Automatic classification method, device and computer readable storage medium

Abstract

The embodiment of the present application discloses an automatic grading method, device and computer-readable storage medium, which can improve the accuracy of grading results. The method of the embodiment of the present application includes: constructing a policy and regulation ontology library and a typical case ontology library; inputting keywords in the policy and regulation ontology and the typical case ontology into a pre-trained word vector model to obtain multiple graded keyword vectors; obtaining graded query information; inputting the graded query information into the word vector model to obtain a corresponding graded query vector, wherein each input word in the graded query information corresponds to a graded query vector; determining a keyword set corresponding to the graded query vector according to the graded query vector and the graded keyword vector, wherein the keyword set includes at least one keyword; inputting the keyword set into an inference engine to perform a forward rule reasoning query on the policy and regulation ontology library or the typical case ontology library, and obtaining a grading result output by the inference engine.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of data processing technology, and in particular, to an automatic grading method, device, and computer-readable storage medium. Background Art

[0002] The hierarchical management of government data resources is a key task in government data sharing and data governance. Hierarchical management helps improve the efficiency of government data resource utilization and promote cross-departmental collaborative sharing. However, due to the large scale and complex types of government data resources, manual classification is inefficient and difficult to cope with the rapidly growing amount of data. In addition, there is a lack of professional hierarchical management tools and technical support, resulting in subjectivity and uncertainty in the classification results. Summary of the invention

[0003] The embodiments of the present application provide an automatic grading method, device, and computer-readable storage medium, which can improve the accuracy of grading results.

[0004] A first aspect of an embodiment of the present application provides an automatic grading method, including:

[0005] Constructing a policy and regulation ontology library and a typical case ontology library, wherein the policy and regulation ontology library stores a policy and regulation ontology having a graph data structure, and the typical case ontology library stores a typical case ontology having a graph data structure;

[0006] Input keywords in the policy and regulation ontology and the typical case ontology into a pre-trained word vector model to obtain multiple hierarchical keyword vectors;

[0007] Acquire hierarchical query information, where the hierarchical query information is query data input by a user, and the hierarchical query information includes at least two input words;

[0008] Inputting the hierarchical query information into the word vector model to obtain a corresponding hierarchical query vector, wherein each input word in the hierarchical query information corresponds to a hierarchical query vector;

[0009] Determine a keyword set corresponding to the hierarchical query vector according to the hierarchical query vector and the hierarchical keyword vector, wherein the keyword set includes at least one keyword;

[0010] The keyword set is input into the inference engine to perform forward rule reasoning query on the policy and regulation ontology library or the typical case ontology library, and the classification result output by the inference engine is obtained.

[0011] Optionally, inputting the hierarchical query information into the word vector model to obtain a corresponding hierarchical query vector includes:

[0012] Inputting each input word in the hierarchical query information into the word vector model in sequence;

[0013] Using the multi-view learning sub-model of the word vector model to extract features from each input word, and obtain sparse feature vectors under multiple views;

[0014] Using the multi-view learning sub-model to fuse the sparse feature vectors to obtain a comprehensive word vector;

[0015] Perform quantum state superposition on the comprehensive word vector to obtain a quantum state vector;

[0016] The quantum state vector is optimized using a generative adversarial network of the word vector model to obtain a hierarchical query vector.

[0017] Optionally, the construction of a policy and regulation ontology library and a typical case ontology library includes:

[0018] Use ontology building tools to extract the ontology of government information disclosure catalogue and obtain the first policy and regulation ontology;

[0019] Storing the first policy and regulation ontology into a policy and regulation ontology library;

[0020] Using the ontology building tool to extract the ontology of typical cases on the government open data platform, and obtaining a first typical case ontology;

[0021] The first typical case ontology is stored in a typical case ontology library.

[0022] Optionally, the method further includes:

[0023] When the government information disclosure catalog is updated, obtaining first update information;

[0024] Using the ontology building tool to extract the ontology of the first update information to obtain a second policy and regulation ontology;

[0025] The second policy and regulation ontology is stored in the policy and regulation ontology library.

[0026] Optionally, the method further includes:

[0027] When the government open data platform is updated, obtaining second update information;

[0028] Using the ontology building tool to extract the ontology of the second update information to obtain a second typical case ontology;

[0029] The second typical case ontology is stored in the typical case ontology library.

[0030] Optionally, the determining, according to the hierarchical query vector and the hierarchical keyword vector, a keyword set corresponding to the hierarchical query vector includes:

[0031] respectively calculating the similarity between each of the hierarchical query vectors and the hierarchical keyword vectors;

[0032] For any of the hierarchical query vectors, taking the keyword corresponding to the maximum similarity as the hierarchical keyword corresponding to the hierarchical query vector;

[0033] A preset number of constituent keyword sets are screened from the hierarchical keywords.

[0034] Optionally, the grading result includes a sharing level, an open level and a grading basis;

[0035] The step of inputting the keyword set into the inference engine to perform forward rule reasoning query on the policy and regulation ontology library or the typical case ontology library, and obtaining the grading result output by the inference engine includes:

[0036] Inputting the keyword set into the inference engine to perform forward rule reasoning query on the policy and regulation ontology library;

[0037] If policies and regulations that include graded content and do not have conflicting clauses are found, the corresponding graded content will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; if policies and regulations that include graded content but have conflicting clauses are found, the graded content in the policies and regulations with the highest effectiveness will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; if policies and regulations that include graded content are not found, the inference engine will be used to perform forward rule reasoning query on the typical case ontology library based on the keyword set; if typical cases that include graded content are found, the corresponding graded content will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; wherein, the graded content includes two aspects: data sharing and data openness.

