255 results about "Business knowledge" patented technology

The invention relates to the technical field of artificial intelligence, and discloses a multi-agent collaborative data question-answering system and method based on a large model, and the system comprises an agent cluster module which is composed of five special agents, namely a data question-answering agent, a data extraction agent, a data analysis agent, a visualization agent and a quality examination agent, and achieves the task decomposition and collaborative execution through dynamic scheduling; the knowledge management module comprises a business knowledge base, a data element knowledge base and a user feedback base, and adopts a hierarchical knowledge fusion technology to provide domain knowledge support for the intelligent agent; and the supporting function module covers a front-end dialogue component and a verification and execution engine and is responsible for interactive interface rendering and result reliability verification. According to the method, the fine tuning requirement on the large model is remarkably reduced, the illusion of the large model is effectively intercepted through a dual verification mechanism, and the accuracy and reliability of question and answer results are improved.

The invention discloses a water conservancy intelligent question-answering system and method based on knowledge enhancement and data driving, and aims at water conservancy business structured and unstructured data query, through loop optimization driven by positive and negative examples, precise classification of secondary intentions of water conservancy problems is realized, knowledge questions and answers, data query and professional water conservancy subclass questions and answers can be distinguished, and the system and the method can be applied to water conservancy business. And the semantic analysis efficiency is improved. Aiming at the problem of low query accuracy of retrieval enhancement generation in the field of water conservancy, a differential water conservancy knowledge base oriented to professional books, industrial standards and laws and regulations is constructed, local and networking information is processed through a multi-source knowledge fusion and conflict resolution mechanism, and the accuracy, interpretability and traceability of question and answer content are improved. Aiming at the problems of complex operation and low semantic query accuracy in query of massive water conservancy business data and monitoring data, multi-layer constraint Text-to-SQL conversion is performed based on water conservancy business knowledge, high-precision semantic query and automatic visual output of water conservancy structured data are realized, and query efficiency is improved.

The invention provides an intelligent question and answer method and device based on large model multi-source information fusion and electronic equipment, relates to the technical field of computers, in particular to the fields of big data, intelligent search and the like, and can be used for application scenes such as intelligent question and answer. According to the specific implementation scheme, pre-retrieval is conducted from a data table structure information base and a data table field information base according to a user problem, and a pre-retrieval result is obtained; retrieving from a business knowledge base, a business rule base and a question number template base according to the user question to obtain a knowledge enhancement result; according to the pre-retrieval result and the knowledge enhancement result, utilizing a query instruction generation model to obtain a query statement; and querying from the data table set according to the query instruction to generate an answer corresponding to the query word. The scheme can improve the professionality of intelligent questioning and answering, and is suitable for application scenes of various industries.

The invention relates to the field of discrete manufacturing intelligence, in particular to an intelligent decision-making and business collaboration method based on an OAG ontology and an LLM large model, which comprises the following steps: collecting historical data and business documents, and performing directional fine tuning on a basic large language model to form the LLM large model; the method comprises the following steps: analyzing discrete manufacturing scene original data and business documents, and constructing an OAG ontology library; receiving field data in real time through the LLM large model, updating the OAG ontology library, and performing feedback optimization on the LLM large model to form bidirectional feedback; integrating real-time data and historical data, inputting the data into an LLM large model, converting the data into business knowledge through layering of an OAG ontology library, and generating a main and standby decision scheme; and establishing an agent federated center, issuing a decision scheme, collecting execution data and feeding back an LLM large model, and realizing cross-scene collaboration and full-process data closed loop. According to the invention, through a technical path of a whole-process data closed loop, a core pain point that an existing LLM does not understand a business is effectively solved, and whole-link value conversion of discrete manufacturing data is realized.

The invention relates to the technical field of system monitoring, and discloses an information system full-link monitoring method based on high-frequency index acquisition optimization, which comprises the following steps: monitoring the running state of an information management system in real time, dynamically adjusting the sampling frequency by means of a customized service key identification component and a comprehensive load prediction model, and performing real-time monitoring on the sampling frequency. Transmitting the target data to the edge computing node; a lightweight monitoring agent is deployed at an edge node, and a wavelet signal decomposition algorithm is adopted to extract features and distinguish data types; constructing an information management business knowledge graph and an entity-relation-business rule base, associating abnormal features, and generating an abnormal root cause report in combination with a time sequence prediction model and a knowledge constraint large language model; based on report and information service priorities, monitoring resources are dynamically allocated in the edge-cloud collaborative architecture, and related model parameters, service association rules and constraint weights are optimized according to operation and maintenance feedback. According to the invention, targeted monitoring requirements in the business peak period and efficient utilization of system resources can be met at the same time.

