An intention understanding agent construction method based on organization unique knowledge governance

Abstract

The application relates to an intention understanding intelligent agent construction method and device based on organization unique knowledge governance, a computer device, a storage medium and a program product. The method comprises the following steps: acquiring a business information query statement of an account input intelligent agent associated with a target organization, and extracting a target semantic element in the business information query statement; matching the target semantic element with data entities in a pre-constructed organization knowledge graph to obtain an entity matching result; generating a data query instruction in a preset structured format according to the business information query statement, the target semantic element and the entity matching result; and generating a business information query result of the business information query statement according to the data query instruction. The method can improve the accuracy and reliability of the business information query of the target organization.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to methods and apparatuses for constructing intention-understanding intelligent agents based on the governance of organization-specific knowledge, computer equipment, storage media, and program products. Background Technology

[0002] As organizations such as businesses and institutions grow in size, their demand for internal business information queries also increases. Due to differences in organizational structure and business types, the same words or phrases may even correspond to completely different concepts in different organizations.

[0003] In related technologies, large language models are used to understand the semantics of user-input queries, and then SQL (Structured Query Language) statements are generated for business information retrieval by matching semantics with rule templates. However, generalized large language models struggle to accurately identify the query intent of users, resulting in low accuracy and reliability of query results. Summary of the Invention

[0004] Therefore, it is necessary to provide a method and device, computer equipment, storage medium, and program product for constructing an intent-understanding intelligent agent based on the organization's unique knowledge governance, which can improve the accuracy and reliability of business information query within the organization and enhance the accuracy and reliability of such queries.

[0005] Firstly, this application provides a method for constructing an intent-understanding intelligent agent based on organization-specific knowledge governance, including:

[0006] Obtain the business information query statement of the intelligent agent inputting the account associated with the target organization, and extract the target semantic elements in the business information query statement; the target semantic elements are the semantic elements in the business information query statement that match the target organization.

[0007] The target semantic element is matched with data entities in a pre-constructed organizational knowledge graph to obtain entity matching results; the organizational knowledge graph is used to store organizational business knowledge data associated with the target organization; the entity matching results include target data entities that match the target semantic element; the target data entities are used to represent data query elements that match the business information query statement;

[0008] Based on the business information query statement, the target semantic element, and the entity matching result, a data query instruction using a preset structured format is generated;

[0009] Based on the data query instruction, generate the business information query result of the business information query statement.

[0010] In one embodiment, the data query elements include a target table name, a target field name, and a target data value. The step of generating a data query instruction using a preset structured format based on the business information query statement, the target semantic element, and the entity matching result includes:

[0011] Based on the business information query statement, construct user intent prompt words;

[0012] Based on the target table name, the target field name, and the target data value, construct query condition suggestions;

[0013] Based on the target semantic elements, construct semantic prompt words;

[0014] The user intent prompt, the query condition prompt, and the semantic prompt are input into a pre-trained generative model, which then generates a data query instruction using a preset structured format.

[0015] In one embodiment, the target semantic element has multiple components, and the step of matching the target semantic element with data entities in a pre-constructed organizational knowledge graph to obtain entity matching results includes:

[0016] Determine the semantic similarity between each target semantic element and multiple data entities in the organizational knowledge graph pairwise, and sort the multiple data entities according to the semantic similarity to obtain the similarity ranking result of the multiple data entities;

[0017] Based on the similarity ranking results of the multiple data entities, the target data entity that matches the target semantic element is determined;

[0018] Output the entity matching result based on the target data entity that matches the target semantic element.

[0019] In one embodiment, after the step of determining the target data entity matching the target semantic element based on the similarity ranking results of the plurality of data entities, the method further includes:

[0020] Obtain the target association relationships between target data entities that match different target semantic elements;

[0021] When the target association relationship between each target data entity is a preset association relationship, the step of outputting the entity matching result based on the target data entity that matches the target semantic element is executed.

[0022] In one embodiment, the method further includes:

[0023] Obtain the target data table that stores the associated data of the target organization, and determine the table name, field names, and data values ​​of the target data table;

[0024] Construct multiple data entities of different types based on the table name, the field name, and the data value, respectively.

[0025] Obtain the governance relationship information of the target organization, and determine the association relationship between the multiple data entities according to the governance relationship information; wherein, the governance relationship information includes one or more of the following: sub-organization affiliation information, organizational personnel affiliation information, and organizational personnel association information of the target organization;

[0026] The organizational knowledge graph is constructed based on the relationships between multiple data entities.

