128 results about "Causal maps" patented technology

The invention relates to the technical field of artificial intelligence, can be applied to business scenes of financial science and technology, medical health and the like, and discloses a multi-modal causal reasoning and interpretation method, device, equipment and medium, and the method comprises the steps: obtaining original data streams of at least two different modals, and extracting modal features; a cross-modal attention mechanism is utilized to fuse modal features, and causal features are extracted through feature distillation; constructing a dynamic causal graph based on causal features, and updating an edge weight through a causal intensity function; identifying the causal relationship in the dynamic causal graph and performing anti-factual reasoning verification to evaluate the reliability of the causal relationship; and generating a causal interpretation result in combination with the dynamic causal graph and the causal relationship reliability. According to the method, the multi-modal data are fused, the causal features are extracted, and dynamic causal graph updating and anti-factual reasoning verification are combined, so that reliable modeling and explanation of the causal relationship in a complex scene are realized, the defects of single modal or simple fusion in the prior art are overcome, and the accuracy and interpretability of causal reasoning are improved.

The invention relates to the technical field of artificial intelligence, can be applied to business scenes such as financial science and technology and medical health, and discloses a strategy generation method and device based on hierarchical reinforcement learning, equipment and a medium. And processing environment state information to generate a sub-target and a specific action, generating a strategy reward signal in combination with state change, carrying out joint training and updating on the dynamic causal graph and the hierarchical reinforcement learning model based on the strategy reward signal, and generating an optimized action strategy. The state evolution relation is modeled by constructing the dynamic causal graph, so that the reinforcement learning can obtain causal understanding of the state change trend, decomposition and optimization of sub-targets and actions are realized in combination with a layered reinforcement learning architecture, the response precision and generalization ability of the action strategy in a complex environment are improved, and the method is suitable for application and popularization. Therefore, the task completion stability and the convergence efficiency of strategy training are improved.

The invention discloses a multi-modal harmful model factor detection method based on a thinking chain, and belongs to the field of harmful model factor detection, and the method comprises the steps: firstly obtaining an image-text model factor, and automatically generating the thinking chain containing a problem summary, an image subtitle, and harmful reasoning and conclusion; constructing an annotation data set according to the annotation data set; then, the multi-modal model is finely adjusted in two stages, in the first stage, the visual encoder and the language model are fully finely adjusted to improve image-text understanding, and in the second stage, the visual encoder is frozen, and the language model is finely adjusted through LoRA to strengthen reasoning; introducing reinforcement learning rewards to optimize implicit harmful discrimination; then extracting an entity relationship to construct a knowledge graph and a causal graph, and embedding into the model; and finally, dynamically updating the atlas through an RAG mechanism, so that a detection result adapts to an emerging model cause in real time.

The invention relates to the technical field of financial risk prediction, in particular to a financial risk prediction method and device, a storage medium and equipment. The method comprises the following steps: acquiring financial data, and extracting multi-modal features; based on the multi-modal features, generating a causal graph by adopting a constraint-based causal discovery algorithm, and quantifying causal intensity by adopting a structural equation model; adopting the quantized causal graph as a constraint of an anti-fact generative adversarial network, generating an anti-fact scene and analyzing a risk conduction path; training a causal graph neural network model based on the multi-modal features, the causal graph and the risk conduction path; and interpretability analysis and risk monitoring are carried out. According to the technical scheme, financial risk causal association can be accurately constructed, an anti-fact scene clear conduction path can be generated, and through interpretable analysis and monitoring, the risk prediction accuracy and interpretability are improved, financial risks are helped to be prevented and controlled in time, and the financial system stability and the risk response capacity are enhanced.

The invention relates to the technical field of financial management, and discloses a causal interpretability and illusion suppression method, system and device for text generation, and the key point of the technical scheme is that the method comprises the following steps: S1, obtaining related data according to a generation target, extracting the causal relationship between entities, and constructing a causal map; s2, extracting a causal chain related to the generated target from the causal atlas, and inputting the causal chain and the input variables into the large language model to obtain a target text; s3, performing anti-fact intervention processing on the input variable, inputting the input variable into the large language model, and recording a logic test result; s4, identifying an entity from the target text, comparing the entity with the fact data related to the generated target, and calculating an entity alignment score; and S5, according to the causal chain, the target text, the logic test result and the entity alignment score, generating an interpretation report and performing structured output, so that the output text has higher logicality and higher credibility.