The invention relates to an auxiliary verbal skill self-adaptive generation method fusing a knowledge graph and reinforcement learning, and belongs to the technical field of artificial intelligence and service compliance management crossing. The method comprises the following steps: collecting service data, extracting entities, attributes and relationships among the entities through natural language processing, constructing a knowledge graph marked with compliance attributes, and establishing an updating mechanism; determining a verbal skill generation main body according to an actual service scene, and integrating a conversation context, a client real-time state and a business target construction environment state; the verbal skill generation subject retrieves associated business knowledge according to the current environment state, and after the seat selects adaptive verbal skill interaction, verbal skill generation strategy parameters are updated through deep reinforcement learning; and constructing a multi-dimensional effect evaluation method based on the business data, feeding back an evaluation result to an iteration link, and optimizing reward calculation and knowledge graph association logic. Dynamic self-adaption of verbal skills in multiple service scenes is realized, and customer demand solving efficiency and business conversion quality are improved.

The invention discloses a data processing method and device, a storage medium and electronic equipment, and relates to the technical field of artificial intelligence, first type memory information is obtained according to the current session context with a user, a model can be ensured to understand the background of a current problem, the problem background is supplemented by using second type memory information, namely the historical session record of the user, and the user experience is improved. According to the method, the service knowledge information related to the question is retrieved, namely the target knowledge information, the target model generates the comprehensive reply according to the information, the context of the instant session is considered, and the personalized preference of the historical session and the current service rule are combined, so that the generated reply information is more comprehensive, and the reply efficiency is improved. Therefore, incoherence or inaccuracy caused by forgetting important details is avoided, and the technical effects of improving the accuracy of problem solving and enhancing individuation and satisfaction of user experience are achieved.

The invention relates to the technical field of data analysis, and discloses a feedback generation method and device based on a business knowledge graph, equipment and a medium, and the method comprises the steps: collecting multi-source business data, cleaning and segmenting the multi-source business data to form standardized text fragments, and carrying out the entity recognition and relation extraction of the text fragments to construct the business knowledge graph; receiving a natural language query instruction, extracting query intention features and key entity references, mapping the key entity references into the business knowledge graph, and performing association traversal under the guidance of the query intention features to obtain associated knowledge sub-graphs; and serializing the associated knowledge sub-graph into a knowledge background context, and jointly inputting the knowledge background context and a natural language query instruction into a pre-training language generation model to generate natural language feedback information. The method can be applied to business scenes such as financial science and technology, medical health and the like, semantic structuring is carried out on text knowledge, reasoning is carried out in knowledge association in combination with query semantics, accurate question answering is achieved, and knowledge retrieval efficiency is improved.

The invention relates to an alarm rule automatic generation and closed loop optimization method and system, and relates to the field of computer system monitoring and testing. The method comprises the following steps: acquiring and cleaning a cluster to generate a standardized enhanced alarm event table based on multi-source heterogeneous data fusion; performing feature extraction on the enhanced alarm event table, constructing a feature vector, performing feature importance analysis on the feature vector by using a tree model, and outputting the feature importance analysis; according to statistical distribution of features in the structured feature vector, automatically generating a rule for identifying an alarm scene in combination with business knowledge; converting the alarm scene rule into a structured format, performing logic conversion through a test case generation engine, and outputting an automatic test case; and executing the automated test case and verifying the accuracy of the automated test case so as to dynamically optimize and adjust the alarm scene rule. According to the method, automatic rule generation and closed-loop optimization of a complex alarm scene can be realized, the alarm test efficiency and the scene recognition accuracy are remarkably improved, and the manual intervention requirement is reduced.

