33results about How to "Improve generation accuracy" patented technology

A text-to-structured query language conversion method based on a large language model comprises the steps that database outline information is cut, and tables and columns needed by query of a non-target structured query language SQL are eliminated through correlation scoring and Steiner tree problem solving; screening context samples, predicting a similarity score from a text to a structured query language Text-to-SQL by using a cross encoder language model, and selecting a sample most related to a query problem; and injecting the cut database table and column information and the screened context sample into a cue word template, and generating an SQL query statement by using a large language model. The method can be used for optimizing the automatic conversion process from the natural language to the SQL and improving the conversion accuracy, and is used for providing a quick and convenient database retrieval mechanism for a non-SQL user and providing applications such as a data access mode in a wider form for mechanisms such as big language model retrieval enhancement generation RAG and the like.

The invention discloses a mobile game dynamic control optimization method based on adaptive deep learning, and the method comprises the following steps: S1, collecting player operation data and scene background parameters, and generating feature data of a unified scale; s2, generating a self-adaptive spatial-temporal feature sequence through the spatial-temporal feature self-adaptive convolutional network; s3, constructing an adaptive memory network, recognizing a recent high-frequency operation mode through a short-term focusing mechanism, and generating operation intention features; s4, generating a self-supervised learning task through a self-supervised learning multi-layer perceptron, and adaptively adjusting the feature weight to generate a control parameter; s5, a feedback score is generated in real time through a feedback scoring mechanism, and the learning rate is adjusted; s6, establishing a satisfaction evaluation mechanism based on feedback scores and data analysis; and S7, recording operation feedback and scoring results in real time, and realizing adaptive optimization based on feedback circulation. According to the method, mobile game control is dynamically optimized through the adaptive deep learning method, and high adaptability and excellent user experience are achieved.

The invention provides a method and equipment for improving Text-to-SQL (Structured Query Language) conversion accuracy, a medium and a program product. The method comprises the following steps: connecting a target database to obtain DDL (Double Data Language) information of all tables in the target database; extracting effective information of all tables in the database based on the acquired data to establish a table association relation graph, and performing near-synonym expansion on DDL information; when a user inputs a natural language question, an entity contained in the question is extracted and combined with the question to perform RAG retrieval, reachability analysis is performed based on a table association relation graph, and DDL information is used as a context to be transmitted to a large model after a correlation table is complemented, so that the large model generates an SQL (Structured Query Language); performing correctness verification on the generated SQL by using the table association relation graph; if the verification is passed, the generated SQL is returned to the user, and if the verification is not passed, automatic correction is tried, and the corrected SQL is returned to the user. According to the method and the device, the SQL generation accuracy can be improved.

The invention discloses an intelligent regulation and control task chain generation method and apparatus, an electronic device and a medium. The method comprises the steps of obtaining content texts of multiple types of regulation and control security regulations and regulation and control operations; inputting the content text into a large language model agent for content extraction so as to extract regulation and control operation sequences and calling contents under various conditions; intelligently generating a knowledge graph in a decision tree form in the knowledge graph through an intelligent agent according to the regulation and control operation sequence and the calling content under various conditions; acquiring a text instruction, performing intention recognition by an intelligent agent, intelligently generating a task chain for regulating and controlling operation logic in the generated knowledge graph after autonomous understanding, and autonomously generating a thinking chain for task execution; and according to the generated thinking chain, the intelligent agent intelligently calls related modules for regulation and control so as to realize intelligent regulation and control. According to the method, the quality and efficiency of obtaining the text information can be improved, the regulation and control thinking chain logic architecture is established by using the decision tree type knowledge graph, the accuracy is improved, and the use threshold of operation is reduced.

