187results about How to "Diversity guaranteed" patented technology

The invention relates to the technical field of generalized quantum evidence theory and image processing, in particular to a ship image data target identification method based on the generalized quantum evidence theory, which comprises the following steps: converting synthetic aperture radar ship image data into complex field synthetic aperture radar data, and constructing training data; generating a composite quantum evidence label according to the potential structure; training the quantum evidence neural network to obtain a standard quantum evidence neural network; inputting to-be-recognized complex field synthetic aperture radar data into the standard quantum evidence neural network, and extracting a complex field feature vector of a to-be-recognized sample; normalizing the complex field feature vector into a generalized quantum basic probability amplitude function, and defining a proposition space under a quantum identification framework; and calculating generalized quantum basic probability distribution of the proposition in the proposition space, and executing double judgment to obtain a recognition result of the original synthetic aperture radar ship image data to be recognized. According to the method, the accuracy of ship image data target identification based on the generalized quantum evidence theory can be improved.

The invention provides a cross-border trade supply chain risk early warning system based on digital twinning, and relates to the technical field of cross-border trade. The cross-border trade supply chain risk early warning system based on digital twinning comprises an entity layer, a data acquisition layer, a digital twinning modeling layer, a data processing layer, a risk early warning layer, a safety protection layer and an application output layer, and the entity layer is a monitoring object basis of the system and covers all key physical objects in a cross-border trade supply chain; comprising suppliers, logistics carriers, storage facilities, customs nodes and purchasers in the cross-border trade supply chain. The system integrates a multi-source data acquisition module, covers full-dimensional data of policies, logistics, finance, markets, Internet of Things perception and the like, ensures the comprehensiveness of risk factor extraction, improves the coverage of risk identification, greatly improves the accuracy and real-time performance of early warning, and solves the problem that a traditional early warning mode is inaccurate and lagged in evaluation.

The invention provides a cooperative path planning method for searching and tracking an underwater target by a multi-USV unmanned cluster, and the method comprises the steps: introducing a pose-control quantity-time to carry out the three-dimensional double-chain coding of a population in an improved genetic algorithm, enabling a control instruction and an accurate execution timestamp to serve as a genetic gene, and enabling the control instruction and the precise execution timestamp to serve as a genetic gene in the crossover and mutation operation, timestamps, control quantities and poses are transmitted synchronously, and it is ensured that offspring individuals can inherit an excellent cooperation mode of a parent; during population initialization, USV is guided to preferentially cover a high-value region through region division and weighted Gaussian distribution, so that blindness caused by random initialization of a traditional genetic algorithm is overcome, and invalid search is avoided; a target for avoiding secondary search is designed in a fitness function, and repeated access to recently searched areas is quantified by introducing a concept of confidence time intervals, so that an algorithm actively explores unexplored water areas, waste of search resources is avoided, and accuracy and planning efficiency of multi-USV collaborative path planning are improved.

The invention discloses a high-precision coordinate calculation method based on Matlab, and belongs to the field of coordinate calculation technology and data processing, and the method comprises the steps: firstly, obtaining the combination of all points and vectors through a specific combination generation function, and then initializing a result vector; aiming at each combination, calculating intersection point coordinates and parameters by virtue of an accurate geometric intersection point calculation function, storing the intersection point coordinates and parameters in a result vector, and distributing the intersection point coordinates and parameters to a working area according to a rule by virtue of a working area assignment function; then, a midpoint coordinate matrix is obtained by applying a midpoint calculation function based on the coordinates of the intersection points, the distance between the points is calculated, a clustering center is determined by adopting a clustering algorithm, points of an intersection point concentrated area are screened out according to distance characteristics in a cluster, and finally, high-precision coordinates are calculated through a weighted average calculation function in combination with priori knowledge and weights. According to the method, the coordinate calculation precision can be effectively improved, good noise immunity is achieved, priori knowledge can be fully fused, and the method is suitable for numerous fields with high coordinate precision requirements.

This invention discloses a target recognition method, docking system, and docking method in a space environment. By adjusting the output intensity of the sunlight simulator's light source and the position and angle of the simulated work equipment, different types of simulated target images are obtained, forming a target photo library. The simulated target images are processed into small target images and then placed into a small target image library. This small target image library is merged with the target photo library augmented by GAN data, and divided into training and testing libraries proportionally. An SA-YOLOX network is constructed. The SA-YOLOX network is trained using the training library to obtain an SA-YOLOX target detection model. The augmented target photo library is tested using the SA-YOLOX target detection model and the testing library to obtain test results. Based on the test results and model metrics, an evaluation result is output. This invention effectively overcomes the performance index problem of the YOLO series algorithms in solving the special situation of high-contrast image target recognition in complex space environments.