The invention relates to the technical field of artificial intelligence, can be applied to the digital medical field and the financial field, and discloses a causal graph model-based data processing method, device and equipment and a storage medium, the method comprises the steps of obtaining target data and a corresponding to-be-explained reasoning result, and constructing a causal graph according to the target data, obtaining a corresponding structured causal graph; performing causal constraint on the structured causal graph by adopting a causal constraint graph network learning mechanism to obtain an optimized causal graph; and performing anti-fact reasoning on the reasoning result based on the optimized causal diagram, and generating an interpretation result of the target data according to an anti-fact reasoning result. By fusing the causal graph model and the graph neural network technology and introducing the anti-fact reasoning module, accurate causal modeling and explanation of the high-dimensional nonlinear data are realized, the limitation of a traditional causal graph model in the aspect of processing the high-dimensional nonlinear data is solved, and the defect of a graph neural network output result in causal explanation is made up.

The invention discloses a management decision-making method and system based on a knowledge base construction technology. The method comprises the following steps: performing sequential relationship extraction on multi-source financial data to obtain a sequential relationship set related to query content; constructing an event-entity incidence matrix corresponding to the time sequence relation set; according to the time sequence relation set and the event-entity incidence matrix, constructing a dynamic knowledge graph; determining causal effect parameters in the causal graph structure by adopting a dual machine learning model; constructing a structural causal model according to the causal graph structure and the causal effect parameters; and generating an anti-fact prediction result by using the structural causal model, and generating a decision scheme corresponding to the query content based on the anti-fact prediction result. The technical problem that decision information including accurate causal basis and prospective simulation information cannot be generated due to the fact that the causal relationship between financial data is difficult to determine and the intervention effect cannot be dynamically deduced in a related management decision method is solved.

The invention discloses a large model fine tuning method based on a causal graph and thinking chain enhancement and a related device, and relates to the technical field of large model fine tuning in the power industry, and the method comprises the steps: carrying out the causal mining of power equipment data, and constructing a power equipment causal graph containing causal weight information; disassembling the input of the large model into a thinking chain, correspondingly generating a chain type causal pair according to the thinking chain, and performing path retrieval matching and causal consistency check through the chain type causal pair and the constructed electrical equipment causal graph to realize alignment of the reasoning process; and exciting a reinforcement learning process through an alignment result of the reasoning process, optimizing a pre-established reinforcement learning reward model, constraining a thinking chain generation process, guiding the large model to generate a thinking chain under a causal constraint condition, and realizing fine tuning of the large model. According to the method, causal reasoning and causality are embedded into a reinforcement learning feedback process of large model fine tuning, so that the large model can learn a basic causal reasoning rule, and the logicality, the interpretability and the robustness of thinking chain reasoning can be improved.

The invention provides a financial account distribution and anomaly detection method and system based on machine learning, and relates to the technical field of machine learning, and the method comprises the steps: building a causal graph model, generating a time sequence attention weight matrix through a conditional random field, carrying out the weighted fusion, and carrying out the attribution of a difference reason; and processing the reconciliation difference feature vector by adopting a reinforcement learning method, generating and optimizing a processing strategy, and finally outputting a difference processing scheme. The account checking accuracy can be improved, abnormal transactions can be automatically found, the manual processing cost is reduced, and financial risks are effectively prevented.

The invention discloses an unsupervised time sequence missing data filling method based on a dynamic cause and effect graph structure, which comprises the following steps: S1, collecting a data sample, and processing to obtain an initial filling complete data table; s2, clustering all the initial filling complete data tables, and selecting representative data tables; s3, constructing a sub-causal graph of each representative data table, and further constructing a total causal graph; s4, on the basis of a causal relationship in the total causal graph, constructing and training a corresponding weighted expandable depth time sequence convolutional network for the to-be-filled target feature column to obtain a missing value filling model, and further performing coarse filling on missing values; and S5, performing refined secondary filling on the target feature column by using the obtained missing value filling model to obtain a complete data table. The invention provides a filling method without real missing labels, optimization is carried out through a self-consistency or structure maintenance principle, the dynamic dependence and real causal relationship between the time series data can still be captured in a dynamic complex scene, and a better data filling effect is achieved.