The invention relates to a business knowledge base construction method, device and equipment for patrol operation and maintenance of a power production environment and a medium. The method comprises the following steps: acquiring an electric power inspection task, and encoding according to the electric power inspection task to obtain a task semantic vector; obtaining a preset knowledge base matrix and a state transition probability matrix; calculating a matching score of the task semantic vector and each state node; selecting state nodes corresponding to the first K highest matching scores as target state nodes, wherein K is a positive integer greater than or equal to 1; extracting a path taking the target state node as a starting point according to the state transition probability matrix to obtain a task matching path; wherein all the task matching paths form a task matching path set; and calculating a risk score of each task matching path to obtain a path risk score, and according to the path risk score, generating an auxiliary decision-making suggestion of the electric power patrol task. By adopting the method, the intelligent management capability of the operation state of the power equipment can be improved.

The invention discloses a customer complaint work order automatic circulation system and method based on composite intention disassembly. The system receives and preprocesses the multi-mode customer complaint data and initializes a service circulation context; business intention identification and business key slot position extraction are carried out; disassembling the composite business intention, and constructing an ordered work order execution queue of a directed acyclic graph structure through cyclic dependence verification; evaluating the work order circulation action with the maximum expected cumulative income by using a reinforcement learning model; when business knowledge query is involved, compliance evidence fragments are retrieved, sorted and output; and generating a candidate reply and a business operation instruction by the fine-tuned large language model, outputting the candidate reply after business compliance verification, and calling an API (Application Program Interface) of an underlying business system to execute entity business handling operation in a cross-system manner. The problems of data interaction and state synchronization among multi-source heterogeneous systems are solved, the automatic execution efficiency of the composite work order at the bottom layer node is improved, and the system unauthorized and dirty data risks caused by an uncertain instruction are effectively avoided.

The invention provides an intelligent document understanding method and system combining natural language processing and deep learning, and relates to the technical field of natural language processing and information management. Fusing the semantic pre-annotation result and the original text feature of the document, inputting the fused semantic pre-annotation result and the original text feature of the document into a cross-modal semantic enhancement model to obtain a document enhanced semantic representation, performing bidirectional semantic interaction with a dynamic business knowledge network based on the document enhanced semantic representation, generating an associated interaction result, and constructing a structured semantic asset; finally, the structured semantic assets are input into an intelligent document application engine, semantic feedback data are collected to optimize model parameters, document understanding accuracy and practicability can be improved, and diversified business requirements are met.

The invention provides an RAG retrieval enhanced SQL generation method based on a multi-dimensional knowledge base, and the method comprises the steps: respectively constructing a technical metadata knowledge base, a business knowledge base and an experience knowledge base based on a target database; respectively slicing knowledge contents of the technical metadata knowledge base, the business knowledge base and the experience knowledge base; receiving a query request of a user, and simultaneously performing parallel retrieval from the technical metadata knowledge base, the business knowledge base and the experience knowledge base to obtain a retrieval result; a large language model is called based on the retrieval result, SQL is output in a staged SQL generation mode, and the large language model adopts a multi-model collaborative architecture and adopts a gradient lifting algorithm for iterative optimization. Through the organic combination of the technical metadata knowledge base, the business knowledge knowledge base and the experience knowledge knowledge base, a comprehensive database query knowledge system is constructed, and the problem that knowledge coverage is not comprehensive in a traditional method is effectively solved.

The embodiment of the invention provides a business data processing method and device, electronic equipment and a storage medium, and relates to the technical field of data processing. The method comprises the following steps: generating a query semantic vector according to a data query problem; performing table data query operation in the business knowledge base according to the query semantic vector to obtain data table information; generating an initial cue word according to the data table information; obtaining a to-be-executed structured query statement according to the initial cue word, the data query problem, the business database and the fine-tuned proprietary large model; inputting the to-be-executed structured query statement into a preset executor, and when the execution result of the preset executor executing the to-be-executed structured query statement is successful, executing query operation from the business database to obtain preliminary result data; and performing presentation processing according to the preliminary result data to obtain to-be-presented data. Conversion from a natural language data query problem to a structured data query problem is realized, and the technical effect of enhancing the accuracy of a query result is achieved.

The invention belongs to the field of artificial intelligence and big data analysis, and particularly relates to an AI-based drive type data analysis method and system.The method comprises the steps that S1, on the basis of natural language query input by a user, candidate SQL and / or DSL statements are generated through a pre-trained big language model; s2, generating a parallel query plan and optimizing the parallel query plan based on the candidate SQL and / or DSL statements and the metadata knowledge graph; s3, extracting authorization data; s4, fusing the extracted data to generate an analysis feature wide table, and identifying key features, abnormal modes and text abstracts in the analysis feature wide table; s5, executing controlled retrieval in the business knowledge graph based on the key features and the abnormal mode, and generating a security evidence sub-graph; s6, inputting the text abstract of the analysis feature wide table and the security evidence sub-graph into an AI inference engine, and outputting a structured report; and S7, displaying the structured report. The problems of high technical threshold and weak safety and interpretability in a data analysis process are solved.