The application discloses a kind of collaborative robot job parameter dynamic optimization method, comprising: real-time acquisition of the multimodal sensor data of collaborative robot in the process of job, multiple job parameters are analyzed based on job quality evaluation model to identify key parameter set from it;Multi-objective function is constructed with job time, end force as optimization target, and multi-objective function is converted into single-target optimization function by weighting sum method;Key parameter set is input to machine learning model, and single-target optimization function is used as optimization target to carry out iterative optimization, in each iteration, the standard deviation σ of the current learning result of machine learning model is obtained, according to the comparison result of standard deviation σ and preset critical value s, dynamically select optimization strategy, when σ≥s, new parameter set is generated using random exploration strategy, when σ<s, new parameter set is generated using guided optimization strategy of machine learning model, and the optimized parameter set is output.

The invention relates to the technical field of computers, in particular to a code review method and device, electronic equipment and a storage medium, which are used for efficiently screening out high-quality and high-diversity training data so as to improve the accuracy of code review. The method comprises the following steps: acquiring a training sample set and a test sample set, and dividing the training sample set into a plurality of training sample subsets; the test sample set comprises test samples reaching expected quality; after a to-be-trained code review model is preliminarily trained based on the training sample set, the test influence degree of each training sample subset on the code review model on the test sample set is determined; for at least one target training sample subset of which the test influence degree meets a preset influence condition in the plurality of training sample subsets, performing clustering according to a sample gradient to obtain a plurality of clusters; and sampling is carried out in each cluster, and the code review model after preliminary training is trained again according to the obtained sampling sample set.

The present application relates to the field of intelligent traffic parking scheduling, and particularly relates to a parking resource comprehensive scheduling strategy optimization method and system based on multi-region cooperation, which specifically as follows: collecting multi-region and multi-type parking resources, commuting parking demand, illegal parking risk and road network operation data according to scheduling period, constructing parking cooperative monitoring sample through standardization processing; calculating time-sharing sharing potential index, underground priority access intensity and road overflow risk elasticity index, constructing hierarchical directed ant colony search graph with heuristic information written; introducing capacity constraint self-repairing mechanism and congestion degree pheromone updating mechanism, solving Pareto solution set through multi-objective ant colony optimization, screening optimal compromise scheme combining fuzzy membership and dynamic preference weight, outputting parking guidance instruction and realizing rolling closed-loop iteration. The present application can realize time-sharing cooperative deployment of parking resources in residential and office areas, balance various parking load, and improve overall utilization efficiency of parking resources.

The invention provides a data processing method and device for instrument detection, equipment and a storage medium, relates to the technical field of digital instrument detection, and solves the problem of insufficient detection precision caused by the technical bottlenecks of data scarcity and difficulty in small target recognition in a power distribution room scene in the prior art. According to the scheme, a large amount of sample data can be automatically generated based on a programmed image generation mechanism, and the target detection model introduced with the convolution attention mechanism is trained, so that a more accurate model is deployed on detection equipment; therefore, high-precision and high-robustness detection of the small-size and low-contrast digital instrument in the to-be-detected place under the complex environment of the power distribution room is realized, and the detection precision of the characters in the digital instrument is effectively improved.

The invention relates to the technical field of disaster early warning processing, in particular to a mountain torrent disaster early warning system based on deep learning and air-space-ground integrated perception. The system comprises the following modules: a space-air-ground sensing module, an edge calculation preprocessing module, a deep learning prediction module and an intelligent decision and early warning module. Wherein the air-space-ground sensing module is used for collecting multi-source data; wherein the multi-source data comprises earth surface deformation data obtained through a spaceborne synthetic aperture radar, high-precision topographic data obtained through an unmanned aerial vehicle laser radar, hydrological sensing data obtained through a ground internet-of-things distributed sensor, and meteorological numerical forecasting data obtained through a meteorological data interface. Through air-space-ground integrated perception, physical constraint deep learning and edge intelligence technologies, mountain torrent multi-source data collaborative acquisition and accurate mountain torrent disaster prediction are realized, so that the comprehensiveness, reliability and timeliness of mountain torrent disaster early warning are improved.

