588 results about "Finite-state machine" patented technology

The invention discloses a hybrid enhanced enterprise smart law platform system based on a generative language large model. Four modules including a hybrid enhanced legal knowledge engine, a multi-modal legal document analysis module, a risk quantitative evaluation module and a compliance verification workflow work cooperatively. The hybrid enhanced legal knowledge engine integrates multi-source data, realizes real-time updating and semantic reasoning, and comprises map construction, a rule base and an incremental learning mechanism; the multi-modal legal document analysis module performs structured analysis on the heterogeneous document to generate a feature vector; the risk quantitative evaluation module is combined with Monte Carlo simulation and an analytic hierarchy process, quantifies the risk according to a compliance reference and analysis characteristics, and outputs a thermodynamic diagram and a report; a compliance verification workflow is driven by a finite-state machine, a verification module and a conflict detection module are integrated, a generative language large model is called to generate an improved scheme, and audit records are solidified and fed back for optimization. And the system runs according to the processes of analysis, supply rule, bias calculation and verification correction, so that the intelligence and accuracy of legal affair processing are improved.

The invention relates to a campus Internet of Things terminal scheduling method based on digital twinning. The method comprises the following steps: defining a general twinning state representation structure; introducing a state mapping relationship, and establishing a one-to-many dynamic mapping relationship between a physical state and a virtual twinning state by using a finite state machine; when the state of the physical terminal changes, triggering multi-level response mapping of the twin in the virtual space; semantic fusion is carried out on twinborn real-time data, a scheduling sensitivity index is introduced, disturbance simulation is carried out based on past scheduling behaviors, and the potential influence of a certain scheduling behavior on the whole system is analyzed; carrying out local division on the situation map; using Bayesian estimation and Monte Carlo combination to sample and evaluate the feasibility and income of each solution set in a period of time in the future; simulating the expected influence of each scheduling strategy through a twin body, and selecting a scheduling path; unified modeling and semantic fusion of multiple types of terminals are realized, the intelligent level and predictive ability of scheduling decision are improved, the collaborative scheduling ability of the terminals is enhanced, and coupling unit level regulation is realized.

The invention discloses a regional energy internet hierarchical control method considering flexible loads, which comprises the following steps of: constructing a global layer, dividing a regional energy internet into a plurality of electrically strongly coupled sub-regions by adopting a dynamic partitioning algorithm, aggregating equivalent adjustable capacities of the flexible loads in the sub-regions, and performing rolling optimization in an hour-level time scale, an improved Benders decomposition algorithm is adopted to accelerate solving and generate a partition-level power regulation instruction; constructing a regional layer, designing a thermodynamic-economic hybrid constraint model according to flexible load physical characteristics and user behavior elasticity, introducing a user comfort recovery time constraint and elastic electricity price response function, and decomposing an adjustment instruction issued by the global layer to each flexible load cluster in a minute-level time scale by adopting a parallel ADMM optimization algorithm; and constructing a load layer, constructing a millisecond-level adjustable load priority switching mechanism based on FPGA hardware acceleration and finite-state machine logic, calculating a load adjustable power boundary in real time, and dynamically adjusting a response queue in combination with system frequency deviation.

The invention discloses a road three-dimensional intelligent inspection system and method based on multi-source perception, particularly relates to the technical field of traffic intelligent inspection, and is used for solving the problem of low reliability of road multi-source data fusion. According to the invention, camera shooting, thermal imaging and a vehicle or airborne platform are deployed in a road scene to collect multi-source data; boolean consistency gating and mutual exclusion override are executed in a unified time window, high-credibility candidates are generated through cross validation, cross-modal external parameters are solved by means of road geometric anchor points and a random consistency model, and long-term stable alignment is ensured in cooperation with external parameter change rate threshold rollback; then, necessary and sufficient condition confirmation is carried out on foreign matters, accumulated water, smog and marking defect according to an inspection anomaly truth table, and a finite-state machine is used for tracking evolution and confirming upgrading nodes to schedule vehicle and machine platforms based on the shortest path of a topological graph and link field facilities to form a closed loop of discovery and disposal, so that the efficiency and reliability are remarkably improved, the cost is reduced, and misjudgment is reduced. The method has the advantages of intelligence, globalization and self-adaption.