[0038] A second aspect of an embodiment of the present application provides an automatic grading device, comprising:

[0039] A construction unit, used to construct a policy and regulation ontology library and a typical case ontology library, wherein the policy and regulation ontology library stores a policy and regulation ontology having a graph data structure, and the typical case ontology library stores a typical case ontology having a graph data structure;

[0040] A first input unit is used to input keywords in the policy and regulation ontology and the typical case ontology into a pre-trained word vector model to obtain multiple hierarchical keyword vectors;

[0041] An acquisition unit, used to acquire hierarchical query information, where the hierarchical query information is query data input by a user;

[0042] A second input unit, configured to input the hierarchical query information into the word vector model to obtain a corresponding hierarchical query vector, wherein each input word in the hierarchical query information corresponds to a hierarchical query vector;

[0043] a determining unit, configured to determine a keyword set corresponding to the hierarchical query vector according to the hierarchical query vector and the hierarchical keyword vector, wherein the keyword set includes at least one keyword;

[0044] The reasoning unit is used to input the keyword set into the reasoning engine to perform forward rule reasoning query on the policy and regulation ontology library or the typical case ontology library to obtain the grading result output by the reasoning engine.

[0045] Optionally, the grading result includes a sharing level, an open level and a grading basis; and the reasoning unit is specifically used for:

[0046] Inputting the keyword set into the inference engine to perform forward rule reasoning query on the policy and regulation ontology library;

[0047] If policies and regulations that include graded content and do not have conflicting clauses are found, the corresponding graded content will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; if policies and regulations that include graded content but have conflicting clauses are found, the graded content in the policies and regulations with the highest effectiveness will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; if policies and regulations that include graded content are not found, the inference engine will be used to perform forward rule reasoning query on the typical case ontology library based on the keyword set; if typical cases that include graded content are found, the corresponding graded content will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; wherein, the graded content includes two aspects: data sharing and data openness.

[0048] Optionally, the second input unit is specifically used for:

[0049] Inputting each input word in the hierarchical query information into the word vector model in sequence;

[0050] Using the multi-view learning sub-model of the word vector model to extract features from each input word, and obtain sparse feature vectors under multiple views;

[0051] Using the multi-view learning sub-model to fuse the sparse feature vectors to obtain a comprehensive word vector;

[0052] Perform quantum state superposition on the comprehensive word vector to obtain a quantum state vector;

[0053] The quantum state vector is optimized using a generative adversarial network of the word vector model to obtain a hierarchical query vector.

[0054] Optionally, the construction unit is specifically used for:

[0055] Use ontology building tools to extract the ontology of government information disclosure catalogue and obtain the first policy and regulation ontology;

[0056] Storing the first policy and regulation ontology into a policy and regulation ontology library;

[0057] Using the ontology building tool to extract the ontology of typical cases on the government open data platform, and obtaining a first typical case ontology;

[0058] The first typical case ontology is stored in a typical case ontology library.

[0059] The apparatus further includes a first updating unit, wherein the first updating unit is configured to:

[0060] When the government information disclosure catalog is updated, obtaining first update information;

[0061] Using the ontology building tool to extract the ontology of the first update information to obtain a second policy and regulation ontology;

[0062] The second policy and regulation ontology is stored in the policy and regulation ontology library.

[0063] The apparatus further includes a second updating unit, wherein the second updating unit is configured to:

[0064] When the government open data platform is updated, obtaining second update information;

[0065] Using the ontology building tool to extract the ontology of the second update information to obtain a second typical case ontology;

[0066] The second typical case ontology is stored in the typical case ontology library.

[0067] Optionally, the determining unit is specifically configured to:

[0068] respectively calculating the similarity between each of the hierarchical query vectors and the hierarchical keyword vectors;

[0069] For any of the hierarchical query vectors, taking the keyword corresponding to the maximum similarity as the hierarchical keyword corresponding to the hierarchical query vector;

[0070] A preset number of constituent keyword sets are screened from the hierarchical keywords.

[0071] A third aspect of the embodiments of the present application provides an automatic grading device, comprising:

[0072] Processor, memory, input-output unit, and bus;

[0073] The processor is connected to the memory, the input and output unit, and the bus;

[0074] A program is stored in the memory, and the processor calls the program to execute the method in the first aspect and any possible implementation manner of the first aspect.

[0075] A fourth aspect of an embodiment of the present application provides a computer-readable storage medium, on which a program is stored. When the program is executed on a computer, the computer executes the method in the first aspect and any possible implementation of the first aspect.

[0076] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0077] The embodiment of the present application can accurately match the user's hierarchical query with the keywords in the policy and regulation ontology library and the typical case ontology library, and generate relevant results using the forward reasoning mechanism. This process improves the accuracy and intelligence level of the hierarchical results, while ensuring that the system can respond to user needs in real time and provide users with more accurate and valuable information. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 This is a flow chart of an embodiment of the automatic grading method in the embodiments of the present application;

[0079] Figure 2 A schematic diagram of a flow chart of an embodiment of inputting hierarchical query information into a word vector model in an embodiment of the present application;

[0080] Figure 3 A flowchart of an embodiment of constructing a policy and regulation ontology library and a typical case ontology library in an embodiment of the present application;

[0081] Figure 4 A schematic diagram of a process for updating a policy and regulation ontology library in an embodiment of the present application;

[0082] Figure 5 A schematic diagram of a process for updating a typical case ontology library in an embodiment of the present application;

[0083] Figure 6 A schematic diagram of a flow chart of an embodiment of determining a keyword set in an embodiment of the present application;

[0084] Figure 7 This is a schematic diagram of the structure of an embodiment of an automatic grading device in the embodiment of the present application;

[0085] Figure 8 It is a structural schematic diagram of another embodiment of the automatic grading device in the embodiments of the present application. DETAILED DESCRIPTION

[0086] The embodiments of the present application provide an automatic grading method, device, and computer-readable storage medium for improving the accuracy of grading results.