The application discloses a kind of big data processing's augmented reality hospital pharmaceutical scheduling management method and system, it is related to medical information processing technical field, its technical solution key points include the following steps: obtaining the AR space positioning data of hospital pharmaceuticals, real-time inventory data, demand data, distribution path data form scheduling core data set, in scheduling core data set Filtering out the data that deviates from standard scheduling threshold value forms to be optimized scheduling data;Filtering scheduling deviation type, scheduling deviation amplitude, scheduling deviation node in to-be-optimized scheduling data, whether need to be optimized scheduling data is judged according to scheduling deviation type and scheduling deviation amplitude Whether collaborative calibration processing is needed, mark the data processed as target scheduling data;Extract scheduling parameter information of pharmaceutical in target scheduling data;Effect is to realize the dynamic accurate update of scheduling data, avoid the risk of inventory backlog, shortage caused by data deviation, guarantee the comprehensiveness and accuracy of scheduling decision.

The embodiment of the invention relates to the technical field of federated learning and information security, in particular to a federated multi-target scheduling method and system for industrial control crowdsourcing testing, which is characterized in that sensitive data such as ability, experience and the like of a tester are only reserved at a local terminal by constructing a framework combining local feature coding and federated model training, so that the test efficiency is improved, and the test efficiency is improved. Outward transmission is not carried out; only the model update quantity after the privacy enhancement processing such as gradient cutting and differential privacy noise injection is uploaded to the federated matching learning server, and only the model update quantity after the privacy enhancement processing is uploaded to the federated learning server, so that the leakage risk of concentrated data storage is avoided from the source, the requirements of privacy regulations such as GDPR and the like are completely met, and the reliability of the system is improved. Compared with traditional privacy technologies such as homomorphic encryption, the method has the advantages that the calculation and communication overhead is greatly reduced, the design of local matching degree calculation is matched, the task matching response speed is greatly increased, and the real-time requirement of industrial control system testing is met.

The invention provides a simulation method and device for positioning an easy-to-wear part of an operating mechanism, equipment and a medium, and effectively solves the problems of various limitations of a traditional positioning method for the easy-to-wear part of the operating mechanism in accuracy, comprehensiveness, capture of a dynamic rule of wear evolution and the like. The method includes; constructing an initial three-dimensional model based on the operating mechanism; performing multi-dimensional configuration on the initial three-dimensional model to obtain a target three-dimensional model, observing a dynamic response generated by the target three-dimensional model during simulation, and outputting action curves and speed curves of a plurality of components; acquiring various contact parameter data of a plurality of contact parts generated when various parts in the plurality of parts are in mutual contact from the action curve and the speed curve, and calculating the various parameter data to obtain abrasion loss and mechanical variables of the plurality of contact parts; and screening out a target contact part from the plurality of contact parts based on the abrasion loss and the mechanical variable so as to position an easy-to-wear part of the operating mechanism based on the target contact.

The application provides a delay test system and method based on a Hall switch device, comprising: a magnetic field generation module containing a programmable current source and a coil to generate a precisely controlled magnetic field change; a temperature control module integrating a temperature sensor and a heating / cooling device to realize constant temperature or temperature scanning test; a signal acquisition module using a high-frequency sampling device to collect Hall device output signals; a data processing module using an embedded processor to realize automatic threshold detection and delay calculation; a general test interface using a test seat with replaceable probes to adapt to different packaged devices; a digital communication interface using an SPI or I2C interface to output test results; in the application, the main structures include a magnetic field generation module, a temperature control module, a signal acquisition module, a data processing module, a general test interface, a digital communication interface and a threshold calibration module, and these modules cooperate with each other to provide a solid foundation for precisely testing the delay characteristics of the Hall switch device.

The application discloses a power transmission line crimping pipe X-ray detection device, system and method, which comprises the following steps: driving a detection module to reach a detection position through a lifting mechanism; driving the detection module to cover the crimping pipe detection range on the left and right sides in the same plane through a swing mechanism; or driving a first receiving plate and a second receiving plate to open through an opening and closing mechanism, and making the first receiving plate and the second receiving plate cover the crimping pipe detection range; and adjusting the angle of an emitter through a swing mechanism to make the emitter align with the first receiving plate or the second receiving plate. The detection position can be adaptively adjusted, the comprehensiveness and accuracy of detection are ensured, the traditional artificial tower climbing detection operation is replaced, and the intelligent detection operation level of the power transmission line crimping pipe is improved.