[0027] In one embodiment, extracting the target semantic elements from the business information query statement includes:

[0028] Determine the target language of the business information query statement;

[0029] According to the preset semantic syntax rules corresponding to the target language, each semantic element in the business information query statement is sequence-labeled to obtain the semantic syntax tag corresponding to each semantic element;

[0030] From the business information query statement, extract the semantic elements that match the semantic syntax tags of the target organization to obtain the target semantic elements.

[0031] Secondly, this application also provides an intent-understanding intelligent agent construction device based on organization-specific knowledge governance, comprising:

[0032] The target semantic element extraction module is used to obtain the business information query statement of the account input agent associated with the target organization, and extract the target semantic elements in the business information query statement; the target semantic elements are the semantic elements in the business information query statement that match the target organization.

[0033] The entity matching module is used to match the target semantic element with data entities in a pre-built organizational knowledge graph to obtain entity matching results; the organizational knowledge graph is used to store organizational business knowledge data associated with the target organization; the entity matching results include target data entities that match the target semantic element; the target data entities are used to represent data query elements that match the business information query statement;

[0034] The data query instruction generation module is used to generate a data query instruction in a preset structured format based on the business information query statement, the target semantic element, and the entity matching result.

[0035] The business information query result generation module is used to generate business information query results for the business information query statement based on the data query instruction.

[0036] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0037] Obtain the business information query statement of the intelligent agent inputting the account associated with the target organization, and extract the target semantic elements in the business information query statement; the target semantic elements are the semantic elements in the business information query statement that match the target organization.

[0038] The target semantic element is matched with data entities in a pre-constructed organizational knowledge graph to obtain entity matching results; the organizational knowledge graph is used to store organizational business knowledge data associated with the target organization; the entity matching results include target data entities that match the target semantic element; the target data entities are used to represent data query elements that match the business information query statement;

[0039] Based on the business information query statement, the target semantic element, and the entity matching result, a data query instruction using a preset structured format is generated;

[0040] Based on the data query instruction, generate the business information query result of the business information query statement.

[0041] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0042] Obtain the business information query statement of the intelligent agent inputting the account associated with the target organization, and extract the target semantic elements in the business information query statement; the target semantic elements are the semantic elements in the business information query statement that match the target organization.

[0043] The target semantic element is matched with data entities in a pre-constructed organizational knowledge graph to obtain entity matching results; the organizational knowledge graph is used to store organizational business knowledge data associated with the target organization; the entity matching results include target data entities that match the target semantic element; the target data entities are used to represent data query elements that match the business information query statement;

[0044] Based on the business information query statement, the target semantic element, and the entity matching result, a data query instruction using a preset structured format is generated;

[0045] Based on the data query instruction, generate the business information query result of the business information query statement.

[0046] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0047] Obtain the business information query statement of the intelligent agent inputting the account associated with the target organization, and extract the target semantic elements in the business information query statement; the target semantic elements are the semantic elements in the business information query statement that match the target organization.

[0048] The target semantic element is matched with data entities in a pre-constructed organizational knowledge graph to obtain entity matching results; the organizational knowledge graph is used to store organizational business knowledge data associated with the target organization; the entity matching results include target data entities that match the target semantic element; the target data entities are used to represent data query elements that match the business information query statement;

[0049] Based on the business information query statement, the target semantic element, and the entity matching result, a data query instruction using a preset structured format is generated;

[0050] Based on the data query instruction, generate the business information query result of the business information query statement.

[0051] The aforementioned business information query method, apparatus, computer equipment, storage medium, and program products based on organizational semantics and organizational knowledge extract target semantic elements matching the target organization from the business information query statement input to the intelligent agent. This accurately extracts key information associated with the target organization from the business information query statement, and then matches these target semantic elements with data entities in a pre-built organizational knowledge graph to obtain entity matching results. These entity matching results include target data entities representing data query elements matching the business information query statement, achieving precise location and enhanced retrieval of data query elements. Furthermore, since the target data entities originate from the pre-built organizational knowledge graph, it ensures that the data query instruction is generated only when the organizational business knowledge data stored in the organizational knowledge graph satisfies the query intent. This achieves a precise understanding of the organizational business information query intent, effectively solving the problem of low accuracy caused by using generalized large models for semantic recognition, which struggles to adapt to the organization's unique knowledge. Thus, an intent-understanding intelligent agent based on organizational-unique knowledge governance is constructed, effectively improving the accuracy and reliability of business information query result generation. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 A flowchart illustrating the steps of an intent-understanding agent construction method based on organization-specific knowledge governance, provided in an embodiment of this application;

[0054] Figure 2 A structural block diagram of an intent-understanding intelligent agent construction device based on organization-specific knowledge governance, provided in an embodiment of this application;

[0055] Figure 3 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0056] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0057] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0058] With the development of AI (Artificial Intelligence) technology, business information retrieval not only needs to meet traditional business scenarios, but also needs to be empowered by AI application scenarios represented by RAG (Retrieval-Augmented Generation), further increasing the requirements for search accuracy and reliability.