The invention relates to a method for generating a complex-shape three-dimensional building model based on frame field deep learning. The problems that in the prior art, when a three-dimensional building model is generated, training difficulty is large, the three-dimensional building model is easily affected by smooth corners, and artifacts are generated are solved. The method comprises the following steps: S1, acquiring remote sensing data and point cloud data; s2, preprocessing the acquired image data; s3, carrying out data normalization processing; s4, designing a frame field and optimizing details; s5, optimizing the activity skeleton model; and S6, generating a multi-shape contour and a three-dimensional model. The method has the advantages that the generation precision and the generation efficiency of the three-dimensional building model are improved, buildings with complex shapes can be processed, and the method has a wide application prospect.

The invention provides a table-to-text conversion method, system and device and a medium, and the conversion method comprises the steps: obtaining a first dictionary recording key query information, a to-be-converted data table and a first prompt word corresponding to the key query information; the first prompt word is used for indicating table type data corresponding to the key query information in the retrieval data table; inputting the data table and the first prompt into a language model, and obtaining target table type data through the language model; according to the target table type data, performing storage updating on the first dictionary to obtain a second dictionary; and inputting the second dictionary and a second prompt into the language model to obtain a target generated text, the second prompt being used for indicating to retrieve dictionary structure data corresponding to the key query information in the second dictionary. According to the conversion method, computing resources and reasoning time consumed by converting the table into the text can be reduced, and the generation accuracy of converting the table into the text is improved. The invention relates to the technical field of artificial intelligence.

The invention provides a data comparison and reconciliation method and system in a cross-network data exchange and transmission scene, and relates to the technical field of big data. A prepositive data reconciliation system receives service data, generates verification information containing a prepositive verification code, a service data format, a data volume and a unique identification code, and sends the verification information to a server; and after secondary packaging, the data is sent to a rear data reconciliation system through a one-way transmission channel. And the rear system de-encapsulates and generates corresponding verification information, compares the verification information with the received information, and marks a verification state according to a result. If verification fails, an abnormal classification and traceability mechanism is triggered, automatic retransmission or manual intervention is executed, the verification code generation model based on the multi-modal confrontation Hash network can extract multi-modal features, a Hash path is dynamically selected according to information entropy to generate a verification code, the verification accuracy and rate of the multi-modal data are improved, and the verification efficiency is improved. And the abnormal classification covers content tampering, protocol abnormity and the like, and transaction records can be searched in the alliance chain according to the transmission sequence number during tracing, so that the problem positioning and solving capabilities are improved.

The invention relates to a satellite neighbor cell determination method and device, equipment, a storage medium and a program product. The method comprises: receiving interface configuration information sent by an adjacent base station of a target satellite base station, the adjacent base station being a base station within a preset range of the target satellite base station, the interface configuration information comprising non-ground network NTN auxiliary information of a cell of the adjacent base station; determining the distance between the cell of the target satellite base station and the cell of the adjacent base station according to the NTN auxiliary information of the cell of the adjacent base station and the NTN auxiliary information of the cell of the target satellite base station; and according to the distance between the cell of the target satellite base station and the cell of the adjacent base station, adjacent cell relation information of the target satellite base station is determined, and the adjacent cell relation information is used for indicating whether the cell of the adjacent base station is the adjacent cell of the cell of the target satellite base station. By adopting the method, the accuracy of the generated neighbor relation can be improved.

The invention discloses an automatic generation method of a test case and a related device, and relates to the field of software test.The automatic generation method comprises the steps that feature vector extraction is conducted on a target uncovered code and a full-quantity code group through a preset feature extraction model, and a calling link feature vector and a calling feature vector of the target uncovered code are obtained; performing feature fusion on the calling link feature vector and the calling feature vector to obtain a calling information feature vector of the target uncovered code, inputting the calling information feature vector into a preset encoder so as to enable the preset encoder to output a weight code set, inputting the weight code set and a test case template into a preset decoder, and outputting the target uncovered code. And obtaining the test case of the target uncovered code. According to the method and the device, the test case of the target uncovered code is generated by utilizing the preset decoder based on the feature weight code carrying the weight representing the association degree of the calling information and the calling target uncovered code and the test case template, so that automatic generation of the accurate test case is realized.

