A recommendation method, apparatus, device and storage medium

Abstract

Embodiments of the present application provide a recommendation method and device, electronic equipment and storage medium, the method comprises: obtaining the search demand of a user; generating a search label and a corresponding label probability; obtaining a probability knowledge model, inputting the search label into the probability knowledge model to obtain feature knowledge data and a corresponding recommendation probability; then matching the search label with the feature knowledge data, comparing the label probability of the search label with the recommendation probability of the matched feature knowledge data; if the label probability of the search label is greater than the recommendation probability of the feature knowledge data, the feature knowledge data is taken as knowledge to be recommended; then based on the knowledge to be recommended, knowledge is recommended to the user. Through the above-mentioned mode, the key word search algorithm in the prior art can usually only match accurate key words, and cannot handle semantic correlation problems such as synonyms, synonyms and the like, thereby improving the accuracy of search.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a recommended method, apparatus, device, and storage medium. Background Technology

[0002] In the field of state-owned asset supervision, there are currently some big data-based supervision systems. These systems typically collect and organize large amounts of data related to state-owned asset supervision, and then use keyword retrieval to provide relevant information. These systems usually use traditional databases to store data and use keyword matching-based retrieval algorithms to find relevant information.

[0003] However, traditional keyword-matching-based retrieval algorithms have some limitations. They can only retrieve results based on keywords provided by the user and cannot accurately understand the user's intent and needs. Therefore, the accuracy of retrieval is not high. Summary of the Invention

[0004] In view of the above problems, embodiments of the present invention are proposed to provide a recommended method, apparatus, device and storage medium that overcomes or at least partially solves the above problems.

[0005] To address the above problems, this invention discloses a recommended method, the method comprising:

[0006] Obtain the user's search requirements;

[0007] Based on the search requirements, search tags and corresponding tag probabilities are generated; the search tags are used to describe the search requirements; the tag probabilities represent the probability that the user expects to use this search tag.

[0008] Obtain a probabilistic knowledge model by inputting the search tags into the probabilistic knowledge model to obtain feature knowledge data and corresponding recommendation probabilities;

[0009] The search tags are matched with the feature knowledge data, and the tag probability of the search tags is compared with the recommendation probability of the matched feature knowledge data.

[0010] If the probability of the retrieved tag is greater than the recommendation probability of the feature knowledge data, then the feature knowledge data is used as the knowledge to be recommended.

[0011] Based on the knowledge to be recommended, knowledge is recommended to the user.

[0012] Optionally, recommending knowledge to the user based on the knowledge to be recommended includes:

[0013] Based on the tag probability of the search tag and the preset segmentation threshold, the search tag is divided into multiple tag sets;

[0014] Combine the recommended knowledge corresponding to the retrieved tags in the same tag set to obtain the recommended knowledge set;

[0015] Based on the set of knowledge to be recommended, knowledge is recommended to the user.

[0016] Optionally, recommending knowledge to the user based on the set of knowledge to be recommended includes:

[0017] The set of knowledge to be recommended is sorted according to the segment order corresponding to the tag set to obtain the sorting result of the set of knowledge to be recommended;

[0018] Based on the ranking of the knowledge set to be recommended, knowledge is recommended to the user.

[0019] Optionally, the probabilistic knowledge model is trained in the following manner:

[0020] Obtain the business data and the type tags corresponding to the business data;

[0021] Acquire business knowledge and business rules;

[0022] Based on the business knowledge, the business rules, the business data, and the type labels corresponding to the business data, feature knowledge data is extracted to obtain the feature knowledge data and the type labels corresponding to the feature knowledge data.

[0023] Obtain the initial probability for the feature knowledge data;

[0024] Based on the feature knowledge data, the type labels corresponding to the feature knowledge data, and the initial probability, a preset probabilistic knowledge model is trained to obtain the probabilistic knowledge model.

[0025] Optionally, the business data and the corresponding type tags include:

[0026] Obtain business data;

[0027] The business data is identified based on a preset labeling model to determine its type and generate a type label.

[0028] Optionally, the preset labeling model is trained in the following way:

[0029] Acquire historical business data, including historical business knowledge and historical business rules;

[0030] Based on the historical business data and preset labels, the initial labeling model is trained to obtain the trained labeling model.

[0031] Optionally, generating search tags and corresponding tag probabilities based on the search requirements includes:

[0032] Obtain the preset demand transformation model;

[0033] Input the search requirements into the requirement transformation model to obtain the search tags and their corresponding tag probabilities.

[0034] Accordingly, embodiments of the present invention disclose a recommendation device, the device comprising:

[0035] The requirement acquisition module is used to obtain users' search requirements;

[0036] The tag generation module is used to generate search tags and corresponding tag probabilities based on the search requirements; the search tags are used to describe the search requirements; the tag probabilities represent the probability that the user expects to use this search tag;

[0037] The knowledge generation module is used to acquire a probabilistic knowledge model, input the search tags into the probabilistic knowledge model, and obtain feature knowledge data and corresponding recommendation probabilities.

[0038] The probability comparison module is used to match the search tag with the feature knowledge data and compare the tag probability of the search tag with the recommendation probability of the matched feature knowledge data.