[0087] The method of the present application can be applied to a server, a terminal or other devices with logic processing capabilities, and the present application does not limit this. For the convenience of description, the following description is made by taking the execution subject as a server as an example.

[0088] The embodiments of the present application will be described below in conjunction with the accompanying drawings.

[0089] See also Figure 1 In the embodiment of the present application, an automatic classification method includes:

[0090] 101. Construct a policy and regulation ontology library and a typical case ontology library. The policy and regulation ontology library stores policy and regulation ontology with a graph data structure, and the typical case ontology library stores typical case ontology with a graph data structure.

[0091] The server's task is to create two ontology libraries: one for storing policies and regulations, and the other for storing typical cases. Each ontology library is based on a graph data structure. This structure is similar to a network, with nodes representing concepts, entities or terms, and edges representing the relationships between them. The policy and regulation ontology library is used to store information such as norms, regulations and articles related to laws and policies, and these information are connected to each other through association rules. The typical case ontology library stores representative cases in history, and the relationship between them reflects the complex association of information such as legal provisions, situations and judgments.

[0092] The server needs to call the graph database to store these graph data structures and ensure their accessibility and query efficiency.

[0093] By building a structured ontology library, the server can transform scattered and isolated policy and regulation data and typical case data into structured and associated graph data, which not only improves the visibility of data management, but also provides a basis for subsequent semantic reasoning and query.

[0094] 102. Input keywords in the policy and regulation ontology and the typical case ontology into the pre-trained word vector model to obtain multiple hierarchical keyword vectors;

[0095] The server traverses all nodes in the ontology library and extracts key concepts and terms (i.e. keywords) from policies, regulations and typical cases. These keywords may include important terms in legal provisions, core keywords of cases, etc. The server then inputs these keywords into a pre-trained word vector model, which is a natural language processing technology that converts each word into a vector representation in a high-dimensional space. These vector representations help subsequent steps to perform accurate query matching by capturing the semantic relationship between words.

[0096] In this way, the server can convert the vocabulary of text data into vector data that can be compared and processed mathematically.

[0097] By using the word embedding model, the server can better understand the semantic similarities between words, making subsequent query matching more accurate. This semantic processing greatly improves the intelligence of data processing, especially when involving a large number of legal terms and complex cases, it can effectively improve query efficiency.

[0098] 103. Obtain hierarchical query information, where the hierarchical query information is query data input by the user, and the hierarchical query information includes at least two input words;

[0099] The server receives the query request from the user, and the query data input by the user will contain at least two input words, such as "intellectual property" and "copyright dispute", etc. These input words represent the areas that the user is concerned about or the legal issues that he wants to query.

[0100] The server treats these input words as hierarchical query information and prepares to pass them to the next word vector model for vectorization. During this process, the server may also perform preliminary data cleaning and processing to ensure that the query words have no spelling errors or invalid characters.

[0101] By obtaining the user's query information, the server can capture the user's query intention and prepare for subsequent query matching. This step ensures that the user's needs can be smoothly connected with the structured data in the system.

[0102] 104. Input the hierarchical query information into a word vector model to obtain a corresponding hierarchical query vector, wherein each input word in the hierarchical query information corresponds to a hierarchical query vector;

[0103] The server inputs each query word entered by the user into the word vector model, and the model converts each query word into a corresponding vector representation. The vector of each word captures the semantic characteristics of the word in a high-dimensional space, allowing the server to compare the similarity of these query words with the keywords in the ontology library.

[0104] For example, the query term "intellectual property" may correspond to one vector, and the query term "copyright dispute" may correspond to another vector. Through this vectorization process, the server lays the foundation for subsequent similarity calculations.

[0105] By converting query terms into word vectors, the server can perform queries at the semantic level, making the match between user queries and data in the system more accurate. It no longer relies solely on simple text matching, but instead performs more intelligent matching based on semantic similarity.

[0106] 105. Determine a keyword set corresponding to the hierarchical query vector according to the hierarchical query vector and the hierarchical keyword vector, where the keyword set includes at least one keyword;

[0107] The server matches the user's hierarchical query vector with the previously generated hierarchical keyword vector. By calculating the similarity between them (e.g., through cosine similarity), the server is able to determine which keywords best match the user's query intent. Keywords with high similarity are selected into the keyword set.

[0108] For example, when a user searches for "copyright dispute", the server may find keywords related to "intellectual property law" and "copyright law" through similarity calculation and classify them into a keyword set.

[0109] The server intelligently selects the most relevant keywords to the user's query through similarity calculation, greatly improving the relevance and accuracy of the query results. This not only makes the query results more in line with user needs, but also helps users to discover other important information related to their query questions in more depth.

[0110] 106. Input the keyword set into the inference engine to perform forward rule reasoning query on the policy and regulation ontology library or the typical case ontology library, and obtain the graded result output by the inference engine.

[0111] The server inputs the generated keyword set into the inference engine, which will infer and query the policy and regulation ontology library or typical case ontology library based on the preset forward inference rules. The inference engine will automatically determine which policies, regulations or cases are most relevant to the words in the keyword set based on the rules, and infer step by step according to the rules to finally obtain a hierarchical query result.

[0112] For example, based on the keyword set "copyright dispute", the server may use inference rules to find relevant legal clauses, precedents, and legal interpretations, and display them in a graded manner based on relevance for user reference.