[0059] Enterprise knowledge search is one type of business information query scenario. The knowledge search needs of enterprise users are significantly different from those of C (Client) end users. As a result, the methods of using generalized large models for text slicing semantic vector retrieval and using text-to-SQL (text-to-SQL statements) generated based on rule template matching for knowledge search cannot meet the needs of enterprises for accurate information retrieval and question answering.

[0060] On the one hand, enterprise users often have a clear purpose when retrieving enterprise data and knowledge, and they possess prior knowledge of the business. This prior knowledge (such as specific business terms and concepts within the enterprise) is only implicitly expressed in the user's query, and cannot be understood by generalized large models. Furthermore, because semantic vector retrieval is based on vector similarity matching, the judgment of the relevance between enterprise knowledge and user intent is ambiguous. In real-world scenarios, it is difficult to determine the threshold for relevance judgment, leading to the erroneous retrieval of content unrelated to the user's actual intent, resulting in inaccurate searches and question-and-answer illusions, leading to low search result reliability. Therefore, compared to C-end users, enterprise users prioritize accuracy over recall.

[0061] On the other hand, rule-template-based Text-to-SQL solutions allow developers to control the matching process through rules, but correspondingly, they also require manual collection and writing of rule templates. Furthermore, the semantic diversity of natural language cannot be exhaustively enumerated by rules; therefore, rule-template-based SQL generation methods can only handle language input with specific patterns, limiting their value and scope. This results in poor scalability and an inability to meet the diverse needs of enterprise users.

[0062] The inventors of this application have discovered that the core problems mentioned above have not yet been fully solved by the technical solutions of related technologies. Natural language possesses inherent flexibility and ambiguity. The conversion from natural language to data query commands inevitably involves many-to-many matching problems (a natural language query can be interpreted as different commands, and vice versa). Furthermore, it requires not only a precise understanding of natural language but also a precise understanding of organizational business knowledge. Converting a natural language query into a precise data query targeting internal enterprise knowledge depends not only on the user's input but also on the structure and characteristics of the target knowledge base.

[0063] Therefore, this application aims to solve the technical problem of how to accurately understand the business information query intent expressed by enterprise users in natural language and transform it into a highly accurate and reliable data query instruction. Based on the technical concept of extracting target semantic elements that match the target organization in the business information query statement and matching the target semantic elements with target data entities in a pre-built organizational knowledge graph, this application achieves high-precision, high-reliability, and highly scalable business information query for the target organization.

[0064] The embodiments of this application will be described in detail below.

[0065] In one embodiment, such as Figure 1 As shown, a method for querying business information is provided. This embodiment illustrates the application of this method to a terminal. It is understood that this method can also be applied to a server, and further to a system including both a terminal and a server, and is implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps:

[0066] Step 101: Obtain the business information query statement of the account input agent associated with the target organization, and extract the target semantic elements in the business information query statement; the target semantic elements are the semantic elements in the business information query statement that match the target organization.

[0067] In practical implementation, the accounts associated with the target organization can be internal accounts of the target organization, such as employee accounts, or accounts authorized by the target organization. Business information query statements can be directly entered text or text obtained through speech-to-text conversion. The intelligent agent can be pre-built for the target organization's business query needs, such as an intelligent assistant for enterprise knowledge management. After receiving business information query statements from accounts associated with the target organization, the intelligent agent can autonomously process the entire business information query process and also possess the ability to autonomously manage the unique knowledge within the target organization. Target semantic elements can specifically be semantic elements such as keywords and key phrases that match the target organization.

[0068] In some examples, target semantic elements in business information query statements can be extracted by using predefined keyword rules and grammar templates for business scenarios; alternatively, business information query statements can be input into a language parsing model for word segmentation and classification, and target semantic elements that match the target organization can be filtered out from the classification results.

[0069] In some embodiments, extracting the target semantic elements from the business information query statement includes:

[0070] Determine the target language of the business information query statement;

[0071] According to the preset semantic syntax rules corresponding to the target language, each semantic element in the business information query statement is sequence-labeled to obtain the semantic syntax tag corresponding to each semantic element;

[0072] From the business information query statement, extract the semantic elements that match the semantic syntax tags of the target organization to obtain the target semantic elements.