[0039] The determination module is used to determine the feature knowledge data as the knowledge to be recommended if the probability of the retrieved tag is greater than the recommendation probability of the feature knowledge data.

[0040] The recommendation module is used to recommend knowledge to users based on the knowledge to be recommended.

[0041] Optionally, the recommendation module includes:

[0042] The splitting submodule is used to divide the search tags into multiple tag sets based on the tag probability of the search tags and a preset segmentation threshold;

[0043] The combination submodule is used to combine the knowledge to be recommended corresponding to the retrieved tags in the same tag set to obtain a set of knowledge to be recommended;

[0044] The recommendation submodule is used to recommend knowledge to users based on the set of knowledge to be recommended.

[0045] Optionally, the recommendation submodule includes:

[0046] The sorting unit is used to sort the knowledge set to be recommended according to the segment order corresponding to the tag set, so as to obtain the sorting result of the knowledge set to be recommended;

[0047] The recommendation unit is used to recommend knowledge to users based on the ranking of the knowledge set to be recommended.

[0048] Optionally, the knowledge generation module includes:

[0049] The first acquisition submodule is used to acquire business data and the type tags corresponding to the business data;

[0050] The second acquisition submodule is used to acquire business knowledge and business rules;

[0051] The feature knowledge determination submodule is used to extract feature knowledge data based on the business knowledge, the business rules, the business data, and the type labels corresponding to the business data, so as to obtain the feature knowledge data and the type labels corresponding to the feature knowledge data.

[0052] The third acquisition submodule is used to acquire the initial probability for the feature knowledge data;

[0053] The training submodule is used to train a preset probabilistic knowledge model based on the feature knowledge data, the type labels corresponding to the feature knowledge data, and the initial probability, so as to obtain the probabilistic knowledge model.

[0054] Optionally, the first acquisition submodule includes:

[0055] The acquisition unit is used to acquire business data;

[0056] The generation unit is used to identify the business data based on a preset labeling model, determine the type of the business data, and generate a type label.

[0057] Optionally, the generation unit includes:

[0058] The acquisition subunit is used to acquire historical business data, which includes historical business knowledge and historical business rules.

[0059] The training subunit is used to train the initial labeling model based on the historical business data and preset labels to obtain the trained labeling model.

[0060] Optionally, the tag generation module includes:

[0061] The requirement transformation model acquisition submodule is used to acquire a preset requirement transformation model;

[0062] The retrieval tag generation submodule is used to input the retrieval requirements into the requirement transformation model to obtain the retrieval tags and their corresponding tag probabilities.

[0063] Accordingly, embodiments of the present invention disclose an electronic device, including: a processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the various steps of the above-described recommended method embodiments.

[0064] Accordingly, embodiments of the present invention disclose a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the various steps of the above-described recommended method embodiments.

[0065] The embodiments of this invention have the following advantages: The recommendation method provided by these embodiments requires obtaining the user's search needs; then, based on the search needs, generating search tags and corresponding tag probabilities; the search tags describe the search needs; the tag probability represents the probability that the user expects to use this search tag; then, obtaining a probabilistic knowledge model, inputting the search tags into the probabilistic knowledge model, obtaining feature knowledge data and corresponding recommendation probabilities; then, matching the search tags with the feature knowledge data, comparing the tag probability of the search tags with the recommendation probability of the matching feature knowledge data; if the tag probability of the search tags is greater than the recommendation probability of the feature knowledge data, then the feature knowledge data is used as the knowledge to be recommended; then, based on the knowledge to be recommended, knowledge is recommended to the user. Through the above method, generating search tags based on the user's search needs and inputting them into the probabilistic knowledge model can solve the problem that existing keyword retrieval algorithms can usually only match precise keywords and cannot handle semantic relevance issues such as synonyms and near-synonyms, thereby improving the accuracy of retrieval. Attached Figure Description

[0066] Figure 1 This is a flowchart of the steps of the recommended method in the embodiments of the present invention;

[0067] Figure 2 This is a schematic diagram of the recommended method in an embodiment of the present invention;

[0068] Figure 3 This is a structural block diagram of the recommended device in the embodiments of the present invention;

[0069] Figure 4 This is a structural block diagram of an electronic device provided in an embodiment of the present invention;

[0070] Figure 5 This is a structural block diagram of a computer-readable storage medium provided in an embodiment of the present invention. Detailed Implementation

[0071] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0072] One of the core concepts of this invention is that the recommendation method provided by this invention requires obtaining the user's search needs; then, based on the search needs, generating search tags and corresponding tag probabilities; the search tags describe the search needs; the tag probability represents the probability that the user expects to use this search tag; then, obtaining a probabilistic knowledge model, inputting the search tags into the probabilistic knowledge model, obtaining feature knowledge data and corresponding recommendation probabilities; then, matching the search tags with the feature knowledge data, comparing the tag probability of the search tags with the recommendation probability of the matching feature knowledge data; if the tag probability of the search tags is greater than the recommendation probability of the feature knowledge data, then the feature knowledge data is used as the knowledge to be recommended; then, based on the knowledge to be recommended, knowledge is recommended to the user. Through the above method, generating search tags based on the user's search needs and inputting them into the probabilistic knowledge model can solve the problem that existing keyword retrieval algorithms can usually only match precise keywords and cannot handle semantic relevance issues such as synonyms and near-synonyms, thereby improving the accuracy of retrieval.