[0113] Through the forward rule reasoning of the inference engine, the server can automatically generate intelligent query results, allowing users to not only see directly related information, but also obtain inference results based on rule reasoning. This reasoning ability expands the scope and depth of user queries and can help users understand the query topic from multiple dimensions.

[0114] Furthermore, the grading results may include sharing level, opening level and grading basis. The sharing levels are divided into three levels: unconditional sharing, conditional sharing and non-sharing. The opening levels are divided into three levels: unconditional opening, opening upon application and non-opening. The server inputs the keyword set into the inference engine to perform forward rule reasoning query on the policy and regulation question bank. If a policy or regulation involving graded content is found, or multiple policies or regulations involving graded content but there is no conflict in terms between these policies or regulations, the corresponding graded content will be used as the grading basis, and the corresponding sharing level and opening level will be determined based on the graded content; if multiple policies or regulations involving graded content are found and there is a conflict in terms between these policies or regulations, the effectiveness will be judged, and the graded content in the policy or regulation with the highest effectiveness will be used as the grading basis, and the corresponding sharing level and opening level will be determined based on the graded content. If no policy or regulation involving graded content is found, the inference engine will perform forward rule reasoning query on the typical case ontology based on the keyword set. If a typical case including graded content is found, the corresponding graded content will be used as the basis for graded classification, and the corresponding sharing level and open level will be determined based on the graded content; if no typical case including graded content is found, a prompt indicating that no result is found will be returned. Graded content refers to content related to data sharing and data openness. The server can determine the corresponding sharing level and open level based on the content related to data sharing and data openness. Taking policy or regulation as an example, for example, a keyword including "personal information" in the keyword set is input into the inference engine for forward rule reasoning query, and it is possible to find laws and regulations including the information that "the relevant competent departments and their staff shall keep confidential the personal privacy, personal information, commercial secrets, confidential business information and other network data known in the performance of their duties in accordance with the law, and shall not disclose or illegally provide them to others". This information is the graded content of the law and regulation. This information involves content related to data sharing and data openness. According to the description of this information, it can be determined that the corresponding sharing level is not to be shared, and the corresponding open level is not to be open.

[0115] In this embodiment, the server can accurately match the user's hierarchical query with the keywords in the policy and regulation ontology library and the typical case ontology library, and generate relevant results using the forward reasoning mechanism. This process improves the accuracy and intelligence level of the query, ensures that the system can respond to user needs in real time, and provide in-depth legal interpretations and related cases based on the latest data, providing users with more accurate and valuable information.

[0116] See also Figure 2 In some embodiments of the present application, step 104 in the aforementioned embodiment inputs the hierarchical query information into the word vector model to obtain the corresponding hierarchical query vector, which may include the following steps 201 to 205:

[0117] 201. Input each input word in the hierarchical query information into the word vector model in sequence;

[0118] The server receives the query information input by the user, which contains multiple input words (such as "trade secrets" and "antitrust", etc.). The server processes these input words one by one, first inputting them into the pre-trained word vector model in turn. The word vector model maps each input word into a vector form, so that the semantics of these words can be processed and calculated through vector representation.

[0119] This process is similar to converting the original words into mathematical form, so that subsequent processing steps can be based on semantic similarity rather than simply the surface of the words.

[0120] By converting each input word into a word vector, the server can effectively process complex semantic relationships in the language and lay the foundation for subsequent multi-view learning and optimization.

[0121] 202. Use the multi-view learning sub-model of the word vector model to extract features from each input word to obtain sparse feature vectors under multiple views;

[0122] The server calls the multi-view learning sub-model in the word vector model to extract features from different perspectives for each word vector. Multi-view learning is a technology that analyzes and models data from different perspectives and can capture multiple semantic features of each input word.

[0123] For example, the word "trade secret" can be understood from multiple perspectives such as law, business, and security. Through multi-view learning, the server extracts multiple feature vectors for each input word and performs sparse processing to make the vector retain only the most important information from each perspective.

[0124] Multi-view learning can deeply mine the semantic information of input words from different angles, enabling the server to capture richer context and semantic details and improve the accuracy of subsequent query matching.

[0125] 203. Use the multi-view learning sub-model to fuse the sparse feature vectors to obtain the comprehensive word vector;

[0126] The server fuses the sparse feature vectors generated in the previous step through the multi-view learning sub-model. The fusion process combines the semantic features extracted from various perspectives to ensure that the multi-dimensional features of each input word are integrated to form a comprehensive word vector.

[0127] This comprehensive word vector contains more semantic information than the word vector from a single perspective and can better reflect the global meaning of the input word.

[0128] Through feature fusion, the server can more accurately represent the semantics of the input word, ensuring that subsequent quantum state superposition and query reasoning can be processed based on richer semantic information, thereby improving the accuracy of the results.

[0129] 204. Perform quantum state superposition on the comprehensive word vector to obtain a quantum state vector;

[0130] The server performs quantum state superposition processing on the comprehensive word vector. Quantum state superposition is a concept used in quantum computing, which refers to the superposition of multiple possibilities, representing the coexistence of multiple states. Here, the server maps the comprehensive semantic state of each word into the quantum state vector through superposition, enabling it to represent multiple semantic levels at the same time.

[0131] Through this quantum state expression, the potential meaning of the input word is expanded, which can not only retain the original semantics of the word but also capture a wider range of semantic associations.

[0132] Quantum state superposition enhances the expressive power of the input word vector, allowing the server to process more complex semantic associations during queries, further improving the intelligence and multi-dimensional expressive power of queries.

[0133] 205. The quantum state vector is optimized using the generative adversarial network of the word vector model to obtain a hierarchical query vector.