The invention provides a heat storage heat pump system control method based on a physical information neural network, and belongs to the technical field of heat storage pump system intelligent control. Aiming at the problems that in the prior art, an algorithm is difficult to adapt to dynamic energy consumption requirements, engineering application of a model is difficult due to building space heterogeneity, high-order RC model prediction credibility is weak, engineering feasibility is poor and the like, a solution combining a physical information sequence to sequence neural network technology and a finite-state machine control strategy is provided. On the model level, a 2R2C resistance-capacitance RC model of building temperature change is established, and then a PI-Seq2seq prediction model is proposed based on the physical model. On the control flow optimization level, on the basis of an industrial and commercial time-of-use electricity price policy, an FSM control model is designed, a system state set is defined, parameters and a transfer function are input, and a control rule is constructed in combination with the working period of a building heat pump and the characteristics of a heat storage tank. And finally, energy consumption cost optimization and indoor temperature stabilization under the peak-valley electricity price are realized.

The invention discloses a secure transmission system based on a multimedia short message, and relates to the technical field of communication and network security, and the system comprises a protocol stack security reinforcement module which carries out the hierarchical reconstruction of an MMS protocol stack, comprises a preprocessing layer, is used for intercepting and filtering illegal characters, and verifies the legality of a message structure through a finite-state machine; the analysis layer is used for pre-judging memory requirements based on message types and lengths by adopting a dynamic memory allocation strategy; the execution layer is integrated with a null pointer checking mechanism, and the validity of a pointer is forcibly verified before a protocol stack calls a function; the encryption and signature module is used for carrying out end-to-end encryption on a message body and metadata by adopting a hybrid encryption algorithm and carrying out message integrity verification through a lightweight hash tree; and the vulnerability real-time monitoring module is embedded into an abnormal behavior detection model based on machine learning and is used for dynamically identifying memory occupation abnormity and illegal instruction calling high-risk operation when the protocol stack runs.

The invention discloses a robot image-free visual target docking method and system based on multi-modal perception, and the method comprises the steps: carrying out the recognition calculation of a preset target in the image data of the environment in front of a robot through a target detection model YOLOv4-tiny, and constructing a robot control feedback quantity; radial distance information of an object in the environment is obtained, a scanning range is divided through a regional strategy, straight line fitting calculation is conducted, and a deviation evaluation value of the robot relative to the yaw angle is constructed; and the robot accumulative mileage is obtained, finite-state machine hierarchical decision making and mistaken stop prevention judgment are conducted in combination with the robot control feedback quantity and the robot relative yaw angle deviation evaluation value, and robot graph-free visual target docking is achieved. According to the invention, full-process autonomous control of the robot from target discovery, navigation on the way, obstacle avoidance to end fine alignment can be realized. The robot graph-free visual target docking method and system based on multi-modal perception can be widely applied to the technical field of graph-free navigation.

The invention provides a method for identifying a power grid frequency signal assessment standard, and the method comprises the steps: constructing a system architecture composed of a master station, a substation and a communication network, and achieving the initialization of equipment through a finite-state machine; the method comprises the following steps: collecting power grid frequency, power generation side frequency difference and communication state multi-dimensional data, processing signals by applying median filtering, a sliding window and an EWMA algorithm, introducing fuzzy logic to evaluate a system state, and dynamically adjusting a response strategy; an improved PID algorithm is adopted to track frequency change, feed-forward compensation and PSO parameter optimization are combined, a hidden Markov model is utilized to predict a communication fault, and local mode switching is triggered in advance; main station, mixing and local three-state switching is carried out according to communication states and load changes, and a frequency modulation dead zone is automatically adjusted according to power grid loads through a dynamic threshold value self-adaptive algorithm; the control effect is verified by using a digital twinning technology; and the safety and reliability of the system are guaranteed by optional quantum encryption and active-active redundancy design. The whole scheme effectively reduces the power loss of power generation enterprises, and maintains the safe and stable operation of a power grid.