[0073] In practical implementation, the target language can be Chinese, English, or a mixture of both. Following the preset semantic grammar rules corresponding to the target language, each semantic element in the business information query statement is sequentially labeled to obtain a semantic grammar tag for each semantic element. The semantic grammar tag can include the semantic element's grammatical attributes and semantic attributes. For example, for the semantic element "Company A," the semantic grammar tag can indicate that this semantic element is an object in Chinese grammar and a specific company name in Chinese semantics. Furthermore, semantic elements that match the semantic grammar tag and the target organization (such as keywords and key phrases associated with the target organization) are extracted, while other semantic elements (such as modal particles unrelated to the target organization) are filtered out.

[0074] In practical applications, although Chinese offers greater flexibility in both semantics and grammar, it's unnecessary to parse all semantic elements of a Chinese sentence; only elements relevant to the enterprise user's search intent need to be extracted. In enterprise business scenarios, user search statements exhibit regularity, with key search intent often concentrated in specific grammatical elements. Therefore, to compensate for the high ambiguity in Chinese grammatical parsing and to selectively extract target semantic elements (such as noun structures representing entities or phrases indicating transaction states), a Chinese search statement parsing model can be employed. This model performs sequence labeling on semantic elements in business information queries, obtaining more precise semantic grammatical tags, and then extracting target semantic elements that match the target organization within the business information query. This Chinese search statement parsing model can be deployed within an intelligent agent, enabling the agent to extract target semantic elements. The Chinese search statement parsing model can be developed based on the Transformer architecture.

[0075] Step 102: Match the target semantic element with data entities in the pre-constructed organizational knowledge graph to obtain entity matching results; the organizational knowledge graph is used to store organizational business knowledge data associated with the target organization; the entity matching results include target data entities that match the target semantic element; the target data entities are used to represent data query elements that match the business information query statement;

[0076] Organizational business knowledge data refers to relevant knowledge data that is related to the business of the target organization. For example, for a company, organizational business knowledge data may include knowledge data such as departmental hierarchy and departmental staffing.

[0077] In practical implementation, data governance can be performed on the target organization's internal data beforehand. This involves first organizing the internal data into organizational business knowledge data, then constructing an organizational knowledge graph. Different types of organizational business knowledge data are stored through data entities within the organizational knowledge graph, and the relationships between these data entities represent the relationships between different types of organizational business knowledge data. By matching the target semantic elements with the data entities in the pre-constructed organizational knowledge graph, entity matching results are obtained, including the target data entities. These target data entities can represent data query elements that match the business information query statement. Specifically, data query elements can include the name of the data table to be queried, field names, data values, query conditions, etc.

[0078] In some embodiments, the method further includes:

[0079] Obtain the target data table that stores the associated data of the target organization, and determine the table name, field names, and data values ​​of the target data table;

[0080] Construct multiple data entities of different types based on the table name, the field name, and the data value, respectively.

[0081] Obtain the governance relationship information of the target organization, and determine the association relationship between the multiple data entities according to the governance relationship information; wherein, the governance relationship information includes one or more of the following: sub-organization affiliation information, organizational personnel affiliation information, and organizational personnel association information of the target organization;

[0082] The organizational knowledge graph is constructed based on the relationships between multiple data entities.

[0083] In practical implementation, the associated data of the target organization can include its various business data, attribute data, and related data of its sub-organizations (such as branch offices). By determining the table name, field names, and data values ​​of the target data table, multiple data entities of different types are constructed based on these table names, field names, and data values. These data entities can be distinguished by type attributes (i.e., distinguishing the types of table names, field names, and data values). Each data entity can also include attributes such as aliases and descriptions, supporting more flexible and comprehensive data entity matching.

[0084] Governance relationship information is used to characterize information related to governance relationships such as decision-making, reporting, and collaboration among various units within the target organization. Governance relationship information includes one or more of the following: sub-organization affiliation information, personnel affiliation information, and personnel association information. Specifically, sub-organization affiliation information characterizes the affiliation of subordinate or affiliated sub-organizations within the target organization; for example, it may include the affiliation information between business regions and specific cities within the enterprise, or the organizational structure of enterprise departments. Personnel affiliation information characterizes the affiliation information of personnel associated with the target organization, such as the departments to which employees belong. Personnel association information characterizes the relationships between various personnel within the target organization, such as departmental or job-level relationships among employees.

[0085] Based on this governance relationship information, the relationships between multiple data entities can be constructed. That is, the edges between each data entity in the organizational knowledge graph are constructed, representing the relationship from data value to field name (i.e., column name of data table) and data table. Based on the relationships between multiple data entities, the organizational knowledge graph is constructed, forming the knowledge base of the target organization.