The invention discloses a test case automatic generation method and device, and relates to the technical field of software testing, and the method comprises the steps: S1, constructing a system behavior graph and a business knowledge graph; s2, integrating the system behavior graph and the business knowledge graph into a multi-layer behavior graph; s3, acquiring test demand data; s4, performing retrieval in the multi-layer behavior map according to the demand data to obtain a retrieval result; s5, fusing the demand data and the retrieval result, and generating an enhancement prompt; s6, generating a test case according to the enhancement prompt; a system behavior graph and a business knowledge graph are independently constructed, and a double-graph technical architecture of cooperative work is adopted, so that double guarantee of technical correctness and business rationality is realized; complete contexts can be obtained from two dimensions of technology implementation and business rules simultaneously through a double-atlas collaborative retrieval technology; high-level business requirements, UI design and specific implementation logic of a system bottom layer are effectively connected, and real efficient and full-automatic test case generation is achieved.

The invention provides a lightweight configurable cue word labeling method and system for a Web system, and relates to the technical field of Web front-end development and man-machine interaction, and the system comprises a front-end labeling plug-in module, a cue word content service module, a cue word management module and an intelligent recommendation module. The front-end labeling plug-in captures text content selected by the user by monitoring a global shortcut key, and allows the user to add prompt words of a text, a rich text or a link type; the prompt word content service module provides an API (Application Program Interface) to realize persistent storage of data; the prompt word management module provides a unified management interface for an administrator; and the intelligent recommendation module automatically judges the prompt word display priority through a candidate recall and sorting algorithm based on the user click behavior data, and continuously optimizes the recommendation result by using an online learning mechanism. The method is integrated to an existing Web system in a non-intrusive mode, business core codes do not need to be modified, the use threshold of the system is remarkably lowered, and precipitation and sharing of business knowledge are promoted.

The invention relates to the technical field of data processing, and particularly provides a double-engine-based SQL (Structured Query Language) generation method, system and device and a medium, and the method comprises the following steps: receiving a natural language query input by a user; analyzing the natural language query by using an intention understanding engine in combination with a dialogue context and a business knowledge graph to generate a structured query intention representation; utilizing an SQL generation and optimization engine to generate at least one candidate SQL statement based on the structured query intention representation and a vectorization database Schema library; performing execution plan analysis on the at least one candidate SQL statement, and performing performance evaluation and optimization rewriting based on an analysis result to generate a final high-performance SQL statement; and executing the final high-performance SQL statement and returning a query result. According to the method, the query success rate and the user experience in a complex scene are greatly improved.

The invention discloses an attack behavior analysis engine construction method and system based on deep learning, and belongs to the technical field of power system network security. The method comprises the following steps of: firstly, performing deep syntax tree analysis on an electric power industrial control protocol by constructing a multilayer feature extraction module for business semantic decoupling, and realizing semantic mapping from a network message to a business operation intention and generating a three-dimensional feature tensor in combination with a pre-constructed electric power business knowledge graph; secondly, designing a dynamic adversarial training mechanism, synthesizing a high-simulation adversarial sample by adopting a Wasserstein generative adversarial network obeying power business logic constraints, and continuously optimizing the robustness of the detection model; and finally, establishing a topology-aware graph neural network attack detection model, converting an equipment connection relationship of physical scenes such as a transformer substation into a graph structure, and analyzing a cross-node attack propagation chain by using a graph convolutional network. According to the method, the detection precision and the response speed of complex attacks such as advanced persistent threats are remarkably improved.

The invention provides a code review method and system based on a multi-dimensional knowledge graph and electronic equipment. The method comprises the following steps: constructing a business knowledge graph, a method calling graph and a code review knowledge graph; code data are obtained, a target code in the code data is determined based on the code review request, and the target code comprises code change content; generating a large model cue word in combination with the business knowledge graph, the method calling graph and the code review knowledge graph; and according to the big model cue word, calling an AI big model to perform joint review on the code change content, and generating a review report which at least comprises a repair suggestion. According to the method, multi-dimensional knowledge of business, code calling and code review is integrated, a more comprehensive and closely associated knowledge system is formed, the generated cue word can accurately guide the AI large model to review code change in combination with multi-field knowledge, blindness of the large model during review is avoided, and review efficiency and accuracy are improved.

