15results about How to "Resolve ambiguity" patented technology

The application discloses a multi-modal-based digital human interaction control method and device, and relates to the technical field of artificial intelligence digital human, and specifically comprises the following steps: collecting multi-source heterogeneous multi-modal data of user interaction with a digital human in real time; constructing a dynamic adaptive multi-modal space-time hypergraph Transformer network, deeply fusing the multi-source heterogeneous multi-modal data, and obtaining a fused interaction feature vector; generating a multi-dimensional interaction situation tensor based on the fused interaction feature vector; generating a digital human response action sequence, voice content, expression posture and interaction strategy according to the multi-dimensional interaction situation tensor, and driving the digital human to execute; acquiring feedback data of the digital human after executing the response in real time, and optimizing parameters of the dynamic adaptive multi-modal space-time hypergraph Transformer network based on the feedback data; and effectively improving the naturalness and response accuracy of digital human interaction, and reducing the interaction misunderstanding rate and user cognitive load.

This invention discloses a Monte Carlo-based mineral assemblage analysis method, belonging to the field of mineral analysis. The invention includes the following steps: multi-source data fusion and preprocessing; physical model construction and initial solution generation; multi-scale feature fusion and extraction; Monte Carlo random sampling and solution space exploration; physical-data dual-driven evaluation and screening; iterative update and model optimization; convergence judgment and result output. This invention effectively solves the problem of multiple solutions in mineral assemblage identification by combining multi-source data fusion, a physical-data dual-driven model, and the Monte Carlo method with multi-scale feature fusion and iterative optimization. This method can integrate geophysical, geochemical, remote sensing, borehole core, and mineral spectral data, and through iterative model updates, gradually converges to the most probable mineral assemblage, providing uncertainty assessment, thereby improving the accuracy and reliability of mineral identification.

This invention relates to the field of petroleum engineering drilling technology, and discloses a method for identifying the pore structure of carbonate rocks based on multimodal deep learning. First, the acquired one-dimensional acoustic waveform data and porosity parameters are normalized, noise-enhanced, and subjected to continuous wavelet transform to obtain two types of inputs: a one-dimensional acoustic waveform curve and a two-dimensional time-frequency graph. Then, a one-dimensional convolutional neural network is used to extract the local phase and amplitude features in the time domain of the acoustic waveform, and the global energy distribution features in the two-dimensional time-frequency graph are extracted. The porosity parameters are embedded as physical constraint features. Finally, the three types of features are concatenated to form a joint feature vector. A classification branch outputs the probabilities of three types of pore structures: fractures, intergranular pores, and caverns, while a regression branch outputs continuous values ​​of the pore aspect ratio. This invention establishes the relationship between acoustic features and pore structure types through multimodal fusion of one-dimensional acoustic waveform curves, two-dimensional time-frequency graphs, and porosity physical parameters, thus realizing the identification of complex pore structures in carbonate rocks.

The invention discloses a self-calibration surface normal field reconstruction method and device based on asynchronous event streams. The method comprises the following steps: controlling an illumination unit to actively scan to generate a time-varying illumination field, and acquiring an event stream through a monocular asynchronous vision sensor; taking the ratio of the ambient light to the surface albedo as a fourth component to construct a joint state vector containing a three-dimensional surface normal vector, deducing the orthogonal constraint of the joint state vector and a four-dimensional differential observation vector, and realizing mathematical decoupling of ambient light interference; representing a time-varying illumination track as weighted superposition of discrete anchor points by adopting a parameterized basis function with a local support characteristic, and establishing a corresponding relation between an asynchronous event and global illumination; eliminating generalized bas-relief transformation fuzziness by utilizing rigid constraint provided by preset spatial geometric configuration of the lighting unit; and the surface normal field and illumination parameters are solved at the same time through global joint iterative optimization. According to the method, precise illumination calibration is not needed, and low-data-bandwidth and high-dynamic-range surface reconstruction can be achieved under the non-darkroom condition.