The invention provides a disease course record generation method and system based on a mobile terminal and multi-modal data, and the method comprises the following steps: firstly constructing a directional collection channel and a ward round session associated with a specific patient through the signal interaction between the mobile terminal and a sickbed; in the session, multi-modal data streams with timestamps, such as voice, images and operation instructions, are obtained in real time, and audio semantic anchor points and visual semantic anchor points are extracted from the multi-modal data streams. And constructing an image-text semantic unit by calculating and associating the audio and video anchor points of which the semantic similarity exceeds a threshold. And taking the unit as a current observation state, and verifying the legality of the state transition through a finite-state machine model. And based on the image-text semantic units passing the verification, template slot filling and term standardization are carried out, and a structured disease course record data packet is generated and written back to the HIS system. The method has the effect of improving the working efficiency and quality of clinical documents.

An end-side-cloud three-in-one active chat robot system for high-emotional merchants comprises an end side, a local edge server and a cloud end, and the end side is used for collecting multi-modal data of a user and performing local lightweight real-time processing and response execution; the local edge server is used for receiving and fusing the multi-modal features and the context information from the end side, and carrying out sentiment calculation, dialogue management and active trigger decision making with medium complexity; the cloud end is used for operating a super-large-scale model and providing global knowledge management, long-term user portrait storage and model training optimization; and the end side, the local edge server and the cloud end carry out cooperative communication through an encrypted channel to form a distributed intelligent processing architecture. According to the method, global optimization is realized by integrating end-side lightweight sensing, edge multi-modal fusion and cloud long-term memory. A composite finite state machine (FSM) active questioning mechanism is combined with sentiment calculation, and is different from traditional rule type triggering. Off-line and on-line fusion scheduling and multi-agent role playing are combined to be applied to a high-emotional-quotient interaction scene, and the simulation is improved. Multi-modal emotion perception circulation is introduced, and the problem of single text emotion misjudgment is solved.

A multi-modulus divider includes a prescaler and a finite state machine. The prescaler includes a combinational logic circuit programmable to divide a frequency of an input clock signal by a first value or a second value. The multi-modulus divider includes a first flip-flop having a data input coupled to an output of the prescaler and an output coupled to an input of the finite state machine. The multi-modulus divider includes a second flip-flop having a data input coupled to the finite state machine such that the data input is configured to receive a programming signal output by the finite state machine and associated with programming the combinational logic circuit. The multi-modulus divider includes a third flip-flop having a data input coupled to an output of the second flip-flop. The third-flip flop also has an output coupled to the combinational logic circuit.

The invention relates to the technical field of mechanical safety monitoring, and discloses a remote monitoring system for instability of a steel structure of hoisting machinery, which comprises the following modules: an inclination monitoring module for collecting inclination angle data of a crane in real time and carrying out signal preprocessing, and a control module for receiving the data and carrying out comparative analysis on the received data and a preset threshold value. And according to an analysis result, a control instruction is sent to the execution module and data is output to the remote transmission module. The execution module, the remote transmission module, the environment compensation module and the data management module are arranged. According to the invention, through multi-module cooperation and algorithm innovation, high-precision monitoring and active prevention and control are realized. A double-axis / three-axis tilt angle sensor is combined with Kalman filtering, the precision reaches + / -0.1 degree, and the reliability is improved through the redundant design of key part equipment. The control module realizes multi-stage alarm through fuzzy PID and a finite-state machine, a Transformer model predicts risks in advance by 10-30 seconds, environment compensation is realized, hybrid networking enhances adaptability, 5G + VR monitoring improves operation and maintenance efficiency, 60% of tipping accidents can be reduced, and multi-scene safety requirements are met.