The invention provides a three-period minimal curved surface generation method and a computer terminal. The generation method comprises the following steps: S1, manufacturing a solid STL model with a required shape; s2, extracting contour coordinate data of the solid STL model; s3, determining a spatial range of the solid STL model; s4, performing grid division on the spatial range, and obtaining coordinate data of each node in a grid; s5, storing each node in the solid STL model as a feature point; s6, calculating an objective function value corresponding to each feature point; and S7, an isosurface is extracted through an isosurface function, and a curved surface is drawn and repaired. According to the method for generating the three-period minimal curved surface, voxel division can be carried out on the solid STL model with the needed shape, gradient change of porosity is achieved through the contour surface extraction algorithm, and the generation precision of the three-period minimal curved surface structure model is improved.

This invention relates to the field of image generation technology, and more particularly to a method for generating bird's-eye view images of reservoir landscapes based on a Stable Diffusion large model. The method includes the following steps: preprocessing acquired reservoir landscape images and semantically annotating them based on water conservancy terminology to construct an image-text pairing dataset; training a pre-trained Stable Diffusion model, incorporating LoRA adaptation parameters into its UNet attention layer, using the image-text pairing dataset to generate a target feature image; and updating the LoRA adaptation parameters based on the difference between the target feature image and the reservoir landscape image to obtain a LoRA model for the reservoir landscape bird's-eye view image. This invention achieves structurally controllable and semantically consistent reservoir landscape generation by fusing structural reference constraints and semantically weighted alignment and introducing LoRA modulation.

The application discloses a museum and art exhibition intelligent operation and visitor experience optimization system and method, relates to the technical field of intelligent management of museums and art exhibitions, and comprises the following steps: obtaining interactive behavior data and identity data of each target object; performing group projection on the interactive behavior data and the identity data, generating group structure data, and performing hierarchical attribution division on the group structure data from an identity dimension, to generate exhibition participation data of each group; and performing scheduling analysis on the correlation degree of the target objects contained in the corresponding group according to the exhibition participation data, to generate behavior-identity deviation data. Through the hierarchical allocation and dynamic reallocation mechanism, the scheme can accurately reflect the resource demand differences of different group levels, and solves the problem of insufficient display strategy adaptability caused by ignoring the behavior differences between group levels in the prior art.

Embodiments of the present disclosure disclose a process page display method and device, electronic equipment and a computer readable medium. A specific embodiment of the method comprises: in response to receiving a process page rendering request, obtaining code information and a set of business processing criteria information; generating initial full business process information; verifying the execution relationship between the initial full business process information, obtaining full business process information; rendering the full business process information and rendering and displaying it on the process page; in response to selecting selection information, generating a matching business batch as a first business batch; and displaying the corresponding business process processing tree on the process page. This embodiment can accurately render the corresponding business process information on the process page after receiving the business data related to the business process generation uploaded to the process page, and can set various page filtering operations for the business process information, greatly improving the rendering efficiency of the business process.

The invention provides a virtual building model generation method and device, equipment, a medium and a program, and relates to the technical field of games. The method comprises the following steps: acquiring at least one group of building structure parameters, wherein the building structure parameters at least comprise a space contour parameter, a structure control parameter and a placement parameter of a virtual connected entity model; generating a space geometry by adopting the space contour parameters; performing structure division on the space geometry by adopting the structure control parameters to obtain a plurality of space regions; generating a virtual segmentation entity model at a target position of a first space region in the plurality of space regions to obtain a basic model of each space region; according to the placement parameters of the virtual connected entity model, the virtual connected entity model is loaded to the basic model of the second space area in the multiple space areas, and a virtual building model corresponding to the building structure parameters is obtained. Therefore, the building model generation precision is improved, and the building model generation scene is expanded.