[0073] In practical applications, the following technical problems exist in the field of state-owned asset supervision:

[0074] Information retrieval accuracy is an issue. Existing keyword retrieval algorithms typically only match exact keywords and cannot handle semantic relevance issues such as synonyms and near-synonyms, resulting in low accuracy of search results. Traditional keyword retrieval algorithms primarily retrieve documents or data containing user-input keywords from a database. However, in the field of state-owned asset supervision, there are often numerous synonyms, near-synonyms, and industry-specific terms and abbreviations, which are semantically related. Traditional keyword retrieval algorithms cannot understand this semantic relevance and can only perform searches based on exact keyword matching, leading to low accuracy of search results. For example, in the field of state-owned asset supervision, a user might search for the keyword "corporate profits," but documents or data in the database might use the synonym "corporate profits" or the near-synonym "corporate revenue." Traditional keyword retrieval algorithms cannot match these related terms, resulting in the inability to retrieve relevant documents or data, thus affecting the accuracy of search results.

[0075] Personalization is a critical issue in information recommendation. Existing recommendation algorithms typically rely solely on users' historical behavior and preferences, failing to adequately consider real-time needs and individual user characteristics. Each user possesses unique interests, preferences, and tastes, but current algorithms primarily rely on group behavior and preferences, neglecting the individual characteristics of each user. This results in poor accuracy and a suboptimal user experience. The inability to fully consider real-time needs and individual characteristics leads to low accuracy in existing recommendation algorithms. Recommendations may not match users' actual needs, reducing user satisfaction. Poor accuracy negatively impacts user experience. Frequently receiving recommendations that don't align with their needs can cause distrust of the recommendation system, decreasing user acceptance and satisfaction.

[0076] Data processing efficiency issues. Current data processing technologies typically use traditional database storage methods, which are inefficient for handling large-scale data, resulting in low data processing efficiency.

[0077] Existing keyword retrieval algorithms cannot handle semantic relevance issues, resulting in low accuracy of search results. Recommendation algorithms also fail to fully consider users' real-time needs and personalized characteristics, leading to poor accuracy of recommendation results and a subpar user experience. Furthermore, existing data processing methods are inefficient in handling large-scale datasets, resulting in low data processing efficiency.

[0078] Reference Figure 1 The flowchart illustrates the steps of the recommended method in an embodiment of the present invention, which may specifically include the following steps:

[0079] Step 101: Obtain the user's search requirements.

[0080] Specifically, it involves obtaining users' search needs, which refers to the description or expression of the information required by users. For example, users can express their needs through keywords, search conditions, questions, etc.

[0081] This invention can be applied to state-owned asset supervision departments, financial institutions, investment institutions, research institutions, and academia. The recommended method of this invention enables users to quickly obtain information related to state-owned asset supervision, including enterprise information, financial data, and market data. This information can provide users with decision support and business guidance, helping them to conduct risk assessment, asset management, investment decisions, and research analysis, thereby improving work efficiency and decision-making accuracy.

[0082] State-owned asset regulatory authorities can utilize the intelligent recommendation and intelligent retrieval functions provided in this invention to quickly acquire and analyze big data related to state-owned asset supervision, providing scientific and accurate support for decision-making. Financial institutions can use the intelligent recommendation and intelligent retrieval functions provided in this invention to conduct risk assessments and credit ratings for state-owned enterprises. By intelligently recommending relevant regulatory indicators and risk warning information, financial institutions can more accurately assess the risk level of enterprises, thereby better managing risks and formulating loan policies. Enterprises can use the intelligent recommendation and intelligent retrieval functions provided in this invention to obtain market data and industry indicators related to state-owned asset supervision, providing scientific and accurate support for their business decisions. The intelligent recommendation of suitable business strategies and market opportunities helps enterprises optimize their business decisions and improve their competitiveness. Research institutions, academia, and consulting firms can use the intelligent recommendation and intelligent retrieval functions provided in this invention to conduct in-depth research and analysis in the field of state-owned asset supervision. By intelligently recommending relevant regulatory indicators and rules, research institutions and academia can better understand and interpret the patterns and trends of state-owned asset supervision, providing references for policy formulation and decision-making.

[0083] Step 102: Based on the search requirements, generate search tags and corresponding tag probabilities; the search tags are used to describe the search requirements; the tag probabilities represent the probability that the user expects to use this search tag.

[0084] Specifically, after obtaining the user's search requirements, keywords are extracted based on these requirements. Keywords can be words entered by the user in the search box or important terms extracted from the question description. These extracted keywords are used as tags, and intelligent association is performed on the tags to classify them and generate search tags and their corresponding probabilities. Classification algorithms or machine learning models can be used to categorize the user's search requirements and generate corresponding category tags. Attribute extraction techniques are used to extract attributes as tags. For example, attributes can be specific attributes mentioned by the user in the question description, such as time, location, and price. Furthermore, the generated search tags and their corresponding probabilities need to be converted into a form that can be used to intelligently recommend knowledge to the user. There can be multiple generated search tags and their corresponding probabilities; for example, the generated search tags and their corresponding probabilities could be ["funds", 5 / 8] and ["profit", 3 / 8].