The application discloses a large language model test learning method, device, equipment and medium, test data is input into a large language model with low rank adaptive parameters, the prediction score of each token in the vocabulary at each generation step is obtained; the token-level evidence quality of the generation step is extracted based on the prediction score, and the token-level cognitive uncertainty of the generation step is calculated; the token-level cognitive uncertainty is smoothed, the stable area of the generation step is determined according to the smoothed uncertainty, and the mask of the model update window is constructed; the mask entropy loss is calculated based on the mask, and the low rank adaptive parameters are updated through back propagation. The application can solve the problem of adaptive drift in model updating in related technologies.

The invention relates to the technical field of government affair informatization, in particular to a government affair knowledge graph construction method and system based on artificial intelligence, and the method comprises the steps: obtaining multiple pieces of multi-source government affair data, carrying out the format unification of the multiple pieces of multi-source government affair data, obtaining multiple unified texts, carrying out the cleaning segmentation of the multiple unified texts, obtaining multiple text segments, and carrying out the segmentation of the multiple text segments; performing subject recognition on the plurality of text segments to obtain a plurality of government affair entities, performing normalized alignment on the plurality of government affair entities to obtain a plurality of unique entities, obtaining a plurality of semantic relationships of the plurality of unique entities, and constructing a triple set by using the plurality of unique entities and the plurality of semantic relationships, and constructing a government affair knowledge graph by utilizing a plurality of triads in the triad set, and automatically generating a rule required by the rule engine by the government affair knowledge graph. The method can solve the problem that intelligent application of government affair knowledge is difficult due to dispersion of multi-source government affair data and low manual integration efficiency.

The invention provides a mail processing method and device, electronic equipment, a storage medium and a program product, and relates to the field of financial science and technology or other related fields. The method comprises the steps of obtaining text content of a mail and metadata related to a sender, processing the text content through a classification model to generate first probability distribution, and generating second probability distribution based on the metadata and historical data of the sender, and the category of the mail can be determined by fusing the first probability distribution and the second probability distribution. Through two-channel processing of email text semantics and metadata, the limitation of single semantic features can be eliminated, the ambiguity and wrong classification problems of semantically similar emails are solved, the email classification accuracy is improved, and the subsequent processing flow can be conveniently carried out based on the email classification result.

The invention provides a fault basin sand body distribution prediction method and device based on fault activity quantitative characterization, equipment, a storage medium and a product, and relates to the technical field of sand body prediction, and the method comprises the steps: obtaining three-dimensional seismic data and logging data of a target region, and carrying out the structural interpretation of the three-dimensional seismic data, and obtaining the structural interpretation data; calculating a thickness difference grid between a fault hanging wall and a fault footwall of the target fault in the target interval based on the structural interpretation data; calculating a fault activity index of the target fault in a target interval deposition period based on the structural interpretation data; based on the structural interpretation data and the logging data, determining a containing space and a supply ratio of the target interval; and generating a sand body thickness distribution diagram of the target interval based on the thickness difference grid, the fault activity index and the containable space and supply ratio. According to the invention, a more accurate sand body distribution prediction result can be obtained.

This invention discloses an adaptive wellbore correction method based on a wellbore response database, belonging to the field of geophysical logging technology. This adaptive wellbore correction method, based on a wellbore response database, is applied to actual logging data from the small-diameter array induction logging instrument HACRT. For wellbore impacts with intrusion, a wellbore correction model with intrusion is introduced. The intrusion impact is calculated using geometric factors, and the wellbore response is calculated using the wellbore response database to obtain a logging response estimate. This invention solves problems such as skin effect correction of the wellbore model response signal, inconsistency between the wellbore model and the actual logging environment, and multiple solutions in wellbore model parameter inversion. It considers engineering problems in actual logging data processing and is adaptable to harsh well conditions such as larger well diameters and low-resistivity salt-cement slurries.

The application provides a decision support method and system based on AIGC integrated clinical data and knowledge base, and relates to the technical field of intelligent diagnosis and treatment. The method comprises the following steps: receiving a natural language query for a patient; standardizing the natural language query by using a term mapping network and an entity recognition model to obtain standardized clinical terms; generating a clinical decision backbone based on the standardized clinical terms and a preset clinical knowledge graph by using a meta-path algorithm; and generating a clinical decision suggestion containing traceability information based on a prompt engineering strategy and the clinical decision backbone by using an AIGC model. The application can realize dynamic traceability and rationality checking of diagnosis and treatment suggestions, significantly improve the accuracy of clinical decision and the trust degree of doctors, and provide reliable technical support for individualized precision medicine.