The invention provides a geological disaster prediction method and device fusing space-time sequence analysis and causal reasoning, and belongs to the technical field of geological disaster monitoring and early warning. Aiming at the problems of non-uniform data space-time reference, lack of causal logic, poor real-time performance and weak scene adaptability of a model in the prior art, the method comprises the following steps: performing standardization processing on acquired multi-source data, processing missing values by adopting an improved K nearest neighbor algorithm in combination with stratum characteristics, and processing abnormal values through a 3 sigma criterion and geological verification; based on an information theory and an improved SURD algorithm, three types of causal entropies among variables are calculated, a time attenuation coefficient is introduced, a core causal chain is constructed, and a dynamic causal graph is constructed; a core causal variable is used as input, a multi-feature attention-multi-relation space-time diagram recursive network model is constructed, a hour-level predicted value is output through space-time diagram convolution, residual training and a geological physical constraint layer, and'causal-space-time 'fusion is realized through a causal weight adjustment model; the method can be widely applied to early warning of geological disasters such as landslide and debris flow.

The invention provides a rock and soil construction quality monitoring and diagnosing method, which comprises the following steps of: modeling historical engineering events and expert rules, constructing an event causal atlas prototype containing weights, and realizing event sequence feature extraction and real-time causal atlas dynamic updating in combination with on-site multi-modal sensing data and construction logs; neural symbol reasoning and tensor completion technologies are adopted to predict a novel causal relationship, and a causal atlas structure is perfected through space, time and logic consistency verification; a multi-layer risk early warning mechanism is set, a causal map local risk assessment and event chain propagation are combined, spatial positioning early warning signals are generated in a grading manner, closed-loop backtracking optimization is supported, causal reasoning accuracy and early warning efficiency are improved, construction process risk identification and dynamic early warning can be realized, and the engineering safety management level is improved.

The invention relates to the technical field of atmospheric environment prediction, and discloses an atmospheric environment multi-parameter intelligent prediction method based on causal VAE and CNN-LSTM. The method comprises the steps of collecting multi-parameter historical data of an atmospheric environment to form an atmospheric parameter sequence; sequence causal features are extracted, a causal graph structure is constructed, a causal VAE model is trained according to the causal graph structure, and potential causal characterization is generated. And extracting spatial features represented by potential causality through a CNN module, capturing time-dependent features of the spatial features through an LSTM module, fusing the two types of features to obtain a fused feature sequence, and outputting a multi-parameter prediction result through a prediction network. And dynamically adjusting hidden variable distribution parameters of the causal VAE according to the prediction error, updating a causal graph structure according to the adjusted parameters, and re-predicting a subsequent atmospheric parameter sequence by using the updated causal graph. The method can improve prediction precision and generalization ability, and enhance model interpretability and dynamic adaptability.

The invention provides a hot continuous rolling process fault diagnosis method and device based on a multilayer dynamic causal diagram, and relates to the technical field of engineering process monitoring. The method comprises the following steps: according to a historical key variable data set, performing causal relationship analysis by using a dynamic causal entropy algorithm to obtain a multi-layer dynamic causal graph, a causal entropy data set and a quality-related causal entropy data set; training a causal entropy prediction model to be trained by using the causal entropy data set and the quality-related causal entropy data set to obtain a causal entropy prediction model, a fault detection threshold and a quality-related fault detection threshold; according to the actual data, performing fault prediction by using a causal entropy prediction model to obtain fault prediction data; and performing fault analysis based on the fault detection threshold and the quality-related fault detection threshold to obtain a fault detection result and a fault tracing path. The hot continuous rolling process fault diagnosis method is based on the multi-layer dynamic causal diagram and is high in judgment precision and high in interpretability.

The invention discloses a CKD special disease database construction method and system, and relates to the technical field of database construction, and the method comprises the steps: obtaining and predicating a multi-source heterogeneous atomic fact; constructing and fusing three sets of ontology models; verifying and deducing facts in a layered manner according to a multi-layer priority rule base; calling a constraint completion engine to generate inference facts for missing or conflicting fields; streaming follow-up visit, test and intervention logs into events and triggering graded alarms; an anti-fact intervention scene is simulated for the high-risk patient, and pilot groups are screened; dynamically evaluating and adjusting the rule priority; and based on a causal chain and an event mode, clustering patients and constructing a generality map, and recording a whole-process operation chain. Semantic management is realized through fusion of ontology, anti-fact intervention simulation is performed on high-risk patients by means of multi-layer rule verification, and database semantic consistency, reasoning depth, decision support and traceability are improved through causal atlas driven clustering.