[0086] In practical applications, intelligent agents can autonomously perform data governance on internal organizational data, thereby constructing an organizational knowledge graph, and can also regularly update the constructed organizational knowledge graph.

[0087] In this embodiment, the association between multiple data entities is determined by governance relationship information, and an organizational knowledge graph is constructed based on the association between multiple data entities. This forms an organizational governance relationship network from data values ​​to field names and data tables, so as to accurately match the target data entity corresponding to the target semantic element. Furthermore, the reliability of the matching result can be verified by the association between multiple data entities, thereby improving the reliability and accuracy of business information query results.

[0088] In some embodiments, the target semantic element has multiple components, and the step of matching the target semantic element with data entities in a pre-constructed organizational knowledge graph to obtain entity matching results includes:

[0089] Determine the semantic similarity between each target semantic element and multiple data entities in the organizational knowledge graph pairwise, and sort the multiple data entities according to the semantic similarity to obtain the similarity ranking result of the multiple data entities;

[0090] Based on the similarity ranking results of the multiple data entities, the target data entity that matches the target semantic element is determined;

[0091] Output the entity matching result based on the target data entity that matches the target semantic element.

[0092] In the specific implementation, the semantic similarity between each target semantic element and multiple data entities in the organizational knowledge graph is used to reflect the degree of matching between each target semantic element and each data entity. Then, the multiple data entities are sorted according to the semantic similarity to obtain the similarity ranking result of the multiple data entities. The data entity with the highest semantic similarity is determined in the ranking result to ensure that the target semantic element and the data entity are matched one-to-one.

[0093] In practical applications, the Embedding semantic vector retrieval method can be used to convert each target semantic element into a feature vector. The semantic similarity between the feature vector and each data entity is calculated to achieve fuzzy matching between the target semantic element and the data entity. Then, a conditional threshold is set to filter out data entities with semantic similarity lower than the conditional threshold. From the filtered data entities, the unique matching target data entity with the highest semantic similarity is determined to ensure the accuracy of the matching.

[0094] In some embodiments, after the step of determining the target data entity matching the target semantic element based on the similarity ranking results of the plurality of data entities, the method further includes:

[0095] Obtain the target association relationships between target data entities that match different target semantic elements;

[0096] When the target association relationship between each target data entity is a preset association relationship, the step of outputting the entity matching result based on the target data entity that matches the target semantic element is executed.

[0097] In practical implementation, multiple different target semantic elements may be extracted from the business query statement, and target data entities corresponding to the target semantic elements are matched. By obtaining the target association relationships between target data entities, the unique knowledge of the target organization implicit in the business query statement can be mined to verify whether there are association relationships between target data entities that match the query intent and whether the organizational business knowledge data in the organizational knowledge graph can satisfy the true query intent in the business query statement, thereby verifying the reliability of the matched target data entities. If a target association relationship can be obtained, or the target association relationship is a preset association relationship, it can be determined that the query intent can be satisfied by the organizational business knowledge data in the organizational knowledge graph, and the matched target data entities are reliable, and then the subsequent steps of outputting entity matching results are executed; if a target association relationship cannot be obtained, or the target association relationship is not a preset association relationship, it can be determined that the query intent cannot be satisfied by the organizational business knowledge data in the organizational knowledge graph, and the matched target data entities are unreliable, then the query process of business query information is terminated, and the query failure result is returned.

[0098] In practical applications, intelligent agents can uncover the unique knowledge of the target organization hidden in business query statements by verifying the target associations and preset associations between target data entities. This can compensate for the adaptability of social big data models to the unique knowledge of the target organization when using social big data models to generate data query instructions.

[0099] In this embodiment, the matching reliability of target data entities is verified by the target association relationship between each target data entity. When the target association relationship is a preset association relationship, the subsequent steps of outputting entity matching results are executed. This can effectively solve the problem of "retrieval illusion", improve the accuracy of target data entity matching, improve the accuracy and reliability of data query command generation, and achieve retrieval enhancement.

[0100] Step 103: Generate a data query instruction using a preset structured format based on the business information query statement, the target semantic element, and the entity matching result;

[0101] In a practical implementation, the business information query statement, target semantic elements, and entity matching results can be constructed into structured information. This structured information is then input into a large model, which generates data query instructions using a preset structured format. For example, it can generate SQL statements corresponding to different databases.

[0102] In practical applications, large models can specifically refer to general-purpose large models, such as socialized large language models. Because the accurately extracted target semantic elements and the precise matching entity matching results provide clear guidance for generating data query instructions for general-purpose large models, they can compensate for the poor adaptability of general-purpose large models to the organizational business knowledge of the target organization. This approach utilizes the versatility and generalization ability of large models while also being able to handle arbitrary natural language inputs, thus resolving the contradiction between flexibility and accuracy.