The present application belongs to the technical field of data processing of land-sea interface zone, and particularly relates to a joint processing method of comprehensive geophysical data of land-sea interface zone, which is used for processing and interpreting data of geophysical prospecting methods at the land-sea interface zone, finding out geological structure features and distribution of adverse geological bodies in the study area, including data processing and comprehensive interpretation of different methods, wherein the method comprises A: conical field source transient electromagnetic method, B: geological radar data processing method, C: high-density resistivity method, and D: micro-motion data processing method, and the specific steps of the conical field source transient electromagnetic method are A1: data arrangement and preprocessing are performed on original measurement data, and the arranged data are converted into.TEM format, so that the data become editable data traces. The present application has higher sensitivity to faults and fracture zones, has strong adaptability to urban environment, and to some extent, overcomes the problem of multi-solution, and has high efficiency, low cost, and higher accuracy and reliability.

The invention discloses a satellite image building element parameterization reconstruction method and system based on physical geometric constraint and differential rendering coupling. The method comprises the following steps: firstly, constructing a multi-scale multi-task network based on Swin Transform to extract image features; secondly, a physically guided shadow consistency constraint is introduced, height prediction is reversely calibrated by using a shadow geometrical relationship, and the ambiguity problem of monocular height estimation is solved; meanwhile, regular vector extraction is carried out on the building contour by utilizing a frame field theory; finally, a reverse modeling strategy based on micro rendering is adopted, roof geometric parameters are automatically solved, and an LOD2-level parameterized three-dimensional model is generated. According to the method, the physical consistency, the contour regularity and the modeling automation level of three-dimensional reconstruction are effectively improved, and the method is suitable for high-precision construction of large-scale urban digital bases.

This invention provides a method for predicting mapping relationships of medical data fields based on world model learning, belonging to the field of model prediction technology. The method includes acquiring a medical data object containing a source field set and a target field set, performing causal state configuration based on a generative world model to generate intermediate world states, performing alignment inference, and combining consistency policy constraints to generate mapping relationship prediction results. Simultaneously, it utilizes backfilling feedback information for closed-loop optimization. This invention improves the accuracy and interpretability of medical data field mapping and reduces the mapping error rate.

The invention discloses a coal field earthquake interpretation multi-mode hidden geological disaster information fusion and dynamic modeling method. The method comprises the following steps: acquiring multi-source heterogeneous geological data of a target coal field area; according to the method, the multi-modal feature tensor fusion model is constructed, so that the problem of multiplicity of solutions of a single data source is effectively solved, and the integrity and consistency of hidden geological disaster features are remarkably improved; conventional seismic attributes are extracted, high-order statistical characteristics of a logging curve and geologic body structure parameters are introduced, lithologic sudden change and tectonic deformation are described from multiple dimensions, and the recognition capacity of a micro-scale hidden disaster body is enhanced; a hybrid neural network combining CNN and LSTM is adopted, historical stratum stress field data and geologic features are fused, a space-time prediction module is constructed, the limitation of traditional static snapshot type interpretation is broken through, a geologic structure stability probability body of a future time sequence can be output, and dynamic prediction of a disaster evolution trend is achieved.

This application discloses a learning method, apparatus, device, and medium for testing a large language model. Test data is input into a large language model with low-rank adaptation parameters to obtain the prediction score of lexical units in the vocabulary at each generation step. Based on the prediction scores, the lexical-level evidence quality of the generation step is extracted, and the lexical-level cognitive uncertainty of the generation step is calculated. The lexical-level cognitive uncertainty is smoothed, and a stable region of the generation step is determined based on the smoothed uncertainty. A mask for the model update window is constructed. Based on the mask, the mask entropy loss is calculated, and the low-rank adaptation parameters are updated through backpropagation. This application can solve the problem of adaptive drift in model updates in related technologies.