The embodiment of the invention provides an intelligent prediction and decision-making method for military demand material guarantee, and the method comprises the steps: constructing a multi-source military demand data fusion framework system, and integrating multi-dimensional material demand feature information; on the basis of military rules and business knowledge, defining and quantifying feature vectors of different combat scenes, and establishing a scene classification model; constructing a military demand large model based on scene dynamic supply, and fusing scene information and material demand feature information by adopting a layered attention mechanism; implementing a multi-stage training strategy to complete military demand large model training; and based on a prediction result of the military demand large model, constructing a dynamic supply decision module based on reinforcement learning, and realizing real-time optimization of a supply scheme. The embodiment of the invention has the advantages of high prediction precision, strong adaptability, decision-making intelligence, high resource utilization rate and good expandability, and realizes intelligent prediction, dynamic planning and accurate supply of military supplies.

The invention discloses a business data question and answer method and device, equipment and a storage medium, and relates to the technical field of natural language processing, and the method comprises the steps: carrying out retrieval recall in a plurality of preset knowledge tables based on a plurality of preset table classification conditions, obtaining a plurality of recall knowledge tables related to the business question text of the target business under each preset table classification condition; determining a target knowledge table from the plurality of recall knowledge tables based on a text similarity index, a query permission index, a query frequency index and a knowledge credibility index between each recall knowledge table and the business question text; inputting the business question text into a business question rewriting model to carry out business knowledge rewriting on the business question text to generate a standard question text; inputting the standard question text and the identification information of the target knowledge table into a statement generation model to generate a query statement; and generating a service answer text based on the service question text and the query statement. By means of the technical scheme, the efficiency and accuracy of business data question answering can be improved.

The invention provides a business auditing method and device, a computer readable storage medium and a computer program product. The method can be applied to the field of financial science and technology. The method comprises the following steps: receiving service information of a to-be-processed service sent by a front-end page; performing natural language conversion on the service information to obtain converted service information; based on the service identifier of the to-be-processed service, acquiring general service knowledge of the to-be-processed service from a knowledge base, and acquiring a service auditing rule of the to-be-processed service from a rule database; filling the converted service information, the general service knowledge and the service auditing rule into a set position in a pre-constructed service auditing prompt word template to generate a target prompt word; and inputting the target prompt word into a preset large model to obtain an audit result of the to-be-processed service, and feeding back the audit result to the front-end page. According to the method, the business auditing efficiency and accuracy are improved.

The invention provides a verbal skill recommendation method based on context and intention prediction, and the method comprises the steps: introducing intention recognition and a potential subsequent intention prediction mechanism based on a affair graph, and fusing the current input information, a business knowledge context and a multi-round dialogue context corresponding to multi-round dialogue historical data; and the current intention of the user is responded, and a more accurate target reply verbal skill is generated. Then, the intention evolution of the potential follow-up intention of the current user is further predicted based on the affair graph, so that the potential demand of the user is pre-judged, and the target guide verbal skill is generated aiming at the potential demand, so that the verbal skill pertinence of the generated verbal skill in a complex dialogue scene is improved through a recommendation mode of combining the reply verbal skill and the guide verbal skill; therefore, the verbal skill generation accuracy is improved, and the user experience and the message reply efficiency are improved. The method can be applied to a seat verbal skill recommendation function in the financial field or the medical field, so that the seat verbal skill generation accuracy is improved.

The invention relates to a dialogue type visual chart generation method and system, belongs to the technical field of application intelligence, and solves the problems of low security and poor usability of a BI system based on a large language model in the prior art. The method comprises the following steps: receiving a visual chart generation demand input by a natural language of a user; enterprise database configuration information related to the demand is extracted from a pre-configured refined business knowledge base; using a large language model to generate SQL query statements meeting requirements according to enterprise database configuration information and general query constraint conditions; executing the SQL query statement in the enterprise database to obtain structured data of a query result; and generating the structured data into a visual chart meeting the requirements by using a large language model. According to the method, automatic adaptation of the large language model and the enterprise database is realized through multi-module collaborative optimization, and the user experience is improved while the safety of the enterprise database is ensured.