The invention discloses an energy storage energy management method, device and system and a medium, the energy storage energy management system comprises a local energy management system, a local energy management system and a local energy system, the local energy management system obtains operation data of each device in the local energy system, the model prediction precision of a locally deployed energy prediction model is improved; based on the data type of the operation fusion data, performing data fusion processing on all the operation data to obtain the operation fusion data; calling an energy prediction model to perform energy prediction on the operation fusion data to obtain a strategy prediction result; when the model prediction precision reaches the target precision, obtaining a corresponding local pre-stored strategy according to the operation fusion data, and adjusting the local pre-stored strategy according to a strategy prediction result to obtain a target control strategy; according to the invention, the accuracy of energy management and control can be improved, so that the reliability of an energy storage energy management system is improved.

The present application relates to radar detection technology field, especially to a kind of signal processing method of Doppler frequency offset interference of height-finding radar, the method is by monitoring the maneuverability of height-finding radar and dynamically adjusting frequency modulation period and frequency modulation bandwidth of symmetrical frequency modulation signal;Based on the frequency modulation period and frequency modulation bandwidth after adjustment generate positive slope frequency modulation signal and negative slope frequency modulation signal and carry out periodic alternate emission, collect and process and difference calculation first echo signal and second echo signal after reflection, obtain after eliminating Doppler frequency offset interference ranging frequency difference, further avoid the height-finding error caused by Doppler shift amount, to obtain accurate target height based on ranging frequency difference calculation.This application is by using the Doppler shift polarity opposite characteristics of positive and negative slope frequency modulation signal, by frequency spectrum line pairing and difference calculation offset Doppler shift component, finally based on the target height of corrected frequency information inversion, to realize high-precision height-finding.

The application relates to the field of artificial intelligence and discloses a word and sentence generation method, which comprises the following steps: obtaining a target character string sequence, the target character string being input by a user in an input method tool; generating a target word and sentence corresponding to the target character string sequence through a target neural network according to the target character string sequence, wherein the target neural network comprises an encoder and a decoder, the encoder is used for obtaining an embedding vector according to the target character string sequence, the decoder is used for generating the target word and sentence according to the embedding vector, the target neural network is obtained by training a training sample, and the training sample comprises a character string sequence and a corresponding word and sentence; and presenting the target word and sentence in an interface of the input method tool. The application inputs a character string without error correction and word segmentation into the target neural network, reduces the error superposition influence of a word and sentence precision caused by an error correction model and a word segmentation model, and improves the generation precision of the word and sentence.

The invention discloses an electric power auditing SQL statement generation method based on a large model, relates to the technical field of information, and solves the technical problem that in the prior art, text data is directly generated into SQL statements, so that some text data with irregular formats are generated into wrong SQL statements, and the efficiency of electric power auditing work is low. The method comprises the following steps: identifying and verifying input data; calculating a data normativity coefficient according to the input data; when the data normativity coefficient is greater than a normativity coefficient threshold value, inputting the input data into an SQL generation model to obtain an SQL statement; otherwise, generating a prompt signal, re-inputting the data, interacting through diversified data, and carrying out standardized verification on the input data before generating the SQL statement, so that the statement generation times of nonstandard information are reduced, and meanwhile, fine adjustment is carried out on the large model in the field of electric power auditing, so that the generation accuracy of the SQL statement is improved.

The application relates to the field of image processing, in particular to a spunlace non-woven fabric defect intelligent identification method and system, which comprises the following steps: dividing a standard non-defect spunlace non-woven fabric image into multiple local image blocks according to spatial positions; constructing a local spectrum prior atlas corresponding to the image spatial positions; performing block processing on a to-be-detected image to obtain multiple original image blocks, and dynamically generating a personalized frequency domain mask for each original image block based on the local spectrum prior atlas to suppress the frequency band corresponding to normal textures; constructing a preliminary defect residual graph and a spectrum graph thereof; in response to the proportion of the medium-low frequency region energy in the total frequency domain energy in the spectrum graph exceeding a preset threshold, updating the mask parameters, applying the updated mask parameters to the personalized frequency domain mask, and obtaining a final defect residual graph; performing threshold segmentation and / or connected domain analysis on the final defect residual graph, and outputting a defect position. The application has the effect of improving defect detection accuracy.