The invention provides a causal enhanced soft measurement modeling method based on joint causal topological reasoning. The method comprises the following steps: firstly, obtaining a joint causal weight between an input variable and a target variable based on multivariable joint causal topological reasoning of a Copula theory; secondly, correcting an ordered prior structure based on dynamic feedback of independent confidence, and performing feedback correction on expert knowledge based on data contribution degree; then, based on correlation graph structure enhancement guided by causal weights, the causal weights are fused on the basis of correlation, a causal graph structure matrix is obtained, meanwhile, a fusion causal time sequence graph network is designed, and extraction of causal and time sequence information and final modeling of target variables are achieved with alignment loss as a constraint. According to the method disclosed by the invention, the correlativity in a data form and the causal relationship in an expert experience form can be organically combined to realize soft measurement modeling with high stability and interpretability, and the method can be widely applied to industrial sites with requirements on soft measurement modeling.

The invention relates to a self-interpretation multi-modal personality assessment method based on a multi-agent hierarchical reasoning chain of a large language model, which comprises the following steps of: firstly, unifying multi-modal features such as voice and vision to a text space through modal self-interpretation personality traits representation of a unified space, and generating text description codes related to personality traits; constructing a causal graph model by using a Bayesian causal reasoning theory, analyzing specific contribution of each modal data to personality traits evaluation, and generating a clear reasoning chain and explanation; a two-stage hierarchical reasoning framework is utilized, coarse granularity personality classification is completed in the first stage, psychological measurement norm data is introduced in the second stage, a'classification-scoring 'dynamic mapping mechanism is established, and fine granularity scoring calibration is achieved. Meanwhile, the interpretability of the model is deepened through a hierarchical confidence transfer mechanism, and it is ensured that each scoring result can be traced to the original feature, the classification credibility and the adjustment rule; according to the method, a multi-modal personality assessment special large language model is constructed, and a more transparent, explainable and credible technical basis is provided for personality assessment.

The invention discloses a deep reinforcement learning optimization method for injection molding process parameters, and belongs to the technical field of intelligent manufacturing. The method comprises the following steps: constructing a dynamic causal graph network through information entropy flow analysis and transfer entropy calculation, and revealing a causal relationship and time delay characteristics among process parameters; manifold learning is adopted to map a high-dimensional parameter space to a low-dimensional manifold, and Riemannian metric guide optimization search is constructed based on the quality gradient; generating enhanced state representation fusing causal association and manifold geometric information; identifying a production element state and selecting a corresponding optimization strategy; a geodesic line is planned in a manifold space to obtain an optimal parameter adjustment path; historical experience is utilized through memory retrieval and case adaptation; cross-task knowledge migration is realized; adopting a depth deterministic strategy gradient algorithm to optimize the decision; and online learning is realized through elastic weight consolidation. According to the method, the problems of black box decision, slow convergence, difficulty in knowledge reuse and the like in the prior art are solved, the optimization efficiency and the interpretability are improved, and the method has the capability of quickly adapting to new tasks.

The invention discloses an event causal relationship identification method based on iterative graph prompt learning, and belongs to the technical field of natural language processing. Comprising the following steps: constructing an initial definite causal graph; based on the initial definite causal graph, performing context modeling on a to-be-recognized text, extracting multi-path information of each event pair in the to-be-recognized text, and inputting the multi-path information into a generative language model to perform causal relationship generation to obtain an initial causal relationship result; based on the initial causal relationship result, adopting a graph structure constraint mechanism to carry out multiple rounds of iterative optimization on the initial definite causal graph, dynamically selecting an edge according to confidence, updating the initial definite causal graph, judging whether an iteration termination condition is met or not, and stopping iteration when the iteration termination condition is met to obtain an optimized definite causal graph; and obtaining an event causal relationship result of the to-be-identified text based on the optimized definite causal graph. According to the method, the event causal relationship identification accuracy is improved.

A government official document fair competition review method and system based on causal inference relates to the technical field of data processing, and comprises the following steps: policy text analysis and causal element extraction, causal graph construction, causal effect estimation, anti-fact simulation and index evaluation, and review conclusion generation. According to the fair competition examination method and system for the government official documents, through a method of combining natural language processing, a causal graph model, quantitative effect estimation and anti-fact simulation, automatic, quantitative and explainable intelligent examination on the fair competition risk of the government official documents is realized, and the potential market effect of policy modification can be predicted.

