35 results about "Relational encoding" patented technology

The invention provides a multi-modal data pairing method and system based on deep learning, and relates to the technical field of data processing, and the method comprises the steps: obtaining a video multi-frame sequence and a target text, and respectively extracting an overlapped frame group set and a standardized text sequence; performing spatio-temporal feature extraction and text dependency relationship coding to obtain a video time sequence vector sequence and a text vector sequence; executing cross-modal alignment search, and constructing a monotonic matching path set; calculating a semantic and action entity relationship consistency score of the paired elements on the path to obtain a comprehensive score; and determining an alignment relationship between the video and the text based on the optimal path. According to the method, accurate matching of the video and the text is realized, and the cross-modal retrieval efficiency is improved.

The invention provides a layout generation method and device based on a large model, electronic equipment and a storage medium, relates to the technical field of artificial intelligence, in particular to the technical fields of computer vision, deep learning, the large model, image processing and the like, and can be applied to scenes such as content generation based on artificial intelligence. According to the specific implementation scheme, visual information of an original image is extracted through a target detection model and a semantic segmentation model; converting the visual information into structured data, and converting the structured data into natural language description by utilizing spatial relationship coding; constructing a multi-modal Prompt based on the natural language description and the user instruction; and inputting the constructed multi-modal Prompt into the large model to obtain layout information of the target object output by the large model in the original image. According to the scheme, the layout generation quality and efficiency can be improved.

The invention discloses a news event prediction method based on heterogeneous evolutionary event clustering, which comprises the following steps: generating event representations based on preliminarily updated entity representations and relationship representations in a constructed entity graph, and constructing an event graph by taking events as nodes and taking heterogeneous relationships between the events as edges; obtaining event clusters through fuzzy clustering and constructing an event cluster graph; using a self-supervised optimization algorithm to optimize the event cluster representation according to the distance and similarity between the event clusters on the event cluster graph; capturing implicit correlation among the event clusters by using an implicit relation encoder, and sequentially updating representation of the event clusters, representation of events, and representation of entities and relations after sparsification and information aggregation; and predicting through the news event model based on convolution. According to the method, the pairwise correlation, the high-order correlation and the multi-step time sequence evolution of the events are effectively modeled, and the method has important application value in the aspects of international situation analysis, social governance, intelligent decision support and the like.

The embodiment of the invention provides a user risk identification method and device, electronic equipment and a storage medium, and relates to the technical field of computers. According to the method, multi-source time series data are fused, local space correlation characteristics of user behaviors are extracted by using a convolutional neural network in sequence, a long-term and short-term memory network captures a long-range time dependency relationship of a behavior sequence, a Transformer encoder mines deep correlation characteristics among key behavior segments, and risk decision is carried out by integrating the three characteristics. According to the method, multi-dimensional collaborative recognition of a complex and hidden risk mode in digital right operation is achieved, the accuracy and robustness of risk recognition are improved, the false alarm rate and the missing report rate are reduced, the limitation of a traditional risk control method or a single model in space-time-semantic joint modeling is overcome, and the method has higher generalization ability and practical application value.

The invention discloses a geospatial data geometric topological relation evaluation method and system based on a large language model, and belongs to the technical field of spatio-temporal information processing, and the method comprises the following steps: S1, data preprocessing and model construction: processing a training data set containing geospatial data and a corresponding geometric topological relation thereof, and constructing a model; the geometric topological relation of the geographic space data is coded into text or vector representation which can be understood by a large language model; the coded training data is utilized to finely adjust a pre-trained large language model, so that the pre-trained large language model can evaluate a topological relation according to input geographic space data description; s2, geographic space data input and feature extraction; and S3, large language model reasoning and topological relation evaluation. According to the method, by combining the language understanding ability of the large model and the spatial topology calculation technology, intelligent and high-precision evaluation of the geographic spatial data geometrical relationship is achieved, and a brand-new and self-adaptive evaluation technology is provided for the field of geographic spatial data processing.

The present invention relates to managing version-controlled datasets. An updated version-controlled dataset is created by applying a structured query to a specific version of a plurality of versions of the version-controlled dataset stored in a backing dataset. The data of the version including the specific version remains unchanged. A version-controlled dataset catalog is updated to include the structured query and lineage associated with the updated dataset version, the lineage indicating that the updated version-controlled dataset includes a successor version of the specific version. A dependency code associated with the updated version-controlled dataset is updated, the dependency code indicating which data of the updated version-controlled dataset matches the data of the specific version and which data of the updated version-controlled dataset is deleted relative to the data of the specific version. A database view of the updated version-controlled dataset is generated by selecting rows of the backing dataset corresponding to the updated version-controlled dataset according to the dependency code and projecting columns.