The embodiment of the invention provides an audio generation mouth shape expression method of a large model based on emotion reference and discrimination, and relates to the technical field of computers, and the method comprises the steps: obtaining audio data and corresponding text data, emotion category labels and emotion reference videos; extracting audio features from the audio data through the large voice model, and extracting text features from the text data through the large voice model; extracting an emotion feature and an emotion reference feature from the emotion category label and the emotion reference video; performing facial parameter prediction based on the audio features, the text features, the emotion features and the emotion reference features by using a facial parameter generation model to obtain predicted facial parameters; the face parameter generation model is a large model obtained by performing emotion discrimination and content discrimination on the basis of historically predicted face parameters and performing model training according to a discrimination result; and generating a facial mouth shape expression according to the predicted facial parameters. According to the method, the expression generation accuracy in the complex emotional scene can be improved.

The present application relates to the field of image processing, and more particularly to a kind of image defogging method and system.The present application carries out blocking to the image to be defogged, for each block region, the RGB vector of all target pixel points in each block region is fitted, and the initial fitting matrix of each block region is obtained;Based on the inner product of the RGB vector of each pixel point in each block region and the initial fitting matrix of the block region, the effective pixel point of each block region is screened;The effective pixel point of each block region is fitted, and the optimization direction matrix of each block region is obtained;The luminance average value, optimization direction matrix of each block region are combined to obtain the fusion feature matrix of each block region;Based on the fusion feature matrix of each block region, the atmospheric light value of each block region is obtained;Based on the atmospheric light value and the transmissivity of each block region, the image after defogging is generated.The present application effectively improves the precision of defogging image.

The application proposes a large-scale MIMO system channel graph construction method based on a latent space diffusion model (LDM). The method uses the spatial consistency of the large-scale MIMO system channel to learn the conditional probability distribution of the statistical channel state information (sCSI) about the user terminal position to construct the channel graph. An improved variational autoencoder (VAE) is used to learn the low-dimensional representation of high-dimensional sCSI, and the LDM is improved to improve the generation accuracy of the low-dimensional representation of sCSI, and then a channel graph containing position and sCSI low-dimensional representation data is constructed. The storage and transmission cost of the channel graph is greatly reduced compared with directly using high-dimensional sCSI, and the sCSI low-dimensional representation in the channel graph can be accurately restored to high-dimensional sCSI using VAE, and the generation process does not depend on the pilot signal, which greatly saves the pilot overhead while accurately generating sCSI.

The invention provides a video frame insertion method and device, electronic equipment, a storage medium and a product, and the method comprises the steps: predicting the motion blur weight of an insertion frame between a first image frame and a second image frame based on a motion blur prediction algorithm; wherein the first image frame and the second image frame are adjacent image frames in the video to be subjected to frame insertion; generating first pixel information of an insertion frame by using the motion blur weight, the pixel information of the first image frame and the pixel information of the second image frame based on an insertion frame generation algorithm; the motion blur prediction algorithm and the frame insertion generation algorithm are trained and determined by taking the minimum loss function when the first sample image frame and the second sample image frame are subjected to sample insertion frame generation as a target. According to the scheme, when the insertion frame between the adjacent image frames is determined, the adjacent image frames are subjected to motion blur analysis, and the motion path and the motion speed can be captured by the motion blur effect, so that the insertion frame generation accuracy can be improved, the frame insertion effect is ensured, and the video fluency is improved.

The present invention discloses a real-time semantic map generation method and device based on a robot, the method comprising: combining a topological map generation method with deep learning feature extraction, the two are operated in parallel with each other when the robot traverses the environment, and are associated and data fused during feature aggregation, and when the robot completes traversal of the environment, a topological map with semantic information is output in real time. The device comprises: a processor and a memory. The present invention combines two topological map construction methods to generate a new semantic map, and proposes a new image feature aggregation based on the combination of deep learning and topological images, which improves the accuracy of path planning, thereby improving the service quality of the robot.