The invention discloses a unified information extraction method based on Transform, belongs to the technical field of natural language processing, and realizes unification of information extraction tasks through a sequence-to-sequence modeling mode. According to the method, parameter redundancy among the independent models is effectively reduced, and accurate multi-dimensional information extraction of entities, relationships, classification and the like can still be realized in a low-resource scene through the pre-training language model. Meanwhile, the finite-state machine is adopted to restrain the generation process, accurate structuring of the output result under the complex text condition is further ensured, and the overall extraction effect and the flexibility of system application are remarkably improved.

The invention relates to the technical field of network security awareness, in particular to a security awareness method and system for a GPU cluster, and the method comprises the steps: collecting multi-source security situation data, and providing basic information for subsequent security analysis; by constructing and updating a cluster security situation map in real time, information timeliness is updated and ensured in real time, and powerful support is provided for risk assessment; a dynamic risk assessment model is utilized to calculate vertex and edge scores, high-risk vertexes and key risk conduction paths are identified, and a basis is provided for resource isolation and scheduling; the targeted instruction is generated according to the security risk score and the preset security policy library, different security risks are effectively handled, and safe and stable operation of the cluster is guaranteed; by converting the instruction into a specific control command and distributing and executing the specific control command, effective implementation of security measures is ensured, and the cluster security protection capability is improved; by setting a finite-state machine monitoring instruction execution state, a recovery mechanism is automatically triggered, and continuous and safe operation of the cluster is guaranteed.

The invention discloses a table-level data classification and grading method and system based on big data, and relates to the technical field of table-level data classification and grading. According to the scheme, second-level sensitive table grading response is achieved through streaming incremental calculation and dynamic semantic coupling; a finite-state machine and a Hash fingerprint technology accurately capture table structure and content mutation, and traditional batch processing delay is avoided; the multi-modal fusion model synchronously analyzes topological features and semantic vectors, and the field-level sensitivity misjudgment problem is solved; the timeliness weight mechanism guarantees real-time evaluation of a dynamic association table relationship, and is particularly suitable for a high-frequency change scene in a big promotion period; and the asynchronous architecture design isolates the calculation load, so that the processing priority of the core price list is effectively guaranteed.

The invention relates to the technical field of vehicle control, and discloses a parking start-stop battery intelligent module management system, which comprises a data acquisition and preprocessing module, a driving intention inference module, a self-adaptive transient power capability evaluation module and a multi-dimensional fusion decision and control module, discrete driving intention states such as congestion crawling and intersection waiting are recognized through a finite state machine; according to the intention state, adaptively selecting an evaluation sub-model from an evaluation sub-model library so as to evaluate the transient power capability of the battery, and outputting a transient power capability score and a corresponding confidence coefficient; and the system fuses the driving intention, the transient power capability score and the confidence coefficient, corrects logic through the preliminary decision matrix and the confidence coefficient, performs rechecking in combination with the whole vehicle level constraint condition, and generates and issues a final start-stop control instruction. According to the method, the problems of single decision and poor situation adaptability in the prior art are solved, the intelligence and reliability of start-stop decision are remarkably improved, and the driving experience is optimized.

The invention discloses a method for deploying a regular expression based on P4 in an intelligent network card / DPU. The method specifically comprises the following steps that the regular expression defined by a user is compiled into a finite-state machine; a matching table structure and a state register are defined in the P4 program and used for storing a state conversion rule of the regular expression, and a current matching state is recorded; and mapping a matching logic and a state register defined by the P4 into a hardware module of the intelligent network card / DPU, and realizing high-speed state conversion and matching by utilizing the parallel processing capability of the intelligent network card / DPU. According to the method for realizing regular expression matching based on P4 in the intelligent network card or the DPU, the regular expression is converted into the finite-state machine, and the programmability and the hardware acceleration capability of the P4 language are combined, so that the network flow detection and analysis efficiency is greatly improved, and the network flow detection and analysis efficiency is improved. And an effective solution is provided for next-generation network security and performance optimization.