A fraud call recognition method and system based on causal graph and agent collaboration comprises the steps of obtaining a fraud ASR text, mapping extracted multi-category to-be-mapped objects into semantic concept variable vectors, then learning a causal structure based on a data matrix formed by all semantic concept variable vectors, constructing a fraud causal graph, and recognizing a fraud call in the fraud ASR text. Calculating a causal effect of the reason variable on the result variable, and storing the causal effect as a relation weight in a causal graph; the method comprises the steps of obtaining a to-be-recognized ASR text, mapping multiple categories of to-be-mapped objects extracted from the to-be-recognized ASR text into semantic concept variable vectors, then retrieving matched causal chains in a causal graph to obtain a plurality of candidate causal chains, and then generating fraud judgment and risk degree evaluation values based on a large language model inference engine. And determining whether the to-be-identified call is a fraud call. The invention relates to the field of telecommunication anti-fraud, can deeply fuse a causal knowledge base and a large language model, and effectively improves the effectiveness, adaptability and reliability of communication anti-fraud identification.

The invention relates to the technical field of target behavior prediction systems based on causal inference and multi-task learning, and particularly discloses a target behavior prediction system based on causal inference and multi-task learning. The system comprises a central coordination server and a plurality of participant clients, constructs a global causal graph through a federated causal discovery algorithm in a collaborative manner, determines a causal feature subset of each prediction task, and carries out multi-task model training through a federated average algorithm under the constraint. In the process, differential privacy and homomorphic encryption technologies are comprehensively applied to protect data privacy. According to the method, more accurate causal discovery and more reliable prediction model training can be realized on the premise of protecting data privacy of all parties.

The invention relates to the field of data processing and information technology operation and maintenance, and discloses an intelligent operation and maintenance decision-making method and system based on multivariate heterogeneous data fusion. Comprising the following steps: constructing a dynamic causal graph reflecting a system state and an environmental factor causal relationship based on a causal structure learning algorithm; constructing a multi-agent game decision model based on the dynamic causal diagram, and respectively setting a global operation and maintenance target and a component local demand as a leader strategy and a follower strategy of the game; and solving game equilibrium by using multi-agent reinforcement learning, and outputting an optimal joint operation and maintenance decision instruction. The system is composed of a data acquisition preprocessing module, a multi-modal feature fusion module, a causal structure learning module, a game decision solving module and an execution monitoring module. According to the method, the causal inference and the game theory are fused, so that the causal logic can be accurately extracted, the decision robustness and interpretability are improved, and the global optimal collaborative configuration is realized.

The invention relates to the technical field of case investigation, and discloses a physical evidence intelligent identification and automatic drawing method and system based on deep learning, and the method comprises the steps: constructing a physical evidence behavior bidirectional causal graph, representation learning and causal relationship modeling of physical evidence nodes and behavior nodes are realized; establishing a physical evidence and behavior joint representation learning framework, and mapping physical evidence features and behavior features to a unified semantic space; constructing a human body behavior simulator based on physics; bidirectional reasoning from physical evidence to behavior and from behavior to physical evidence is supported, and consistency is optimized; constructing a scene hypothesis verification system, and generating and verifying case scene hypotheses; developing an automatic drawing system, and visually presenting a reasoning result; according to the method, the bidirectional reasoning capability from physical evidence to behavior and from behavior to physical evidence is realized through the physical evidence behavior bidirectional causal graph and the bidirectional reasoning engine, and the problem of reasoning unidirectivity of a traditional method is solved.

The invention discloses a causal decoupling method and device based on multi-scale noise and adversarial supervision, and the method comprises the steps: carrying out the simulation of the causal relationship of variables in a causal graph, so as to generate observation image data, and constructing a training set and a test set according to the observation image data, the corresponding causal label information and the causal graph; constructing a causal decoupling model, and performing adversarial supervision training under multi-scale noise by using the training set; and obtaining anti-fact intervention data by using the test set and the trained causal graph matrix and utilizing the trained observation data coding module and observation data decoding module. According to the method, an auto-encoder and causal acyclic constraints are fully combined, the discrimination module is trained under multi-scale noise, and high-quality adversarial supervision is performed, so that the model representation learning ability is improved, the representation understanding of the model on data with causal relationships is enhanced, the accuracy of implicit causal network prediction is improved, and the prediction efficiency is improved. And the causal decoupling accuracy is improved.