This invention discloses a multi-factor driven logical sequence generation method and apparatus, comprising: a feature input module, a multi-factor fusion module, a dynamic constraint module, a logical generation module, a reordering and evaluation module, and a feedback learning module. The method involves receiving text information from multiple sources, performing structured parsing and normalization to obtain a standardized set of text feature vectors; extracting multi-factors and constructing multi-factor constraints; generating a multi-factor driven logical sequence based on the set of multi-factor constraints; performing dynamic information updates and output constraint adjustments; generating candidate logical sequences; and designing multiple-dimensional indicators for re-evaluation. This invention significantly improves the accuracy, adaptability, and optimality of logical text sequences by addressing the problems of insufficient input information processing, lack of dynamic adaptability, and insufficient post-optimization mechanisms in the logical text sequence generation process.

The application discloses a kind of steel rail defect three-dimensional generation reconstruction methods based on ultrasonic sound field guidance, the B scanning data of steel rail is converted into defect cross section drawing by the cyclic generation of the adversarial network of ultrasonic sound field guidance, then defect cross section aperture and distance and other information are extracted from defect cross section drawing, then the transverse size of defect is calculated in combination with the amplitude of B scanning data, and the position of steel rail defect is converted from B scanning coordinate system to steel rail coordinate system by three-dimensional space transformation, to obtain the spatial position of steel rail defect in steel rail;Finally, the aperture of defect, transverse size and spatial position are transmitted into ultrasonic detection three-dimensional platform, to output the three-dimensional model of steel rail defect, complete the three-dimensional reconstruction of steel rail defect.

The application discloses a four-dimensional flow field magnetic resonance data synthesis method and system based on CT images, and is applied to the technical field of medical images, and the method comprises the steps of: acquiring a CTA image; performing image segmentation and three-dimensional reconstruction processing on the CTA image to obtain a target three-dimensional model of a thoracic aorta system; performing fluid dynamics simulation processing based on the target three-dimensional model to obtain blood flow field simulation parameters of the thoracic aorta system; and performing four-dimensional flow field magnetic resonance data synthesis processing based on the blood flow field simulation parameters to obtain magnetic resonance synthesis data corresponding to the CTA image. The application can generate 4D Flow MRI data from a CT image, and effectively improves the generation accuracy of the 4D Flow MRI data.

The application discloses a network security threat event report generation method and system, and belongs to the network security field.The method comprises the following steps: acquiring a network security threat event log, pre-processing the log, extracting key information by using a pre-trained ERNIE-Gram model, and generating structured knowledge points; integrating the knowledge points into a feature vector, inputting the feature vector into a KAN model for nonlinear feature extraction and multi-level feature fusion, and generating fusion feature data; inputting the fusion feature data into an OverLoCK model, performing multi-stage feature extraction and aggregation through a dynamic context guiding mechanism of the model, and generating summary semantic analysis information; standardizing the semantic analysis information, inputting the semantic analysis information into a pre-trained federal decision tree model, matching a decision path of the federal decision tree model according to a global split rule, and outputting an interpretable security event report containing a confidence distribution.The application can improve the accuracy and interpretability of threat event report generation.

The invention discloses an intersection reference line generation method, a medium, equipment and a vehicle, and the method comprises the steps: mining self-vehicle passing data fragments related to an intersection scene in full-scene data to form a self-vehicle passing data fragment set, and carrying out the defect data filtering and abnormal data filtering of the self-vehicle passing data fragment set, the full-scene data is recorded real driving data; taking scene data in the self-vehicle passing data fragment set before the self-vehicle passes through the intersection as training data, taking a real track in the self-vehicle passing data fragment set when the self-vehicle passes through the intersection as a training label, and constructing a training set according to the training data and the training label; training a preset intersection reference line generation model by using the training set, wherein the intersection reference line generation model is created based on a deep learning model; and deploying the trained intersection reference line generation model in an automatic driving system to generate an intersection reference line. In this way, the intersection reference line can be accurately generated in various driving environments.