The invention relates to the technical field of IPv6 self-configuration, in particular to an IPv6 network mode self-adaptive configuration method and system based on an RA message. In the method, a finite-state machine establishes a state transition path according to the combination relation of M and O flag bits in an RA message cache set and prefix field content, converts a network mode recognition process into a quantifiable state judgment sequence, and sends the state judgment sequence to the RA message cache set; configuration ambiguity caused by flag bit fluctuation in traditional static judgment is eliminated, pattern recognition under network environment changes has continuity and self-consistency, the classification regression tree constructs a branch judgment path with the duration, the advertisement interval and the prefix consistency of an RA source as characteristic variables, an optimal node is selected through a hierarchical splitting algorithm, and the judgment accuracy is improved. The dynamic mapping of stateful and stateless configuration paths is realized, the misjudgment rate of configuration branches is reduced, and the address allocation delay is optimized, so that the IPv6 self-configuration has a structured decision-making capability and a parameter mapping self-optimization characteristic, and the connection reliability, the address generation speed and the configuration accuracy are improved.

The invention discloses a cigarette cut tobacco batch production circulation control system and method based on single-point condition judgment, and the system comprises a work order parameter verification module which is used for receiving a work order issued by a scheduling system, and carrying out the work order information integrity verification, process parameter compliance verification and equipment suitability verification; the regular expression configuration module is used for providing a visual interface and configuring a regular judgment rule set of equipment starting conditions, ending conditions and abnormal ending conditions; and the single-point judgment execution mechanism is embedded into the PLC control layer, acquires equipment parameters in real time, dynamically evaluates regular conditions based on a finite-state machine model, and triggers an equipment start-stop instruction. According to the system, the cigarette cut tobacco batch switching time is remarkably shortened, and the work order circulation efficiency is improved; the manual intervention frequency is further reduced, and the misoperation rate is further reduced; and the problem of production interruption caused by disconnection of the equipment state and the process parameters is solved.

Techniques for constraining a language model's generation / decoding process using a finite state machine are described. A finite state machine may represent information corresponding to APIs, arguments and argument values that are available / supported by the system. In some embodiments, a language model (LM) is constrained to generate tokens representing valid API calls based on the finite state machine. The system may enable generation of argument values from a defined set or free-form generation of argument values. The system may also enable unconstrained generation of a response by the LM. In some embodiments, a trie data structure is used to determine the possible next tokens that the LM can generate from.

The invention provides a non-deep traffic analysis streaming matching search analysis method, which comprises the following steps of: constructing a deep data packet detection architecture on the basis of a data platform development kit (DPDK); aiming at the non-encrypted traffic, establishing a multi-mode recognition mechanism combining a regular expression and a feature bit stream mode; aiming at the encrypted traffic, constructing an encrypted traffic feature library according to the statistical features, the protocol features and the behavior features; real-time risk detection of network traffic is realized by adopting a streaming matching search technology; and constructing an intelligent decision and response mechanism, and integrating flow analysis and identification results. According to the non-deep traffic analysis streaming matching search analysis method provided by the invention, real-time analysis, risk identification and supervision of network non-encrypted traffic and encrypted traffic are realized by taking a data platform development kit (DPDK) as a basis and combining a high-performance streaming regular expression engine and a finite-state machine principle; the method can be widely applied to scenes of network communication supervision, data security protection, malicious traffic monitoring and the like.