[0085] Step 103: Obtain the probabilistic knowledge model by inputting the search tag into the probabilistic knowledge model to obtain feature knowledge data and the corresponding recommendation probability.

[0086] Specifically, a probabilistic knowledge model is obtained, which can output feature knowledge data, the corresponding tags for the feature knowledge data, and the corresponding recommendation probabilities. Inputting the search tags obtained based on the user's search needs into the probabilistic knowledge model yields multiple pieces of feature knowledge data, the corresponding tags for the feature knowledge data, and the corresponding recommendation probabilities. For example, the feature knowledge data, the corresponding tags for the feature knowledge data, and the corresponding recommendation probabilities could be ["Enterprise Large Fund Approval", "Funds", 5 / 8].

[0087] Step 104: Match the search tag with the feature knowledge data, and compare the tag probability of the search tag with the recommendation probability of the matching feature knowledge data.

[0088] Specifically, the search tags are matched with feature knowledge data to obtain feature knowledge data that matches the search tags. Then, the tag probability corresponding to the search tag is compared with the recommendation probability of the feature knowledge data that matches the search tag.

[0089] Step 105: If the probability of the retrieved tag is greater than the recommendation probability of the feature knowledge data, then the feature knowledge data is used as the knowledge to be recommended.

[0090] Specifically, if the probability of a search tag is greater than the recommendation probability of the feature knowledge data, then the feature knowledge data is considered as recommended knowledge. For example, if the recommendation probability of a piece of feature knowledge data is 0.5, and the probability of a search tag matching this feature knowledge data is 0.6, then the probability of the search tag is greater than the recommendation probability of the feature knowledge data, and thus this feature knowledge data can be recommended, i.e., it is considered as recommended knowledge. If the recommendation probability of a piece of feature knowledge data is 0.6, and the probability of a search tag matching this feature knowledge data is 0.5, then the probability of the search tag is not greater than the recommendation probability of the feature knowledge data, and thus it is not recommended, i.e., this feature knowledge data is not considered as recommended knowledge.

[0091] Step 106: Based on the knowledge to be recommended, recommend knowledge to the user.

[0092] Specifically, the system compares the probability of the retrieved tags with the recommendation probability of the matching feature knowledge data to obtain the knowledge to be recommended, and then recommends knowledge to the user. This embodiment of the invention can intelligently match and recommend relevant state-owned asset supervision data according to the user's needs and query conditions, reducing the time and cost for users in the information retrieval process, improving user work efficiency and decision-making accuracy. By providing data and information related to user needs, intelligent recommendation technology can help users obtain the information they need more quickly, improving work efficiency and decision-making accuracy.

[0093] In this embodiment of the invention, recommending knowledge to the user based on the knowledge to be recommended includes:

[0094] Based on the tag probability of the search tag and the preset segmentation threshold, the search tag is divided into multiple tag sets;

[0095] Combine the recommended knowledge corresponding to the retrieved tags in the same tag set to obtain the recommended knowledge set;

[0096] Based on the set of knowledge to be recommended, knowledge is recommended to the user.

[0097] Specifically, after obtaining the knowledge to be recommended, the knowledge is divided into several segments based on the tag probability of the search tags that match the knowledge to be recommended, using methods such as taking the median or quartiles, to form a preset segment threshold. The search tags are then divided into multiple tag sets based on the preset segment threshold. Finally, the knowledge to be recommended corresponding to the search tags in the same tag set is combined to obtain the set of knowledge to be recommended.

[0098] In this embodiment of the invention, recommending knowledge to the user based on the set of knowledge to be recommended includes:

[0099] The set of knowledge to be recommended is sorted according to the segment order corresponding to the tag set to obtain the sorting result of the set of knowledge to be recommended;

[0100] Based on the ranking of the knowledge set to be recommended, knowledge is recommended to the user.

[0101] Specifically, after obtaining the set of knowledge to be recommended, the set of knowledge to be recommended is sorted according to the segment order corresponding to the tag set to obtain the sorting result of the set of knowledge to be recommended. Based on the sorting result of the set of knowledge to be recommended, the knowledge that the user needs is recommended to the user in a personalized and intelligent way.

[0102] This invention intelligently retrieves state-owned asset supervision data based on user query conditions, providing data and information relevant to user needs. Based on the user's query conditions and provided information, keywords are extracted, tags are intelligently associated, tags are categorized, search tags are generated, and relationships are established between tags to generate the user's search tags and corresponding tag probabilities. Based on the user's search tags and corresponding tag probabilities, feature knowledge data output by a probabilistic knowledge model is matched to obtain recommended knowledge. This recommended knowledge is then combined according to the tag probabilities of the search tags and preset segmentation thresholds. Utilizing the rules and correlations in the recommended knowledge and search tags, and sorting and outputting the recommended knowledge set based on the search tags, intelligent matching and retrieval of state-owned asset supervision data is performed, presenting user-relevant data and information to improve the efficiency and relevance of information acquisition.