[0134] The server uses a generative adversarial network (GAN) to optimize the quantum state vector. GAN consists of a generator and a discriminator. The generator tries to generate realistic data (here, the optimized quantum state vector), while the discriminator determines whether the generated data meets the standard. Through this continuous game, the generative adversarial network can optimize the input quantum state vector to make it more accurate and useful.

[0135] After GAN optimization, the hierarchical query vector finally obtained by the server has higher representation ability and can accurately capture the semantic requirements of user queries.

[0136] The optimization of the generative adversarial network enables the server to generate higher-quality hierarchical query vectors, ensuring the accuracy and reliability of the query results. This step improves the performance of the query system in complex semantic scenarios, making the query results more in line with the actual needs of users.

[0137] In this embodiment, the server can generate high-precision hierarchical query vectors based on multi-view learning, quantum state superposition, and optimization of generative adversarial networks. This process not only improves the system's ability to understand complex semantics, but also makes the query results more accurate and intelligent, allowing users to quickly obtain the policies, regulations, and typical case information that best match their needs.

[0138] See also Figure 3 In some embodiments of the present application, the step 101 in the aforementioned embodiment of constructing a policy and regulation ontology library and a typical case ontology library may include the following steps 301 to 304:

[0139] 301. Use the ontology building tool to extract the ontology of the government information disclosure catalog to obtain the first policy and regulation ontology;

[0140] The server uses an ontology building tool to process the government information disclosure directory. The main function of this tool is to analyze the policy and regulatory information in the directory and convert it into a structured ontology form. The server parses the information disclosure directory, identifies the core terms, regulatory clauses, and the relationships between them, and then represents these elements as a graph data structure, namely the first policy and regulatory ontology.

[0141] The process of ontology extraction includes steps such as data cleaning, concept identification, and relationship establishment to ensure that useful semantic information is extracted from unstructured data.

[0142] By using ontology building tools, the server can automatically extract useful regulatory information from complex government information disclosure catalogs and structure it into an ontology to facilitate subsequent query and reasoning operations. This step improves the semantic and intelligent management capabilities of policy and regulatory information.

[0143] 302. storing the first policy and regulation ontology into the policy and regulation ontology library;

[0144] After extracting the first policy and regulation ontology, the server stores it in a pre-built policy and regulation ontology library. The ontology library is managed through a graph database or other ontology storage tools to ensure data persistence and accessibility. During storage, the server may also perform necessary updates or index construction on the ontology library to improve query performance.

[0145] The server will ensure the integrity and consistency of the ontology library so that subsequent reasoning and query operations can proceed smoothly.

[0146] After storing the extracted policy and regulatory ontology into the ontology library, the server can manage and query relevant regulatory information more efficiently, ensure that the query results are consistent with the latest policies and regulations, and improve the accuracy and timeliness of information retrieval.

[0147] 303. Use the ontology building tool to extract the ontology of typical cases on the government open data platform and obtain the first typical case ontology;

[0148] Similar to the process of processing the government information disclosure catalog, the server uses the ontology construction tool to process the typical case data on the government open data platform. The server analyzes these cases, identifies important concepts (such as case type, judgment results, legal clause citations, etc.) and their relationships, and converts this information into a typical case ontology.

[0149] This process can transform unstructured case data into structured semantic data and form the first typical case ontology.

[0150] By structuring the key information of typical cases into ontologies, the server can more intelligently understand the semantic associations between cases and improve the case association and relevance that users obtain when querying.

[0151] 304. Store the first typical case ontology into the typical case ontology library.

[0152] The server stores the extracted first typical case ontology into the typical case ontology library to ensure that the case data can be efficiently managed and queried by the system. Similar to the policy and regulation ontology library, the server needs to perform necessary updates and index optimization on the ontology library to support fast query and reasoning.

[0153] The storage structure of the ontology library enables efficient query of semantic associations between cases, and users can quickly locate relevant typical cases through queries.

[0154] By storing the extracted typical case ontology into the case ontology library, the server can better manage, query and reason about typical cases, improve the speed and accuracy of information retrieval, and help users quickly obtain typical cases related to policies and regulations.

[0155] In this embodiment, the server can automatically extract and construct the ontology of policies, regulations and typical cases, and store them in the corresponding ontology library. This intelligent processing method greatly improves the structuring and operability of data, allowing the system to provide more accurate and comprehensive results in subsequent query and reasoning processes, helping users to better understand and apply policies, regulations and case information.

[0156] See also Figure 4In some embodiments of the present application, the automatic classification method may further include the following steps 401 to 403:

[0157] 401. When the government information disclosure directory is updated, obtain the first update information;

[0158] The server monitors the update of the government information disclosure directory. When it detects changes in the directory (such as the addition or modification of policy and regulatory information), the server will automatically obtain the updated content. This update information is called "first update information", which usually includes newly released regulations, revised terms, and deleted outdated information.

[0159] The server collects these update data through automated crawling tools or API interfaces and performs preliminary processing on them to ensure the integrity and correctness of the update information.

[0160] By real-time monitoring and automatic acquisition of updated information, the server can ensure the timeliness of policy and regulatory data, keep the regulatory information in the system synchronized with the latest content released by the government, and improve the accuracy and timeliness of queries.

[0161] 402. Using an ontology building tool to extract the ontology of the first update information to obtain a second policy and regulation ontology;

[0162] The server uses the ontology building tool to process the first updated information and extract the updated policy and regulation ontology. This process is similar to the previous ontology extraction, which analyzes the new regulatory data, identifies key terms and relationships, and converts them into a structured ontology form, namely the second policy and regulation ontology.

[0163] This process ensures that new or modified regulatory information can be accurately mapped into the ontology structure and seamlessly connected with existing data.

[0164] By extracting the ontology of updated information, the server can intelligently process new regulations, so that the policy and regulatory ontology library in the system is always kept up to date and complete, ensuring that subsequent reasoning and queries can use the latest data.