The invention discloses a hanging basket motion control method and system based on operation state recognition, particularly relates to the technical field of bridge construction automation, and is used for solving the problem of poor synchronous control of hanging basket movement. According to the method, working conditions are accurately recognized through multi-source sensing data fusion and a finite-state machine, servo system parameter self-adaptive setting is achieved based on working condition parameter mapping, and an anti-interference movement track is generated by adopting a model prediction control algorithm fusing pose feedback and wind load feedforward; walking synchronization and platform stability are achieved through cross coupling control and posture closed-loop fine adjustment, emergency response is provided through the independent safety guarding thread, control and protection double closed loops are formed, and therefore the motion precision, the anti-interference capacity and the intrinsic safety level of the hanging basket under the variable-load and strong-disturbance working conditions are effectively improved, and the safety of the hanging basket is improved. And the energy consumption and the mechanical vibration frequency in the moving process of the hanging basket are reduced.

The invention relates to the technical field of unmanned aerial vehicle control and management, in particular to an unmanned aerial vehicle parameter consistency configuration method based on template management, which adopts a structured parameter configuration template in a JSON format, comprises four blocks of metadata, parameter definition, constraint rules and digital signatures, establishes an MAVLink communication link with an unmanned aerial vehicle flight controller through a template management terminal, and provides an unmanned aerial vehicle parameter consistency configuration method. And executing an automatic configuration process based on a finite-state machine, and ensuring parameter logic consistency by using a difference comparison and constraint verification mechanism. According to the method, temporary storage, batch writing and atomic submission of parameter changes are achieved through the atomization transaction submission protocol, the integrity of the configuration process is guaranteed, the digital signature auditing log containing the timestamp, the device identifier, the template hash and the change details is generated after configuration is completed, whole-process tracing is supported, and the method is high in practicability and easy to popularize. Headless mode operation, parameter white list control and hardware abstraction layer adaptation are supported, and high automation, standardization and safety of unmanned aerial vehicle parameter configuration can be achieved.

The invention discloses an interpretable end-to-end automatic driving method based on a multi-modal large model, and belongs to the technical field of data processing, and the method comprises the following steps: S1, obtaining a driving environment image, carrying out the feature extraction and coding, and generating a visual representation; s2, generating scene semantic representation by using the semantic answer; s3, determining a state by using a finite state machine; s4, generating a final action according to the visual representation and the scene semantic representation, and constraining the final action through the state of the finite-state machine; and S5, based on the final action, generating a language instruction. According to the method, navigation information and human natural language instructions are encoded into prompts, and large model behavior generation is guided.

The invention belongs to the technical field of data processing. The invention discloses a matrix multiplication calculation task processing system, method and device, a storage medium, a program product and a chip, and the system comprises a storage module which is used for storing a preset weight matrix in advance; the linear transformation control module comprises a finite-state machine used for matrix multiplication operation scheduling and is used for generating an input request signal so that the matrix multiplication calculation task processing system can receive an input matrix; the operation module is used for executing matrix multiplication operation on the input matrix and a preset weight matrix to obtain a matrix multiplication operation result and generate an output matrix; and the data flow control module is used for controlling a plurality of groups of data flows formed by the input data in the input matrix according to rows to sequentially enter the operation module according to the input request signal until all the data in the input matrix execute and complete the matrix multiplication operation. The problems that in the prior art, a matrix multiplication accelerator lacks flexibility in a linear layer implementation task of artificial intelligence hardware, the data migration cost is high, and the energy efficiency ratio is low are solved, and the method is suitable for self-defined extension based on an RISC-V instruction set architecture, can be used as a coprocessor of an RISC-V processor, and can be used as a coprocessor of the RISC-V processor. And linear layer calculation in the neural network is accelerated, especially in an artificial intelligence task, the calculation efficiency can be remarkably improved, the power consumption can be reduced, and the real-time processing requirement can be met.

