; Perform incremental updates on the community knowledge base and the community knowledge graph according to the incremental knowledge data.

9. The method according to claim 1, wherein The method further includes: Determine the target types of learning transformation metrics corresponding to each of the target knowledge data from a preset metric mapping relationship library, where the metric mapping relationship library stores the mapping relationships between different knowledge data and different learning transformation metric types, and the mapping relationships are determined based on the relationships between historical learning data and historical business data and expert experience; Periodically obtain the learning transformation metric data of the target type; When all the learning transformation metric data meet the preset requirements, determine that the learning state of the target object is good; When any learning transformation metric data does not meet the preset requirements, send a prompt message to the target object, where the prompt message is used to prompt the target object to relearn the target knowledge data corresponding to the learning transformation metric data.

10. An online community knowledge sharing system based on deep learning, characterized in that, It includes: An acquisition module, configured to acquire the user attribute information of the target object and the user behavior records of the target object within a preset time period; A feature encoding module, configured to perform feature encoding on the user attribute information and the user behavior records respectively to obtain a static feature vector corresponding to the user attribute information and a dynamic feature vector corresponding to the user behavior records, and splice the static feature vector and the dynamic feature vector into a mixed feature vector; A matching module, configured to determine the target multimodal knowledge feature matching the mixed feature vector from the community knowledge base, and determine the multimodal target knowledge data corresponding to the target multimodal knowledge feature, where the community knowledge base stores multiple multimodal knowledge features, and each multimodal knowledge feature is spliced by feature vectors corresponding to the corresponding multimodal knowledge data; A recommendation module, configured to determine the knowledge learning paths corresponding to the multiple target knowledge data according to the community knowledge graph, and recommend the target knowledge data to the target object according to the knowledge learning paths, where the community knowledge graph is used to reflect the attributes and mutual relationships of the multimodal knowledge data corresponding to each multimodal knowledge feature in the community knowledge base.

11. A computer program product, characterized in that, It includes: A computer program, where when the computer program is executed by a processor, it implements the online community knowledge sharing method based on deep learning according to any one of claims 1 to 9.

12. An electronic device, characterized in that, It includes: A memory and a processor, where a computer program is stored in the memory, and the processor is configured to execute the online community knowledge sharing method based on deep learning according to any one of claims 1 to 9 through the computer program.

Images