The invention discloses a scene perception information fusion decision-making method for industrial field multi-source heterogeneous data, which comprises an edge layer, a cloud layer and an interaction layer which form a three-level collaborative architecture, and belongs to the technical field of industrial field data processing. The method comprises the specific steps that S1, an edge layer collects industrial field multi-source heterogeneous original data and carries out self-adaptive preprocessing to obtain cleaned multi-source feature data; s2, the edge layer processes the multi-source feature data through a lightweight scene perception model, and outputs a preliminary scene judgment result; s3, the cloud layer constructs and updates a causal relationship graph based on industrial business knowledge and incremental data of the edge layer, processes data uploaded by the edge layer through a causal reasoning fusion model, outputs a decision result and issues the decision result through the interaction layer; and S4, the cloud layer optimizes parameters of the causal reasoning fusion model and the lightweight scene perception model based on a decision execution result fed back by the edge layer, and issues the optimized parameters to the edge layer through the interaction layer to realize dynamic iteration.

The invention discloses a business full-process generation system and method based on an enterprise knowledge graph enhancement model, and belongs to the technical field of intelligent software codes, and the system comprises a perception and understanding engine which is used for receiving and analyzing the multi-modal original data of an enterprise by using a preset industry knowledge enhancement model, obtaining analysis content, and generating an enterprise knowledge graph enhancement model; constructing a business knowledge graph based on the analysis content; the design and planning engine is used for performing automatic design of the system architecture based on the business knowledge graph to obtain a design scheme blueprint; and the construction and delivery engine is used for generating a plurality of code components based on the design scheme blueprint and assembling and deploying all the generated code components to obtain a software system. The system can improve the current situation of serious disjunction between service cognition and software implementation.

The invention discloses a data processing method and device in agent dialogue, equipment, a medium and a product, and relates to the technical field of big data. According to the method, the user attribute, the service rule, the scene information and other structured knowledge are extracted in response to the dialogue data and stored in the dynamic cache layer, and the timeliness tag is added, so that the problem of dialogue memory loss of an existing agent dialogue system is effectively solved, and the interaction efficiency and the user experience are improved. Then, target structured knowledge is screened in combination with dialogue relevance and business relevance and then classified and stored according to a three-level hierarchical storage architecture, and then the value of the memory unit is evaluated based on timeliness, scene matching degree and cross-session relevance strength and retrieved to generate response data. Therefore, physical isolation of user features and business knowledge is realized through three-level hierarchical storage, mutual interference of different types of data during retrieval is reduced, retrieval accuracy and efficiency are improved, and timely response of an intelligent agent dialogue system to user requirements is guaranteed.

The invention provides an intelligent plan generation method and system based on RAG, and belongs to the technical field of artificial intelligence, and the method comprises the steps: carrying out the fusion processing of collected multi-source risk data; performing semantic vectorization representation on the multi-source risk fusion data; performing similarity recall on the business knowledge base according to the semantic vector; generating cue words based on the semantic vector and the recall data, and inputting the cue words into a risk plan generation model based on a large language model to generate a target risk plan; and in the plan execution stage, when the plan deviation between the expected effect of the target risk plan and the actual execution result exceeds a preset threshold value, performing plan updating according to the reinforcement learning model. The method has the beneficial effects that in the plan execution stage, the quantitative feedback index based on plan deviation and the reward function of reinforcement learning jointly form a dynamic optimization closed loop of the plan, it is ensured that the plan can adapt to constantly changing actual situations, and the flexibility and effectiveness of the risk business plan are improved.

The embodiment of the invention provides a multi-modal data retrieval method and device, equipment, a medium and a product, and the method comprises the steps: obtaining a natural language problem of a user, carrying out the intention understanding of the natural language problem, and obtaining the query intention of the natural language problem; performing retrieval path selection on the query intention, and determining at least one target modal type matched with the natural language question; the natural language question is distributed to vector databases corresponding to the target modal type to be retrieved, and candidate multi-modal retrieval results of the natural language question are obtained; and verifying the candidate multi-modal retrieval result according to a pre-constructed multi-modal business knowledge graph to obtain a target multi-modal retrieval result of the natural language problem. By utilizing the method, a collaborative mechanism of semantic comprehension, routing decision and knowledge enhancement is introduced, so that the cross-modal semantic comprehension capability and the information retrieval performance are remarkably improved.