The present application relates to the field of visual detection technology of optical module industry, and provides a method and device applied to AOI detection. A low-power microscopic imaging module of a fixed shot visual subsystem is controlled to acquire a global image of a detection object on an object platform, and each flying shot point coordinate is located from the global image; according to the each flying shot point coordinate, a motion shaft is controlled to drive the flying shot visual subsystem to move, and a plurality of local high-definition images are acquired by performing flying shot on each flying shot point; the acquired plurality of local high-definition images are spliced into a complete to-be-detected image with the global image as a background; each regional image is obtained by performing regional cutting on the to-be-detected image according to the category of the to-be-detected object; and the defect category of each regional image is determined according to a defect feature, and a defect detection result is generated. The present application solves the problem that large-area defects are not fully extracted and are incomplete, and ensures the comprehensiveness and accuracy of detection.

The invention relates to the technical field of instant food production, and discloses a making process of a convenient mushroom hotpot, which comprises the following steps of: firstly, selecting basic mushrooms of which the first type is shiitake mushroom, coriaria sinica, pleurotus citrinopileatus, black fungus, pleurotus cornucopiae, velvet antler mushroom, cordyceps flower and hericium erinaceus, and the second type is tremella aurantialba and bamboo fungus; cleaning and removing impurities, adding water into the first type of mushrooms, heating at 100-120 DEG C for 30 minutes, adding the second type of mushrooms, and continuously heating for 20 minutes to obtain a mixed system containing cooked mushrooms and mushroom stock; and after solid-liquid separation, adding seasonings into the mushroom stock soup, heating to prepare a mushroom soup bag, sealing the cooked mushrooms to form a mushroom bag, and packaging the mushroom soup bag and the mushroom bag together to obtain the product. According to the scheme, the problems that basic mushroom combination of an existing convenient mushroom hotpot is not clear, the taste and the umami are insufficient due to the fact that heating is not conducted according to characteristics differentially, and the flavor is poor due to the fact that mushroom soup is not original soup are solved, and the product is convenient to eat and stable in quality.

The application discloses a fistula risk assessment system and method based on a flora gene database, and belongs to the technical field of data evaluation, and comprises: a specimen collection and sequencing module: used for collecting fecal specimens of high-complexity fistula subjects and healthy subjects, extracting intestinal flora DNA from the specimens, performing PCR amplification and macrogene sequencing, and obtaining high-quality clean macrogene sequencing data; a database construction module: used for receiving high-quality sequencing data output by the specimen collection and sequencing module, sequentially performing sequencing data preprocessing, Alpha and Beta diversity analysis, PCA / PCoA dimensionality reduction analysis and intergroup difference flora screening, and constructing a high-complexity fistula specific flora gene database; and a model training module: used for constructing and training a high-complexity fistula risk assessment model of a deep learning network based on the flora gene database. The application can objectively quantify the risk of disease, and the result is stable and has strong generalization.

The present application relates to zero-carbon park management technical field, specifically relates to a kind of multi-source carbon data intelligent management and control method and system for zero-carbon park.The technical scheme includes the following steps: constructing multidimensional perception network, real-time acquisition multi-source carbon related data;And it is preprocessed, obtains carbon data set.Combined with emission source type, obtain park carbon emission evaluation data.Through park carbon emission evaluation data, construct multidimensional zero-carbon park carbon emission evaluation index system, calculate the comprehensive weight and comprehensive evaluation score of each evaluation index.Judge whether the comprehensive evaluation score meets the qualified standard, and combined with park zero-carbon development goal and park carbon emission evaluation data, construct intelligent emission reduction decision model and generate emission reduction scheme.Through data preprocessing algorithm, eliminate abnormal data, fusion multi-source data, improve data reliability;Realize high-precision carbon emission calculation, and calculation error can be controlled in very small range;Effectively improve park zero-carbon transformation efficiency.

The present application relates to the technical field of flexible job shop scheduling in intelligent manufacturing and production scheduling, and particularly relates to an optimization method for solving flexible job shop scheduling with adjusted resource constraints. The method comprises the following steps: initializing parameters and randomly generating an initial population; using a crossover operator and a mutation operator in sequence to evolve the current population; executing a hybrid decoding strategy on the current population; sorting the individuals in the current population in ascending order of maximum completion time to form an elite population, and updating the elite population through problem-specific local search; judging whether the evolution condition is met, and if yes, executing mathematical evolution based on CP, and outputting a final solution when the running time reaches the total running time. The present application has the positive effects of reducing resource waiting time, improving machine utilization, and improving the resource utilization efficiency and scheduling performance of the entire workshop production.