[0165] 403. Store the second policy and regulation ontology into the policy and regulation ontology library.

[0166] After extracting the second policy and regulation ontology, the server will store it in the policy and regulation ontology library. At this time, the server not only stores the new ontology data, but also performs necessary merging, deduplication or version control on the existing ontology library to ensure that all data remains consistent and available.

[0167] The server may also update the index of the ontology library to improve query performance, and label or annotate new data to facilitate subsequent reasoning and query.

[0168] After the updated policy and regulation ontology is stored in the ontology library, the system can timely reflect the latest changes in laws and regulations, helping users obtain the latest and most accurate legal information. This process improves the efficiency of data maintenance and the accuracy of query results.

[0169] In this embodiment, the server can update policy and regulatory data in real time to ensure that the policy and regulatory database in the system is always consistent with the official release. This automated processing method effectively reduces the workload of manual maintenance, while improving the timeliness and accuracy of queries, and providing users with more intelligent regulatory information services.

[0170] See also Figure 5 In some embodiments of the present application, the automatic grading method may further include the following steps 501 to 503:

[0171] 501. When the government open data platform is updated, obtain the second update information;

[0172] The server monitors the updates of the government open data platform in real time. When new or modified typical case data is added to the platform, the server will automatically obtain these updated data. These data are called "second update information" and usually include new case details, judgment results or modified case descriptions.

[0173] The server collects this data through automatic crawling or API interface to ensure that all updates are reflected in the system in a timely manner.

[0174] By real-time monitoring and automatically acquiring updated data, the server can ensure that the typical case information database is always kept up to date, allowing users to obtain the latest case information and ensuring the timeliness and accuracy of query results.

[0175] 502. Using an ontology building tool to extract the second updated information, to obtain a second typical case ontology;

[0176] The server uses the ontology construction tool to analyze and process the acquired second update information, extract key information, and construct the second typical case ontology. This process includes semantic analysis of the case, identification of important concepts and relationships in the case, and conversion of this information into structured ontology data.

[0177] This step ensures that the updated case data can be seamlessly integrated into the existing typical case ontology system.

[0178] By performing ontology extraction on updated typical case data, the server can intelligently convert new data into structured ontology, making the case information in the system more complete and accurate, and facilitating subsequent query and reasoning.

[0179] 503. Store the second typical case ontology into the typical case ontology library.

[0180] After the server completes the extraction of the second typical case ontology, it will store it in the typical case ontology library. The server will perform necessary optimization and update operations based on the existing structure of the ontology library to ensure that the newly added case data is consistent with the existing data and improve the query efficiency of the system.

[0181] In addition, the server may optimize the index of the ontology library to ensure that users can quickly retrieve and obtain the latest case information.

[0182] By storing the updated case ontology in the ontology library, the server can ensure that the case data in the system is always kept up to date, and users can obtain the most comprehensive and accurate results when conducting case queries, further improving the quality of information services.

[0183] In this embodiment, the server can effectively process the updated data of the government open data platform, ensuring that the information in the typical case ontology library remains up-to-date and complete, and users can obtain the most relevant and latest case information when querying. This automated update and processing mechanism greatly improves the system maintenance efficiency and the accuracy of query results.

[0184] See also Figure 6 In some embodiments of the present application, the step 105 in the above embodiment determines the keyword set corresponding to the hierarchical query vector according to the hierarchical query vector and the hierarchical keyword vector, and may include the following steps 601 to 603:

[0185] 601. Calculate the similarity between each hierarchical query vector and hierarchical keyword vector respectively;

[0186] The server calculates the similarity between each hierarchical query vector and the hierarchical keyword vector. The similarity is usually calculated based on the distance metric of the vector (such as cosine similarity, Euclidean distance, etc.). The server compares each query vector entered by the user with all possible hierarchical keyword vectors one by one to measure the semantic similarity between them.

[0187] The purpose of this process is to find the keyword that is closest to the user's query semantics, thereby facilitating subsequent reasoning and query operations.

[0188] By accurately calculating the similarity, the server can match the most relevant keywords based on the query vector, ensuring that the system can intelligently understand the user's query intent and provide users with more accurate results.

[0189] 602. For any hierarchical query vector, use the keyword corresponding to the maximum similarity as the hierarchical keyword corresponding to the hierarchical query vector;

[0190] For each query vector, the server selects the keyword with the highest similarity as the best matching keyword for the query vector. This keyword represents the core semantics of the query vector.

[0191] The server will traverse all hierarchical query vectors and find the best keywords for each query vector. These keywords will serve as a bridge between the query vector and the system's internal policies and regulations or typical cases, helping the system better understand user input.

[0192] By selecting the most similar keywords, the server can find the best semantic match for each query vector, ensuring the accuracy and relevance of the query process and making the query results more targeted.

[0193] 603. Filter a preset number of constituent keyword sets from the hierarchical keywords.

[0194] The server selects a certain number of keywords from all matching hierarchical keywords according to the system preset requirements to form the final keyword set. The selection can be based on the degree of similarity or specific rules, such as selecting only the top N keywords.

[0195] This keyword set will serve as the basis for the final query or reasoning. The server will use these keywords to search and reason in the ontology library to obtain relevant policies, regulations or typical cases.

[0196] By filtering out the right number of keywords, the server can effectively narrow the query scope, making the query process more efficient while ensuring the relevance and accuracy of the output results. This step ensures that the final query results can actually meet the needs of users.

[0197] In this embodiment, the server can accurately match the user's hierarchical query with the keywords in the system and generate an optimized keyword set. This method not only improves the intelligence level of the query, but also ensures the accuracy of the output results, so that users can quickly find the policies, regulations or typical case information most relevant to their needs.