The invention belongs to the technical field of natural language processing and artificial intelligence, and relates to a biomedicine named entity recognition method based on causal graph guided anti-fact analysis, which comprises the specific steps of causal graph construction, false node anti-fact analysis, false link anti-fact analysis, consistency constraint and model training and optimization. According to the method, a dual anti-fact analysis mechanism based on syntactic analysis and adversarial disturbance is introduced into large-scale language model training, so that false factors in input features can be effectively identified and intervened. Therefore, the model can focus on core features with real causal association in an entity identification task, so that the robustness and the cross-domain generalization ability of the model are remarkably enhanced. Experimental results prove that the Micro-F1 scores on a plurality of biomedical named entity recognition reference data sets are all advanced to the prior art.

The invention relates to a causal structure discovery method based on evolutionary neural architecture search and reinforcement learning, and the method comprises the steps: obtaining to-be-detected observation data, inputting the to-be-detected observation data into a causal structure detection model, and obtaining a directed acyclic graph composed of highest confidence coefficient edges; the causal structure detection model is obtained by training an Actor-Critic model by using a training set; the Actor-Critic model comprises an actor network model and a commentator network model; in the training process, the model configuration of the Actor-Critic model is optimized based on an evolutionary algorithm, and the optimal configuration is obtained; and inputting the observation data in the training set into the actor network model under the optimal configuration to obtain a candidate directed acyclic graph, and evaluating the candidate directed acyclic graph by the commentator network model under the optimal configuration to assist in strategy optimization to obtain a candidate directed acyclic graph. And strengthening the actor network model under the optimization strategy through a reinforcement learning algorithm to tend to output an optimal causal graph structure.

The invention relates to the field of multi-agent reinforcement learning, and discloses a reinforcement learning method based on causal reasoning and a hierarchical attention mechanism, which makes up the defects of inaccurate role allocation and low cooperation efficiency in a traditional multi-agent system and improves the overall cooperation performance. According to the method, a causal perception multi-agent cooperation model is constructed, firstly, a dynamic causal graph is constructed by using a structural causal model, and a causal influence vector between agents is calculated by using an optimized variational distribution estimator; then generating a causal-guided hierarchical attention weight based on a causal influence vector; then cooperative information transmission is realized through a cross-agent attention sharing mechanism, and enhanced feature representation is generated by using a cross-layer fusion mechanism; a causal perception role selector is further designed based on the global causal contribution degree and the local causal contribution degree, and dynamic allocation of agent roles is achieved; and finally, performing model training optimization through a causal influence internal reward mechanism.

The invention discloses a data identification method and system based on data high-dimensional feature deconstruction. The method comprises the following steps: decoupling multi-modal time series data into independent implicit factors with clear physical meanings through a domain knowledge constrained depth generation model; based on the implicit factor sequence, using time sequence causal discovery and anti-fact intervention to construct a causal graph and generate a causal effect vector; constructing a differentiable identification strategy network taking the implicit factor and the causal effect as input, and optimizing system meta parameters through a meta learner according to feedback; robustness loss feedback optimization is generated through causal consistency verification and abnormal injection, meanwhile, a structured identification graph is output, and an incremental model library is established. The system correspondingly comprises four function modules. According to the method, the problems that the features cannot be explained, the causal mechanism is missing and the model adaptability is insufficient are solved, and intelligent data identification which can be explained and is robust and has the sustainable evolution capacity is achieved.

The invention discloses a question and answer method and system based on large model causal graph discovery and causal graph enhanced reasoning, and belongs to the technical field of natural language processing and artificial intelligence. The method comprises the following steps: firstly, carrying out hierarchical expansion from an initial root node by adopting a breadth-first search strategy, and efficiently constructing a reliable causal graph conforming to directed acyclic graph constraints with linear complexity in combination with real-time loop detection; in the question and answer stage, key entities in user questions are analyzed, and causal paths connecting the key entities are retrieved in a causal graph; and taking the retrieved structured causal path as a constraint condition to be injected into a decoding process of a large-scale language model, and generating a natural language answer which is strict in logic and can trace reasoning steps. According to the method, the resource consumption of large-scale causal discovery is remarkably reduced, the accuracy and interpretability of the answers of the questions and answers are effectively improved, and technical support is provided for medical treatment, finance and other scenes needing high-reliability reasoning.