The invention provides a microorganism-drug association prediction method based on multi-scale adaptive graph representation learning. The method comprises the following steps: constructing a microorganism-drug heterograph comprehensive feature matrix; processing the matrix through a spectrum kernel attention mechanism, generating a spectrum kernel feature matrix in a spectrum space by using a Chebyshev polynomial, and explicitly fusing graph topology information in combination with a multi-head attention mechanism to capture a global topology dependency relationship and obtain global feature representation; the matrix is processed based on a structure perception adaptive neighborhood sampling mechanism, a neighborhood sampling strategy is dynamically predicted and adjusted according to structural features such as degrees and centrality of nodes, local neighborhood information is obtained in a mixed sampling mode, and local feature representation is obtained; and fusing global and local feature representations, and calculating a potential association score between the microorganisms and the drugs through a prediction module. According to the method, the global structure dependence and the local structure heterogeneity of the graph can be effectively combined, and the accuracy and the robustness of microorganism-drug association prediction are remarkably improved.

This invention belongs to the technical field of bearing production equipment, and particularly relates to a fully automatic bearing retainer laser cladding equipment. It includes a worktable, with two sets of feeding and discharging conveyor belts arranged side-by-side on the outer side of the worktable above the protective housing. These belts are used for receiving workpieces input inward and output outward. A transverse feeding mechanism is installed on the worktable above the discharging conveyor belt. A linkage drive assembly is provided between the transverse and longitudinal feeding mechanisms. A tilting mechanism is installed on the worktable below the laser cladding assembly, and a clamping mechanism is installed below the worktable corresponding to the auxiliary feeding mechanism. This invention is rationally designed, simple in structure, and easy to process. It integrates automatic feeding, laser cladding, and unloading functions, improving the linkage between equipment, ensuring smooth production, reducing labor intensity, guaranteeing processing progress, and meeting usage requirements.

The application relates to the technical field of artificial intelligence, and discloses a highway side slope deformation intelligent identification and prediction system based on space-air-ground stereoscopic monitoring, which comprises a space-air-ground stereoscopic monitoring module, a data fusion processing module, an intelligent identification module, a deformation prediction module and a central control module; the space-air-ground stereoscopic monitoring module cooperatively collects side slope deformation data by using multi-source heterogeneous sensors, including space satellite remote sensing data, sky unmanned aerial vehicle aerial photography data and ground Internet of Things sensor data, realizes all-weather and multi-scale monitoring of side slope surface displacement, internal stress, crack expansion and environmental factors, guarantees the comprehensiveness and real-time performance of the monitoring data, and eliminates the limitations of a single data source; the data fusion processing module performs space-time registration, denoising and normalization processing on the multi-source monitoring data, guarantees the accuracy and reliability of real-time deformation risk diagnosis, and reduces the false alarm rate.

The invention discloses a self-adaptive stress release workbench for restraining deformation of a laser cutting plate, and relates to the field of laser cutting machining, the self-adaptive stress release workbench comprises a rack, a plate clamping frame, a workpiece receiving assembly and a self-adaptive supporting assembly, the plate clamping frame is slidably connected to the rack through a first rail, and the workpiece receiving assembly is slidably connected to the rack through a second rail; the workpiece receiving assembly is slidably connected to the middle of the rack through a second rail and located below the plate clamping frame, the self-adaptive supporting assembly is arranged on the rack in a liftable mode, and the working end of the self-adaptive supporting assembly makes contact with the bottom of a plate clamped and fixed to the plate clamping frame. The working end of the self-adaptive supporting assembly can respond to the stress change in the plate in the cutting process of the plate to generate multidirectional micro displacement, so that dynamic supporting and stress reduction of the plate are achieved. In the cutting process, the internal stress of the plate can be responded and reduced in real time through the dynamic supporting system, and meanwhile precise cooperative control over the supporting assembly and the workpiece receiving system is achieved.