[0103] In this embodiment of the invention, the probabilistic knowledge model is trained in the following manner:

[0104] Obtain the business data and the type tags corresponding to the business data;

[0105] Acquire business knowledge and business rules;

[0106] Based on the business knowledge, the business rules, the business data, and the type labels corresponding to the business data, feature knowledge data is extracted to obtain the feature knowledge data and the type labels corresponding to the feature knowledge data.

[0107] Obtain the initial probability for the feature knowledge data;

[0108] Based on the feature knowledge data, the type labels corresponding to the feature knowledge data, and the initial probability, a preset probabilistic knowledge model is trained to obtain the probabilistic knowledge model.

[0109] Specifically, the business data and the corresponding type tags are first obtained. The business data can be data related to the operation of the state-owned assets supervision platform, such as corporate operating profit, return on net assets, and debt-to-equity ratio.

[0110] Acquiring business knowledge and rules can involve relevant information and regulations within the state-owned assets supervision field. For example, this could include state-owned assets supervision standards and enterprise performance evaluation criteria. This can be achieved through cooperation with state-owned assets supervision departments, expert consultation, and literature research. By gaining a thorough understanding of state-owned assets supervision regulations and guidelines, one can acquire knowledge about enterprise financial data, operating indicators, industry classifications, and other related aspects.

[0111] Based on business knowledge and rules, knowledge extraction is performed to extract feature knowledge data, which is then represented in a form suitable for machine learning and inference. For example, key indicators and ratios in financial data can be extracted, and a network of relationships between enterprises can be established. The feature knowledge data is matched with the type labels corresponding to the business data to obtain the feature knowledge data and its corresponding type labels. Based on standard rules and empirical splitting rules, the feature knowledge-related data information is combined to obtain the feature knowledge data, its corresponding type labels, and the corresponding recommendation probabilities. This combined feature knowledge-related data is then used to train the model. This is a dynamic process that requires continuous updates and iterations. During training, the combined data can be evaluated using methods such as cross-validation and metric evaluation to ensure its accuracy and reliability. This is achieved through regularly updating the data, retraining the combined data, and evaluating the model's performance. Furthermore, the model can be improved and optimized based on new data and domain knowledge.

[0112] Initial probabilities are obtained for the feature knowledge data. Based on the feature knowledge data, its corresponding type labels, recommendation probabilities, and professional knowledge in the field of state-owned asset supervision, a pre-defined probabilistic knowledge model is constructed, and initial empirical values ​​are generated. This pre-defined probabilistic knowledge model contains some basic rules and knowledge that can be used for preliminary classification and labeling of state-owned asset supervision data. These initial empirical values ​​can be provided by professionals or extracted from existing data. The pre-defined probabilistic knowledge model is trained using machine learning algorithms. By using previously acquired business data and its corresponding type labels, the machine learning algorithm can automatically adjust the parameters and weights of the pre-defined probabilistic knowledge model based on the features and labels of the data, improving the model's accuracy and generalization ability. The training of empirical values ​​can be implemented using supervised learning, unsupervised learning, or semi-supervised learning methods. After training the pre-defined probabilistic knowledge model, the probabilistic knowledge model is calculated. The probabilistic knowledge model is an optimized and adjusted intelligent model. By setting and training knowledge probability thresholds, it performs probability matching for different types of knowledge, enabling a more accurate understanding and application of state-owned asset supervision data. Probabilistic models incorporate knowledge and rules learned from training data, as well as updates to empirical values.

[0113] In this embodiment of the invention, obtaining the business data and the type tag corresponding to the business data includes:

[0114] Obtain business data;

[0115] The business data is identified based on a preset labeling model to determine its type and generate a type label.

[0116] Specifically, business data is acquired by connecting to the state-owned assets supervision platform, large-scale model data, and relevant external data files. This business data can be data related to the operation of the state-owned assets supervision platform, such as enterprise operating profit, return on net assets, and debt-to-equity ratio—in other words, data related to the actual production and operation indicators of state-owned assets supervision enterprises. After acquiring the business data, it is preprocessed to form standardized data. A preset labeling model is then obtained, and the standardized data generated from the preprocessed business data is input into the preset labeling model. The preset labeling model can intelligently label the preprocessed data. Based on the content and context of the business data, the preset labeling model can automatically identify and affix appropriate labels, outputting the labeled business data and its corresponding type labels. For example, for financial data, labels such as "operating revenue" and "net profit" can be affixed; for enterprise names, labels such as "state-owned enterprise" and "listed company" can be affixed.

[0117] In this embodiment of the invention, the preset labeling model is trained in the following manner:

[0118] Acquire historical business data, including historical business knowledge and historical business rules;

[0119] Based on the historical business data and preset labels, the initial labeling model is trained to obtain the trained labeling model.

[0120] Specifically, historical business data is acquired, including historical business knowledge and rules. This data is preprocessed using machine learning and natural language processing technologies to train an initial labeling model. Based on the historical business data and pre-set labels, the initial labeling model is trained to obtain a trained model. This model can learn and understand relevant knowledge and rules in the field of state-owned asset supervision, as well as the meaning and correlation of financial data and operating indicators. Through training, the model can automatically identify and label different types of data. After labeling is completed, the labeling results are verified and corrected through a combination of intelligent verification by the machine learning model and manual review. If labeling errors or omissions are found, corrections can be made, and the model can be used again for labeling. After labeling is completed, all the final labeling data and related information are compiled into a label dataset and a tag set for output.