[0103] In some embodiments, the data query elements include the target table name, the target field name, and the target data value. Generating a data query instruction using a preset structured format based on the business information query statement, the target semantic element, and the entity matching result includes:

[0104] Based on the business information query statement, construct user intent prompt words;

[0105] Based on the target table name, the target field name, and the target data value, construct query condition suggestions;

[0106] Based on the target semantic elements, construct semantic prompt words;

[0107] The user intent prompt, the query condition prompt, and the semantic prompt are input into a pre-trained generative model, which then generates a data query instruction using a preset structured format.

[0108] In practical implementation, user intent prompts refer to prompts that represent the user's original intent. The entire original text of the business information query statement can be constructed into prompts that adapt to the input format of the generative model. Through user intent prompts, it can be ensured that the generative model always takes solving the initial problem as the ultimate goal and prevents it from deviating from the user's intent in the process of generating data query instructions.

[0109] Semantic cue words refer to semantic metadata that generative models can understand. Specifically, they can be constructed using target semantic elements based on the model structure of the generative model. Precise guidance from semantic cue words can avoid ambiguity in the generative model's understanding, thereby improving the accuracy of data query commands.

[0110] Query condition prompts are prompts that guide the generation of various query conditions in a data query instruction. They can be constructed from the target table name, target field name, and target data value to fit the input format of the generative model. During the generation of the data query instruction, the generative model can verify whether the query conditions meet the constraints of the query condition prompts (such as verifying whether the query conditions match the target table name, target field name, and target data value), further improving the accuracy of the generated data query instructions.

[0111] In this embodiment, although the large model is trained on massive amounts of data with a large number of parameters, it can handle any natural language input. However, the core issue is the conversion from fuzzy to precise. After the large model learns the input natural language, how to convert it into precise instructions remains fuzzy and uncertain. The large model's processing involves converting text into high-dimensional vector spaces that lack interpretability, and then calculating the output result through vector probabilities. This process is an uncontrollable black box for users, often resulting in unstable behavior in actual use. Furthermore, developers find it difficult to analyze and trace the root causes of problems, causing the generalized large model to fail to meet the maintainability and controllability requirements of enterprise users. Meanwhile, in actual business systems, this matching process needs to be controllable and scalable, allowing system developers to make specific adjustments to the matching process based on business needs and to trace the root causes of errors.

[0112] Based on this, this embodiment inputs user intent prompts, query condition prompts, and semantic prompts into a pre-trained generative model. The generative model then generates data query instructions in a preset structured format. Since the user intent prompts, query condition prompts, and semantic prompts are already precise prior knowledge, the generative model only needs to be responsible for constructing the data query instructions (e.g., generating different SQL statements to adapt to different databases). Therefore, when the query results are incorrect, developers can more easily attribute and trace the cause of the problem (e.g., tracing the construction of user intent prompts, query condition prompts, and semantic prompts, and each previous step), without needing to parse the complex structure of the generative model. This meets the scalability and maintainability requirements of organizational users such as enterprise users, while providing accurate business information query results.

[0113] Step 104: Generate the business information query result of the business information query statement according to the data query instruction.

[0114] In practical implementation, the intelligent agent can query target data in the database through data query commands and directly output the target data to obtain business information query results. Alternatively, it can organize the target data into business information query results in natural language. For example, in the context of internal enterprise knowledge question answering, the intelligent agent can act as an intelligent assistant for enterprise knowledge management. The intelligent assistant can be deployed with a large language model. The intelligent assistant obtains relevant target data through data query commands and then inputs the data query commands into the large language model to generate answer text in the language of business queries.

[0115] The embodiments of this application have the following advantages: By extracting target semantic elements matching the target organization from the business information query statement input to the intelligent agent, the key information associated with the target organization in the business information query statement is accurately extracted. Then, based on the target semantic elements, data entities in a pre-built organizational knowledge graph are matched to obtain entity matching results. The entity matching results include target data entities used to represent data query elements matching the business information query statement, achieving precise positioning and enhanced retrieval of data query elements. Simultaneously, since the target data entities originate from the pre-built organizational knowledge graph, it can be ensured that the data query instruction is generated when the organizational business knowledge data stored in the organizational knowledge graph can satisfy the query intent. This achieves a precise understanding of the organizational business information query intent and effectively solves the problem of low accuracy caused by using generalized large models for semantic recognition, which is difficult to adapt to the unique knowledge of the organization. Thus, an intent-understanding intelligent agent based on the governance of unique organizational knowledge is constructed, which effectively improves the accuracy and reliability of business information query result generation.