The invention relates to the technical field of adsorbent performance optimization, in particular to a CO2 humidifying adsorbent adsorption performance optimization method and system based on multi-feature matching, and the method comprises the steps: constructing a synthetic functional group set, selecting adsorbents based on the synthetic functional group set to obtain a pre-selected adsorbent set, sequentially extracting pre-selected adsorbents from the pre-selected adsorbent set, and extracting the pre-selected adsorbents from the pre-selected adsorbent set; the method comprises the following steps: extracting a pre-selected adsorbent, determining a pre-selected adsorption characteristic parameter group of the extracted pre-selected adsorbent, constructing a simulation model according to the pre-selected adsorption characteristic parameter group to obtain a simulation adsorbent, performing adsorption condition optimization on the extracted pre-selected adsorbent based on the simulation adsorbent to obtain target adsorption data, and summarizing the target adsorption data to obtain a target adsorption data set. According to the method, the efficiency of optimizing the adsorption performance of the CO2 adsorbent can be improved, and the dependence on experimental resources in the optimization process is reduced.

This invention relates to the field of radio frequency communication technology and discloses a communication frequency selection method and apparatus based on RSSI. The method includes: sweeping a target frequency band using step units to obtain a set of sweep points; performing peak detection on the sweep point set to obtain the peak received signal strength and interference frequencies; calculating at least one third-order intermodulation point within the target frequency band based on each interference frequency; fitting the sweep point set, each interference frequency, and each third-order intermodulation point to obtain a target sweep point set; performing weighted calculations on each target sweep point in the target sweep point set based on antenna weighting coefficients and the center frequency to obtain a received signal strength weighted value; and selecting the optimal frequency point based on each received signal strength weighted value. Therefore, implementing this invention can improve the accuracy and reliability of communication frequency selection, extend communication distance, and ensure the comprehensiveness and accuracy of frequency selection, thereby improving the stability and anti-interference capability of wireless audio communication.

The present invention provides a polishing pad, by a design of a subdivided structure characteristic in a thickness direction, can provide a physical characteristic suitable for various polishing purposes to various polishing objects, and, unlike the existing polishing pad, in at least a part of the structure, a regenerative or recyclable material is applied, so that environmental friendliness can be achieved. Specifically, the polishing pad includes a polishing layer including a polishing variable layer having a polishing surface, and a polishing invariable layer disposed on the back surface of the polishing surface of the polishing variable layer, the polishing invariable layer can include a cured product of a composition including thermosetting polyurethane particles and a binder.

This invention belongs to the field of integrated circuit equipment technology and discloses a high-damping vibration-damping base for integrated circuits, including a lower base. The integrated circuit equipment is placed inside the upper base and is centered and clamped by four first flexible square rings in the clamping assembly. When the integrated circuit equipment is used and subjected to vibration during operation, the upper base vibrates. The force on the upper base is transmitted to the second block, the second spherical rod, the elastic damper, the first spherical rod, and the first block, causing the elastic damper to extend and retract, thus mitigating the impact force generated during vibration. Since the multiple second spherical rods and the first spherical rods can move in a circular motion around the second block and the first block, when the integrated circuit equipment vibrates irregularly, the impact force generated in multiple directions of movement of the integrated circuit equipment can be mitigated, reducing the vibration frequency of the upper base relative to the lower base.

The application provides an oversampling method, system, controller and storage medium for an unbalanced data set, wherein the oversampling method comprises: mapping a data set to a feature space by using a pre-trained deep learning model, wherein the data set comprises multi-class sample data, and the multi-class sample data is divided into majority class sample data and minority class sample data according to data quantity; optimizing the feature space to enable same-class sample data in the multi-class sample data to be gathered; generating synthetic samples of the minority class sample in the feature space; and verifying the synthetic samples, and supplementing the at least minority class sample with the verified synthetic samples. The application solves the problems of existing oversampling methods, such as easy model overfitting, sample confusion, poor sample quality and the like.

The present disclosure relates to a layered parallel inference acceleration system, method and computer program product. The system comprises a text content analysis module for analyzing and processing user input text content to obtain a vector sequence; a candidate token sequence generation module for determining a corresponding set of to-be-predicted positions according to the current position of each of the plurality of vectors, and determining the feature information corresponding to each to-be-predicted position in the set of to-be-predicted positions; and predicting at least one candidate token according to the feature information, and combining each corresponding at least one candidate token in the set of to-be-predicted positions to generate a plurality of candidate token sequences; and a token sequence verification module for screening a candidate token sequence with the most number of candidate tokens that are continuously verified from the start position from the plurality of candidate token sequences as a target token sequence. In this way, the diversity and quality of the generated content can be ensured while reducing the model inference delay.