[0198] See also Figure 7 In the embodiment of the present application, an automatic grading device includes:

[0199] A construction unit 701 is used to construct a policy and regulation ontology library and a typical case ontology library, wherein the policy and regulation ontology library stores a policy and regulation ontology in a graph data structure, and the typical case ontology library stores a typical case ontology in a graph data structure;

[0200] The first input unit 702 is used to input keywords in the policy and regulation ontology and the typical case ontology into a pre-trained word vector model to obtain multiple hierarchical keyword vectors;

[0201] An acquisition unit 703 is used to acquire hierarchical query information, where the hierarchical query information is query data input by a user;

[0202] The second input unit 704 is used to input the hierarchical query information into the word vector model to obtain a corresponding hierarchical query vector, wherein each input word in the hierarchical query information corresponds to a hierarchical query vector;

[0203] A determining unit 705, configured to determine a keyword set corresponding to the hierarchical query vector according to the hierarchical query vector and the hierarchical keyword vector, wherein the keyword set includes at least one keyword;

[0204] The reasoning unit 706 is used to input the keyword set into the reasoning engine to perform forward rule reasoning query on the policy and regulation ontology library or the typical case ontology library to obtain the classification result output by the reasoning engine.

[0205] In this embodiment, the server can accurately match the user's hierarchical query with the keywords in the policy and regulation ontology library and the typical case ontology library, and generate relevant results using the forward reasoning mechanism. This process improves the accuracy and intelligence level of the query, ensures that the system can respond to user needs in real time, and provide in-depth legal interpretations and related cases based on the latest data, providing users with more accurate and valuable information.

[0206] Optionally, the grading result includes a sharing level, an open level, and a grading basis; the reasoning unit 706 is specifically used to:

[0207] Input the keyword set into the inference engine to conduct forward rule reasoning query on the policy and regulation ontology library;

[0208] If policies and regulations that include graded content and do not have conflicting clauses are found, the corresponding graded content will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; if policies and regulations that include graded content but have conflicting clauses are found, the graded content in the policies and regulations with the highest effectiveness will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; if policies and regulations with graded content are not found, the inference engine will be used to perform forward rule reasoning queries on the typical case ontology library based on the keyword set; if typical cases that include graded content are found, the corresponding graded content will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; among them, the graded content includes two aspects: data sharing and data openness.

[0209] Optionally, the second input unit 704 is specifically used for:

[0210] Input each input word in the hierarchical query information into the word vector model in turn;

[0211] The multi-view learning sub-model of the word vector model is used to extract features from each input word to obtain sparse feature vectors under multiple views.

[0212] Use the multi-view learning sub-model to fuse the sparse feature vectors to obtain the comprehensive word vector;

[0213] Perform quantum state superposition on the comprehensive word vector to obtain a quantum state vector;

[0214] The quantum state vector is optimized using the generative adversarial network of the word vector model to obtain a hierarchical query vector.

[0215] Optionally, the construction unit 701 is specifically used for:

[0216] Use ontology building tools to extract the ontology of government information disclosure catalogue and obtain the first policy and regulation ontology;

[0217] Storing the first policy and regulation ontology into the policy and regulation ontology library;

[0218] The ontology construction tool is used to extract the ontology of typical cases on the government open data platform to obtain the first typical case ontology;

[0219] Store the first typical case ontology into the typical case ontology library.

[0220] The device also includes a first updating unit, which is used to:

[0221] When the government information disclosure catalog is updated, obtain the first update information;

[0222] Using an ontology building tool to extract the ontology of the first update information to obtain a second policy and regulation ontology;

[0223] Store the second policy and regulation ontology into the policy and regulation ontology library.

[0224] The device also includes a second updating unit, which is used to:

[0225] When the government open data platform is updated, obtain the second update information;

[0226] Using an ontology building tool to extract the ontology of the second update information, and obtaining the second typical case ontology;

[0227] Store the second typical case ontology into the typical case ontology library.

[0228] Optionally, the determining unit 705 is specifically configured to:

[0229] Calculate the similarity between each hierarchical query vector and hierarchical keyword vector respectively;

[0230] For any hierarchical query vector, the keyword corresponding to the maximum similarity is used as the hierarchical keyword corresponding to the hierarchical query vector;

[0231] A preset number of constituent keyword sets are filtered from the hierarchical keywords.

[0232] In this implementation, the functions of each unit are the same as those mentioned above. Figures 1 to 6 The steps in the illustrated embodiments correspond to each other and will not be described again here.

[0233] See also Figure 8 Another embodiment of the automatic grading device in the embodiment of the present application includes:

[0234] Processor 801, memory 802, input and output unit 803 and bus 804;

[0235] The processor 801 is connected to the memory 802, the input and output unit 803 and the bus 804;

[0236] The memory 802 stores a program, and the processor 801 calls the program to execute Figures 1 to 6 Steps in the illustrated embodiment.

[0237] In this embodiment, the function of the processor 801 is the same as that of the aforementioned Figures 1 to 6 The steps in the illustrated embodiments correspond to each other and will not be described again here.

[0238] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0239] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as 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 mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0240] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0241] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0242] If 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 the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, read-only memory), random access memory (RAM, random access memory), disk or optical disk and other media that can store program code.