[0116] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0117] Based on the same inventive concept, this application also provides an apparatus for constructing an intent-understanding intelligent agent based on organization-specific knowledge governance, as described above. The solution provided by this apparatus is similar to the solution described in the above method; therefore, specific limitations in one or more apparatus embodiments provided below can be found in the limitations of the intent-understanding intelligent agent construction method based on organization-specific knowledge governance described above, and will not be repeated here.

[0118] In one exemplary embodiment, such as Figure 2 As shown, an intent-understanding intelligent agent construction device based on organization-specific knowledge governance is provided, comprising:

[0119] The target semantic element extraction module 201 is used to obtain the business information query statement of the account input agent associated with the target organization, and extract the target semantic elements in the business information query statement; the target semantic elements are the semantic elements in the business information query statement that match the target organization.

[0120] The entity matching module 202 is used to match the target semantic element with data entities in a pre-built organizational knowledge graph to obtain entity matching results; the organizational knowledge graph is used to store organizational business knowledge data associated with the target organization; the entity matching results include target data entities that match the target semantic element; the target data entities are used to represent data query elements that match the business information query statement;

[0121] The data query instruction generation module 203 is used to generate a data query instruction in a preset structured format based on the business information query statement, the target semantic element and the entity matching result;

[0122] The business information query result generation module 204 is used to generate the business information query result of the business information query statement according to the data query instruction.

[0123] In some embodiments, the data query elements include the target table name, the target field name, and the target data value, and the data query instruction generation module 203 includes:

[0124] The user intent prompt word construction submodule is used to construct user intent prompt words based on the business information query statement;

[0125] The query condition suggestion construction submodule is used to construct query condition suggestion words based on the target table name, the target field name, and the target data value;

[0126] The semantic prompt word construction submodule is used to construct semantic prompt words based on the target semantic element;

[0127] The data query instruction generation submodule is used to input the user intent prompt, the query condition prompt, and the semantic prompt into a pre-trained generative model, and generate a data query instruction using a preset structured format through the generative model.

[0128] In some embodiments, the target semantic elements are multiple, and the entity matching module 202 includes:

[0129] The similarity ranking submodule is used to determine the semantic similarity between each target semantic element and multiple data entities in the organizational knowledge graph, and to rank the multiple data entities according to the semantic similarity to obtain the similarity ranking result of the multiple data entities.

[0130] The target data entity matching submodule is used to determine the target data entity that matches the target semantic element based on the similarity ranking results of the multiple data entities.

[0131] The entity matching result output submodule is used to output the entity matching result based on the target data entity that matches the target semantic element.

[0132] In some embodiments, the entity matching module 202 further includes:

[0133] The target association relationship acquisition submodule is used to acquire the target association relationship between target data entities that match different target semantic elements;

[0134] The target association determination submodule is used to execute the step of outputting the entity matching result based on the target data entity that matches the target semantic element when the target association relationship between each target data entity is a preset association relationship.

[0135] In some embodiments, the apparatus further includes:

[0136] The target data table information determination module is used to obtain the target data table that stores the associated data of the target organization, and determine the table name, field names and data values ​​of the target data table;

[0137] The data entity construction module is used to construct multiple data entities of different types based on the table name, the field name, and the data value, respectively.

[0138] An entity relationship construction module is used to obtain the governance relationship information of the target organization and determine the association relationship between the multiple data entities according to the governance relationship information; wherein, the governance relationship information includes one or more of the following: sub-organization affiliation information, organizational personnel affiliation information, and organizational personnel association information of the target organization;

[0139] An organizational knowledge graph construction module is used to construct the organizational knowledge graph based on the relationships between multiple data entities.

[0140] The modules in the aforementioned intent-understanding intelligent agent construction device based on organization-specific knowledge governance can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can invoke and execute the operations corresponding to each module.

[0141] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 3 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores, but is not limited to, related data of a specific organization. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network connection. When executed by the processor, the computer program implements a method for constructing an intent-understanding intelligent agent based on organization-specific knowledge governance.

[0142] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 3As shown, the computer device includes a processor, memory, input / output interfaces, a communication interface, a display unit, and an input device. The processor, memory, and input / output interfaces are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interfaces are used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a method for constructing intelligent agents based on organization-specific knowledge governance and intent understanding. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0143] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0144] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the above-described intention-understanding agent construction method based on organization-specific knowledge governance.

[0145] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described method for constructing an intent-understanding intelligent agent based on organization-specific knowledge governance.

[0146] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the above-described method for constructing an intent-understanding agent based on organization-specific knowledge governance.