[0121] This invention utilizes a labeling model to automatically classify and label state-owned asset supervision data. This reduces the workload of manual labeling and improves data processing efficiency. The labeling model can quickly and accurately mark data, providing tagged data for subsequent processing and analysis, saving significant time and labor costs.

[0122] Based on the labeled data, probabilistic knowledge models for the state-owned assets supervision field are generated using inference algorithms and probabilistic models. These probabilistic knowledge models can extract key knowledge and rules from the data and transform them into a form that is understandable and applicable to state-owned assets supervision operations. The application of probabilistic knowledge models can improve the accuracy and reliability of data analysis and reduce the impact of human factors on the results.

[0123] Based on the generated probabilistic knowledge model, this invention can intelligently recommend relevant state-owned asset supervision data according to user needs and query conditions, helping users quickly obtain the information they need. It can intelligently retrieve state-owned asset supervision data based on user query conditions, providing data and information relevant to user needs. These intelligent functions can improve user work efficiency and reduce the time and cost of information retrieval.

[0124] This invention enhances the utilization value of state-owned asset supervision data. Intelligent labeling technology tags the data, making it easier to understand and apply. Probabilistic knowledge models extract key knowledge and rules from the data, helping users better understand and analyze it. Intelligent recommendation and retrieval functions provide personalized and accurate data recommendations and searches to meet specific user needs. The application of these technologies improves data utilization efficiency and value, helping users better utilize state-owned asset supervision data for decision-making and analysis. This invention effectively processes and analyzes state-owned asset supervision data, improving its utilization efficiency and value by enhancing data processing efficiency, increasing data analysis accuracy, providing intelligent recommendation and retrieval functions, and enhancing data utilization value.

[0125] In this embodiment of the invention, generating search tags and corresponding tag probabilities based on the search requirements includes:

[0126] Obtain the preset demand transformation model;

[0127] Input the search requirements into the requirement transformation model to obtain the search tags and their corresponding tag probabilities.

[0128] Specifically, a pre-defined demand transformation model is obtained. User search demands are input into this model, which then outputs search tags and their probabilities. User search demands refer to a user's description or expression of the required information, which can be expressed through keywords, query conditions, questions, etc. User search demands are processed and identified through keyword extraction and question analysis. Based on the information provided by the user, classification algorithms or machine learning models are used to categorize the user's search demands and generate corresponding category tags. Attribute extraction techniques are used to extract attributes as tags. Attributes can be specific attributes mentioned by the user in the question description, such as time, location, and price. Relationships are established between the identified and parsed tags through tag classification and attribute extraction methods. Knowledge graphs or association rule mining techniques can be used to establish these relationships.

[0129] Reference Figure 2 The diagram illustrates a recommendation method in an embodiment of the invention. Taking the field of state-owned asset supervision as an example, the recommendation method is described exemplarily, and may specifically include the following steps:

[0130] Intelligent labeling. Business data is acquired through API interfaces, databases, large models, and related files, and then labeled to obtain the business data and its corresponding tags. All the final labeled data and related information are combined into a label dataset and a tag set. State-owned asset supervision data is automatically labeled and classified using predefined rules and models. The system analyzes and processes the input state-owned asset supervision data and related information according to pre-set tags and classification rules, automatically labeling it as the corresponding category or attribute, generating a tag set and dataset for state-owned asset supervision data. This reduces the workload and error rate of manual labeling, improving the accuracy and efficiency of data processing.

[0131] Probabilistic knowledge model generation. This involves extracting key knowledge and rules from the data, extracting feature labels, and extracting relevant knowledge content to generate a knowledge sequence. By analyzing and mining the labeled data, information such as correlations, patterns, and importance is extracted. Based on state-owned asset supervision standards and rules, and empirical decomposition rules, a probabilistic knowledge model is assembled and generated, which is the initial knowledge model X in the diagram. The output of the probabilistic knowledge model is feature knowledge data, labels, and corresponding recommendation probabilities. Data M and data N in the diagram both belong to feature knowledge data, and the empirical value X in the diagram refers to the recommendation probability.

[0132] Intelligent Recommendation. Based on the generated initial knowledge model, by analyzing user needs and query conditions, an initial knowledge model is first generated. Through model training, the knowledge model is assigned training experience values. Based on the experience knowledge model, a probabilistic knowledge model is generated through probability model calculation, which is the experience knowledge model Y in the diagram. The knowledge model and experience knowledge model are then matched with the user's needs in a personalized and intelligent manner. Based on the matching results, relevant regulatory measures and suggestions are intelligently recommended. Intelligent recommendation can intelligently recommend state-owned asset supervision data based on the experience knowledge model. For example, when a user needs to query the financial data of a certain company, intelligent recommendation can intelligently generate a financial knowledge model of that company from the database based on the company's query tag model. Combining this with historical training experience values, a financial operation knowledge model and a probabilistic model of the company are generated, and targeted recommendations are made based on the tag model. Intelligent recommendation can improve user work efficiency and reduce the time and cost of information retrieval.