Claims

1. An automatic grading method, It is characterized in that include: Constructing a policy and regulation ontology library and a typical case ontology library, wherein the policy and regulation ontology library stores a policy and regulation ontology having a graph data structure, and the typical case ontology library stores a typical case ontology having a graph data structure; Input keywords in the policy and regulation ontology and the typical case ontology into a pre-trained word vector model to obtain multiple hierarchical keyword vectors; Acquire hierarchical query information, where the hierarchical query information is query data input by a user, and the hierarchical query information includes at least two input words; Inputting the hierarchical query information into the word vector model to obtain a corresponding hierarchical query vector, wherein each input word in the hierarchical query information corresponds to a hierarchical query vector; Determine a keyword set corresponding to the hierarchical query vector according to the hierarchical query vector and the hierarchical keyword vector, wherein the keyword set includes at least one keyword; Inputting the keyword set into the inference engine to perform forward rule reasoning query on the policy and regulation ontology library or the typical case ontology library, and obtaining the classification result output by the inference engine; The grading results include sharing level, open level and grading basis; The step of inputting the keyword set into the inference engine to perform forward rule reasoning query on the policy and regulation ontology library or the typical case ontology library, and obtaining the grading result output by the inference engine includes: Inputting the keyword set into the inference engine to perform forward rule reasoning query on the policy and regulation ontology library; If policies and regulations that include graded content and do not have conflicting clauses are found, the corresponding graded content will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; if policies and regulations that include graded content but have conflicting clauses are found, the graded content in the policies and regulations with the highest effectiveness will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; if policies and regulations that include graded content are not found, the inference engine will be used to perform forward rule reasoning query on the typical case ontology library based on the keyword set; if typical cases that include graded content are found, the corresponding graded content will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; wherein, the graded content includes two aspects: data sharing and data openness.

2. The method according to claim 1, It is characterized in that The step of inputting the hierarchical query information into the word vector model to obtain a corresponding hierarchical query vector comprises: Inputting each input word in the hierarchical query information into the word vector model in sequence; Using the multi-view learning sub-model of the word vector model to extract features from each input word, and obtain sparse feature vectors under multiple views; Using the multi-view learning sub-model to fuse the sparse feature vectors to obtain a comprehensive word vector; Perform quantum state superposition on the comprehensive word vector to obtain a quantum state vector; The quantum state vector is optimized using a generative adversarial network of the word vector model to obtain a hierarchical query vector.

3. The method according to claim 1, It is characterized in that The construction of the policy and regulation ontology library and the typical case ontology library includes: Use ontology building tools to extract the ontology of government information disclosure catalogue and obtain the first policy and regulation ontology; Storing the first policy and regulation ontology into a policy and regulation ontology library; Using the ontology building tool to extract the ontology of typical cases on the government open data platform, and obtaining a first typical case ontology; The first typical case ontology is stored in a typical case ontology library.

4. The method according to claim 3, It is characterized in that The method further comprises: When the government information disclosure catalog is updated, obtaining first update information; Using the ontology building tool to extract the ontology of the first update information to obtain a second policy and regulation ontology; The second policy and regulation ontology is stored in the policy and regulation ontology library.

5. The method according to claim 3, It is characterized in that The method further comprises: When the government open data platform is updated, obtaining second update information; Using the ontology building tool to extract the ontology of the second update information to obtain a second typical case ontology; The second typical case ontology is stored in the typical case ontology library.

6. The method according to claim 1, It is characterized in that The step of determining a keyword set corresponding to the hierarchical query vector according to the hierarchical query vector and the hierarchical keyword vector comprises: respectively calculating the similarity between each of the hierarchical query vectors and the hierarchical keyword vectors; For any of the hierarchical query vectors, taking the keyword corresponding to the maximum similarity as the hierarchical keyword corresponding to the hierarchical query vector; A preset number of constituent keyword sets are screened from the hierarchical keywords.

7. An automatic grading device, It is characterized in that include: A construction unit, used to construct a policy and regulation ontology library and a typical case ontology library, wherein the policy and regulation ontology library stores a policy and regulation ontology having a graph data structure, and the typical case ontology library stores a typical case ontology having a graph data structure; A first input unit is used to input keywords in the policy and regulation ontology and the typical case ontology into a pre-trained word vector model to obtain multiple hierarchical keyword vectors; An acquisition unit, used to acquire hierarchical query information, where the hierarchical query information is query data input by a user; A second input unit, configured to input the hierarchical query information into the word vector model to obtain a corresponding hierarchical query vector, wherein each input word in the hierarchical query information corresponds to a hierarchical query vector; a determining unit, configured to determine a keyword set corresponding to the hierarchical query vector according to the hierarchical query vector and the hierarchical keyword vector, wherein the keyword set includes at least one keyword; An inference unit, used for inputting the keyword set into an inference engine to perform forward rule inference query on the policy and regulation ontology library or the typical case ontology library, and obtaining a grading result output by the inference engine; The grading result includes a sharing level, an open level and a grading basis; the reasoning unit is specifically used for: Inputting the keyword set into the inference engine to perform forward rule reasoning query on the policy and regulation ontology library; If policies and regulations that include graded content and do not have conflicting clauses are found, the corresponding graded content will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; if policies and regulations that include graded content but have conflicting clauses are found, the graded content in the policies and regulations with the highest effectiveness will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; if policies and regulations that include graded content are not found, the inference engine will be used to perform forward rule reasoning query on the typical case ontology library based on the keyword set; if typical cases that include graded content are found, the corresponding graded content will be used as the basis for classification, and the corresponding sharing level and open level will be determined based on the graded basis; wherein, the graded content includes two aspects: data sharing and data openness.

8. An automatic grading device, It is characterized in that include: Processor, memory, input-output unit, and bus; The processor is connected to the memory, the input and output unit, and the bus; A program is stored in the memory, and the processor calls the program to execute the method according to any one of claims 1 to 6. 9 . A computer-readable storage medium having a program stored thereon, wherein when the program is executed on a computer, the computer is caused to execute the method according to claim 1 .

Images