[0147] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0148] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0149] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0150] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for constructing an intent-understanding intelligent agent based on organization-specific knowledge governance, characterized in that, The method includes: The process involves obtaining a business information query statement from an account associated with a target organization, and extracting target semantic elements from the query statement. This includes: determining the target language of the query statement; performing sequence labeling on each semantic element in the query statement according to preset semantic syntax rules corresponding to the target language to obtain a semantic syntax tag corresponding to each semantic element; and extracting semantic elements from the query statement whose semantic syntax tags match the target organization to obtain the target semantic elements. The target semantic element is matched with data entities in a pre-constructed organizational knowledge graph to obtain entity matching results. The organizational knowledge graph is used to store organizational business knowledge data associated with the target organization. The entity matching results include target data entities that match the target semantic element. The target data entities are used to represent data query elements that match the business information query statement. The organizational knowledge graph is constructed based on the relationships between multiple data entities. The relationships between the multiple data entities are determined according to the governance relationship information of the target organization. The multiple data entities are constructed based on the table name, field name, and data value of the target data table. The target data table stores the associated data of the target organization. Based on the business information query statement, the target semantic element, and the entity matching result, a data query instruction using a preset structured format is generated; Based on the data query instruction, generate the business information query result of the business information query statement.

2. The method according to claim 1, characterized in that, The data query elements include the target table name, target field name, and target data value. The step of generating a data query instruction using a preset structured format based on the business information query statement, the target semantic elements, and the entity matching results includes: Based on the business information query statement, construct user intent prompt words; Based on the target table name, the target field name, and the target data value, construct query condition suggestions; Based on the target semantic elements, construct semantic prompt words; The user intent prompt, the query condition prompt, and the semantic prompt are input into a pre-trained generative model, which then generates a data query instruction using a preset structured format.

3. The method according to claim 1, characterized in that, The target semantic element has multiple components, and the matching of the target semantic element with data entities in the pre-constructed organizational knowledge graph to obtain entity matching results includes: Determine the semantic similarity between each target semantic element and multiple data entities in the organizational knowledge graph pairwise, and sort the multiple data entities according to the semantic similarity to obtain the similarity ranking result of the multiple data entities; Based on the similarity ranking results of the multiple data entities, the target data entity that matches the target semantic element is determined; Output the entity matching result based on the target data entity that matches the target semantic element.

4. The method according to claim 3, characterized in that, After the step of determining the target data entity matching the target semantic element based on the similarity ranking results of the plurality of data entities, the method further includes: Obtain the target association relationships between target data entities that match different target semantic elements; When the target association relationship between each target data entity is a preset association relationship, the step of outputting the entity matching result based on the target data entity that matches the target semantic element is executed.

5. The method according to any one of claims 1 to 4, characterized in that, The method further includes: Obtain the target data table that stores the associated data of the target organization, and determine the table name, field names, and data values ​​of the target data table; Construct multiple data entities of different types based on the table name, the field name, and the data value, respectively. Obtain the governance relationship information of the target organization, and determine the association relationship between the multiple data entities according to the governance relationship information; wherein, the governance relationship information includes one or more of the following: sub-organization affiliation information, organizational personnel affiliation information, and organizational personnel association information of the target organization; The organizational knowledge graph is constructed based on the relationships between multiple data entities.

6. A device for constructing an intent-understanding intelligent agent based on organization-specific knowledge governance, characterized in that, The device includes: The target semantic element extraction module is used to obtain the business information query statement of the account input agent associated with the target organization, and extract the target semantic elements in the business information query statement, including: determining the target language of the business information query statement; performing sequence labeling on each semantic element in the business information query statement according to the preset semantic grammar rules corresponding to the target language to obtain the semantic grammar tag corresponding to each semantic element; and extracting the semantic elements that match the semantic grammar tags of the business information query statement with the target organization to obtain the target semantic elements. An entity matching module is used to match the target semantic element with data entities in a pre-built organizational knowledge graph to obtain entity matching results. The organizational knowledge graph is used to store organizational business knowledge data associated with the target organization. The entity matching results include target data entities that match the target semantic element. The target data entities are used to represent data query elements that match the business information query statement. The organizational knowledge graph is constructed based on the relationships between multiple data entities. The relationships between the multiple data entities are determined according to the governance relationship information of the target organization. The multiple data entities are constructed based on the table name, field name, and data value of the target data table. The target data table stores the associated data of the target organization. The data query instruction generation module is used to generate a data query instruction in a preset structured format based on the business information query statement, the target semantic element, and the entity matching result. The business information query result generation module is used to generate business information query results for the business information query statement based on the data query instruction.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

9. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 5.

Images