[0133] Intelligent retrieval. Based on the user's query conditions, tags are split and modeled to form a tag model, which is the retrieval tag. The tag model and knowledge model are combined according to weights and segmentation thresholds, that is, according to the tag probability of the retrieval tag and a preset segmentation threshold. The knowledge model is then sorted and output based on the user's tag model. Embodiments of this invention implement intelligent retrieval functionality. Intelligent retrieval can intelligently retrieve state-owned asset supervision data based on the user's query conditions using a knowledge model. Users can input keywords, conditions, or questions, and intelligent retrieval will intelligently match and retrieve data and information relevant to the user's needs based on the knowledge model. Intelligent retrieval can improve the user's information acquisition efficiency and help users better understand and analyze state-owned asset supervision data.

[0134] This invention utilizes technologies such as intelligent labeling, probabilistic knowledge model generation, intelligent recommendation, and intelligent retrieval to process and analyze state-owned asset supervision data. These technologies automatically classify and label data, generate probabilistic knowledge models that can be understood and applied by state-owned asset supervision operations, and provide intelligent recommendation and retrieval functions, thereby improving the utilization value and efficiency of state-owned asset supervision data. State-owned asset supervision departments can use the recommendation method of this invention to quickly obtain information related to state-owned asset supervision and make personalized recommendations based on individual preferences and historical behavior, improving work efficiency and decision-making accuracy.

[0135] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of the present invention are not limited to the described order of actions, because according to the embodiments of the present invention, some steps can be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions involved are not necessarily essential to the embodiments of the present invention.

[0136] Reference Figure 3 The diagram illustrates a structural block diagram of a recommended embodiment of the present invention, which may specifically include the following modules:

[0137] The requirement acquisition module 201 is used to acquire users' search requirements.

[0138] The tag generation module 202 is used to generate search tags and corresponding tag probabilities based on the search requirements; the search tags are used to describe the search requirements; the tag probabilities represent the probability that the user expects to use this search tag;

[0139] The knowledge generation module 203 is used to obtain a probabilistic knowledge model, input the search tag into the probabilistic knowledge model, and obtain feature knowledge data and corresponding recommendation probabilities;

[0140] The probability comparison module 204 is used to match the search tag with the feature knowledge data and compare the tag probability of the search tag with the recommendation probability of the matched feature knowledge data.

[0141] The determining module 205 is used to determine the feature knowledge data as the knowledge to be recommended if the probability of the retrieved tag is greater than the recommendation probability of the feature knowledge data.

[0142] The recommendation module 206 is used to recommend knowledge to the user based on the knowledge to be recommended.

[0143] In this embodiment of the invention, the recommendation module includes:

[0144] The splitting submodule is used to divide the search tags into multiple tag sets based on the tag probability of the search tags and a preset segmentation threshold;

[0145] The combination submodule is used to combine the knowledge to be recommended corresponding to the retrieved tags in the same tag set to obtain a set of knowledge to be recommended;

[0146] The recommendation submodule is used to recommend knowledge to users based on the set of knowledge to be recommended.

[0147] In this embodiment of the invention, the recommendation submodule includes:

[0148] The sorting unit is used to sort the knowledge set to be recommended according to the segment order corresponding to the tag set, so as to obtain the sorting result of the knowledge set to be recommended;

[0149] The recommendation unit is used to recommend knowledge to users based on the ranking of the knowledge set to be recommended.

[0150] In this embodiment of the invention, the knowledge generation module includes:

[0151] The first acquisition submodule is used to acquire business data and the type tags corresponding to the business data;

[0152] The second acquisition submodule is used to acquire business knowledge and business rules;

[0153] The feature knowledge determination submodule is used to extract feature knowledge data based on the business knowledge, the business rules, the business data, and the type labels corresponding to the business data, so as to obtain the feature knowledge data and the type labels corresponding to the feature knowledge data.

[0154] The third acquisition submodule is used to acquire the initial probability for the feature knowledge data;

[0155] The training submodule is used to train a preset probabilistic knowledge model based on the feature knowledge data, the type labels corresponding to the feature knowledge data, and the initial probability, so as to obtain the probabilistic knowledge model.

[0156] In this embodiment of the invention, the first acquisition submodule includes:

[0157] The acquisition unit is used to acquire business data;

[0158] The generation unit is used to identify the business data based on a preset labeling model, determine the type of the business data, and generate a type label.

[0159] In this embodiment of the invention, the generating unit includes:

[0160] The acquisition subunit is used to acquire historical business data, which includes historical business knowledge and historical business rules.

[0161] The training subunit is used to train the initial labeling model based on the historical business data and preset labels to obtain the trained labeling model.

[0162] In this embodiment of the invention, the label generation module includes:

[0163] The requirement transformation model acquisition submodule is used to acquire a preset requirement transformation model;

[0164] The retrieval tag generation submodule is used to input the retrieval requirements into the requirement transformation model to obtain the retrieval tags and their corresponding tag probabilities.