The invention relates to the technical field of artificial intelligence, in particular to a method, device and equipment for simulating a finite-state machine based on a spiking neural network and a storage medium. The method comprises the following steps: acquiring an input signal; according to the input signal, a corresponding output signal is output through a DTSRNN model which is trained in advance, the DTSRNN model is a model which simulates the state transition process of an FSM by combining the characteristics of the DTSRNN and an SNN, and the output signal is a response signal generated by the DTSRNN model according to the input signal and internal state transition logic. According to the embodiment of the invention, the DTSRNN model combining the characteristics of the DTRNN and the SNN is provided, and the information can be processed in a discrete and sparse mode, which is highly matched with the discrete state conversion process of the FSM, thereby achieving the efficient simulation of the FSM behavior, and greatly improving the performance of a neural network model for simulating the FSM.

A method for converting unstructured data. The method includes receiving a custom-defined data schema that is constructed according to a structured data syntax. The method includes converting the custom-defined data schema into a language model modifier that restricts outputs based on preceding outputs and integrating the language model modifier with an autoregressive machine-learned language model (LLM) to modify output scores of the autoregressive LLM. When receiving a data file that includes unstructured data, the method includes generating a first output from the autoregressive LLM and receiving a set of tokens representing candidates of a second output succeeding the first output. Each token is associated with a score. The method further includes identifying a rule in the language model modifier using the first output, modifying scores of the tokens that violate the rule and selecting one of the tokens as the second output based on the modified scores.

The invention provides an Internet of Vehicles secure communication method and system based on a block chain, and the method comprises the steps: generating a unique vehicle digital gene through fusing a hardware response value and a static identifier of a built-in PUF circuit of a vehicle, and deploying a finite-state machine smart contract capable of managing factory, use, transaction and other states for the vehicle on the block chain. And in a vehicle use state, activating the key pair to perform secure communication, and anchoring the consensus-verified interaction behavior to the contract in a reputation abstract form. And when responding to an ownership transfer request jointly signed by the original vehicle owner, the new vehicle owner and the third party, the contract enters a transaction state and starts a permission decay mechanism to gradually reduce the permission of the original vehicle owner. And after the consensus is passed, completely and synchronously transferring the public key in the contract and the anchored reputation abstract control right to the new vehicle owner through one atomization transaction, and finally updating the contract state to the vehicle use state of the new vehicle owner. According to the invention, full-life-cycle safety management and reputation inheritance of the vehicle are realized.

The invention discloses a book-based subject knowledge system construction method and device, and the method comprises the steps: representing a directory as a directed tree, and dividing the directed tree into a skeleton module and a plurality of chapter modules; carrying out the association and expansion of a title through a large language model, and generating an abstract; knowledge concepts, subject terms and structural relations in titles in all the modules are extracted and subjected to double correction to obtain a plurality of knowledge system trees, and then a mother tree obtained after the skeleton module is processed and a sub-tree obtained by the chapter module are spliced; and finally, giving a plurality of knowledge trees obtained according to a plurality of groups of book directories in the same subject field, and combining the knowledge trees into a unified knowledge system tree based on a finite-state machine principle. According to the method, the knowledge system is constructed and decoupled into an extraction stage and a fusion stage based on LLM, the problem of a pain point of drifting between knowledge concepts and attributes when a simple chapter title is faced is solved, and strict control on semantic consistency and hierarchical rationality in a path merging process is ensured based on a state machine.

The invention provides a constraint method and device for model structured output based on a finite-state machine, and relates to the technical field of network security, and the method comprises the steps: converting a structured constraint rule into a finite-state machine, constructing a shared vocabulary finite-state machine in combination with a model vocabulary, and dynamically limiting an output token range and a field sequence in a reasoning process. It is ensured that model output strictly conforms to preset field definition, value range and grammar rules, the problems of field missing and format disorder in a traditional mode are thoroughly solved, and results can be directly analyzed and utilized by a downstream system. According to the method, low resource occupation, high reusability and quick response are realized while output accuracy and structuring are guaranteed, and key technical support is provided for large-scale application of a large model in security scenes such as attack detection and vulnerability recognition.