[0165] The recommendation method provided in this invention requires obtaining the user's search needs; then, based on the search needs, generating search tags and corresponding tag probabilities; the search tags describe the search needs; the tag probability represents the probability that the user expects to use this search tag; next, obtaining a probabilistic knowledge model, inputting the search tags into the probabilistic knowledge model, obtaining feature knowledge data and corresponding recommendation probabilities; then, matching the search tags with the feature knowledge data, comparing the tag probability of the search tags with the recommendation probability of the matching feature knowledge data; if the tag probability of the search tags is greater than the recommendation probability of the feature knowledge data, then the feature knowledge data is used as the knowledge to be recommended; then, based on the knowledge to be recommended, knowledge is recommended to the user. Through the above method, generating search tags based on the user's search needs and inputting them into the probabilistic knowledge model can solve the problem that existing keyword retrieval algorithms can usually only match precise keywords and cannot handle semantic relevance issues such as synonyms and near-synonyms, thereby improving the accuracy of retrieval.

[0166] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0167] like Figure 4 The diagram illustrates a structural block diagram of an electronic device provided in an embodiment of the present invention, comprising:

[0168] It includes a processor, a memory, and a computer program stored in the memory and capable of running on the processor. When the computer program is executed by the processor, it implements the various processes of the above-described recommended method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here.

[0169] like Figure 5 The diagram illustrates a structural block diagram of a computer-readable storage medium provided in this embodiment of the invention. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the various processes of the above-described recommended method embodiments and achieves the same technical effects. To avoid repetition, it will not be described again here.

[0170] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0171] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0172] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0173] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0174] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0175] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0176] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0177] The above provides a detailed description of a recommended method, apparatus, device, and storage medium provided by the present invention. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A recommendation method, characterized in that, The method includes: Obtain the user's search requirements; Based on the search requirements, search tags and corresponding tag probabilities are generated; the search tags are used to describe the search requirements; the tag probabilities represent the probability that the user expects to use this search tag. Obtain a probabilistic knowledge model by inputting the search tags into the probabilistic knowledge model to obtain feature knowledge data and corresponding recommendation probabilities; The search tags are matched with the feature knowledge data, and the tag probability of the search tags is compared with the recommendation probability of the matched feature knowledge data. If the probability of the retrieved tag is greater than the recommendation probability of the feature knowledge data, then the feature knowledge data is used as the knowledge to be recommended. Based on the knowledge to be recommended, knowledge is recommended to the user.

2. The method according to claim 1, characterized in that, The process of recommending knowledge to the user based on the knowledge to be recommended includes: Based on the tag probability of the search tag and the preset segmentation threshold, the search tag is divided into multiple tag sets; Combine the recommended knowledge corresponding to the retrieved tags in the same tag set to obtain the recommended knowledge set; Based on the set of knowledge to be recommended, knowledge is recommended to the user.

3. The method according to claim 2, characterized in that, The process of recommending knowledge to users based on the set of knowledge to be recommended includes: The set of knowledge to be recommended is sorted according to the segment order corresponding to the tag set to obtain the sorting result of the set of knowledge to be recommended; Based on the ranking of the knowledge set to be recommended, knowledge is recommended to the user.

4. The method according to claim 1, characterized in that, The probabilistic knowledge model is trained in the following manner: Obtain the business data and the type tags corresponding to the business data; Acquire business knowledge and business rules; Based on the business knowledge, the business rules, the business data, and the type labels corresponding to the business data, feature knowledge data is extracted to obtain the feature knowledge data and the type labels corresponding to the feature knowledge data. Obtain the initial probability for the feature knowledge data; Based on the feature knowledge data, the type labels corresponding to the feature knowledge data, and the initial probability, a preset probabilistic knowledge model is trained to obtain the probabilistic knowledge model.

5. The method according to claim 4, characterized in that, The acquisition of business data and the corresponding type tags for the business data includes: Obtain business data; The business data is identified based on a preset labeling model to determine its type and generate a type label.

6. The method according to claim 5, characterized in that, The preset labeling model is trained in the following way: Acquire historical business data, including historical business knowledge and historical business rules; Based on the historical business data and preset labels, the initial labeling model is trained to obtain the trained labeling model.

7. The method according to claim 1, characterized in that, The step of generating search tags and corresponding tag probabilities based on the search requirements includes: Obtain the preset demand transformation model; Input the search requirements into the requirement transformation model to obtain the search tags and their corresponding tag probabilities.

8. A recommended device, characterized in that, The device includes: The requirement acquisition module is used to obtain users' search requirements; The tag generation module is used to generate search tags and corresponding tag probabilities based on the search requirements; the search tags are used to describe the search requirements; the tag probabilities represent the probability that the user expects to use this search tag; The knowledge generation module is used to acquire a probabilistic knowledge model, input the search tags into the probabilistic knowledge model, and obtain feature knowledge data and corresponding recommendation probabilities. The probability comparison module is used to match the search tag with the feature knowledge data and compare the tag probability of the search tag with the recommendation probability of the matched feature knowledge data. The determination module is used to determine the feature knowledge data as the knowledge to be recommended if the probability of the retrieved tag is greater than the recommendation probability of the feature knowledge data. The recommendation module is used to recommend knowledge to users based on the knowledge to be recommended.

9. An electronic device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and capable of running on the processor, wherein the computer program, when executed by the processor, implements the steps of the recommended method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, which, when executed by a processor, implements the steps of the recommended method as described in any one of claims 1-7.

Images