An inductive link prediction method based on subgraph and path comparative representation learning is characterized in that an overall knowledge graph comprises a plurality of structured triples et = (s, rT, o), surrounding subgraphs containing target nodes are extracted, isolated nodes and a complete neighbor relation are reserved, and node features of complete semantic information between neighbor nodes are obtained. And extracting a topological relation path between the head entity s and the tail entity o from the knowledge graph. Local structures and relation semantics are obtained through node-level relation coding, sub-graph-level aggregation perception message passing and an attention mechanism. And carrying out modeling on paths of adjacent relations of the sub-graphs to obtain features in the sub-graphs. A topological structure of a target map is associated with path representation, marginal-based loss is incorporated into distance scores of a positive path sample and a negative path sample, the positive path sample and the negative path sample which need to be distinguished are compared and learned through semantic information conveyed by a relation path, and a loss function is optimized through joint training.

This invention proposes a method, system, device, and medium for relation completion in cement-based materials, belonging to the field of cement-based composite materials technology. The method includes: constructing a set of triples for a material spectrum based on original text samples of the cement-based material's formulation, process, and properties; training an encoder based on the triple set to obtain a first relation encoding model; mixing unlabeled samples into the original text samples and using the first relation encoding model as a model base, semi-supervised training of the first relation encoding model using a self-adversarial loss function to obtain a second relation encoding model, and extracting the source node embedding, target node embedding, and relation embedding of all triples; constructing a positive and negative sample pair input self-interference decoder to predict missing relations in the material spectrum; traversing the material spectrum, inputting the embeddings of any two nodes, and completing the missing relations based on the self-interference decoder. This invention achieves relational semantic reasoning for cement-based materials, thereby improving the accuracy of relation completion.

The invention relates to the technical field of road disease detection and maintenance, in particular to an enhanced road disease knowledge extraction method and system based on large model self-learning, and the method comprises the following steps: carrying out the entity and relation extraction of a road disease field standard document through a pre-trained large model BERT, and generating an initial knowledge triple; a road disease knowledge extraction decision table is constructed, a hierarchical relation coding rule is defined, an inter-entity relation strength value is quantized, and the relation strength value is calculated through dynamic fusion of an expert initial score, a data-driven co-occurrence frequency and an attention weight; the method has the beneficial effect that the propagation risk of the error relationship is effectively reduced through the constraint of the relationship weight table and an expert rechecking mechanism. In 1000 pieces of test data, the conflict detection module successfully recognizes and corrects 83.2% of semantic conflicts (such as'seam strip 'error association' subgrade subsidence '), and the false alarm rate is lower than 5%.

The application discloses a kind of fusion object position relationship's image description method and device, which comprises the following steps: S1.input image to be described, extract the global feature information of image to be described and candidate frame feature information;S2. according to the feature information extracted, construct object relationship scene graph;S3. the node information of each kind contained in object relationship scene graph is initially encoded;If there is proportion imbalance phenomenon in the position relationship proportion between object nodes in object relationship scene graph compared with the original position relationship of object, then according to the proportion imbalance degree, the object relationship coding feature matrix obtained by initially encoding node information is secondarily encoded;S4. the object relationship scene graph and node information after coding are input into joint decoder for joint decoding, and the text information of the image to be described is predicted.The application can fully fuse the object position relationship in image, realize efficient and accurate image description.

The application discloses a kind of small sample learning method and device based on sample pair relation propagation.The method can obtain better relation coding by explicitly modeling and propagating the relation between the sample pairs of support set-query set.Through the introduction of pseudo-relation node, the characteristic information of the query set sample itself can be effectively retained. In addition, the application further provides an effective transduction learning strategy, which can better mine the relationship information between the query set samples, thereby obtaining more accurate classification results. Compared with the prior art, the present application better mines the potential information contained in the sample pairs of support set-query set in each task, and has higher accuracy and better generalization ability when processing a completely new task.

The invention relates to the field of machine learning, in particular to an industrial internet graph data representation learning method based on edge relation coding, which comprises the following steps of: firstly, randomly initializing embedded vectors with equal dimensions for each node and each edge respectively; constructing a self-center network, updating an edge vector in the self-center network, executing biased random walk on nodes, and sampling a node sequence according to a certain length every time to generate a set; and extracting nodes in a node context window according to the set to generate a context node sequence, and summarizing to obtain a corresponding set. And finally, node vectors are updated in sequence, a final embedding vector is obtained through iteration for a certain number of times, and then a node embedding matrix is constructed. Therefore, the problem that the performance of a downstream task is affected by loss of edge relation information in a graph due to neglect of edge difference in related technologies is solved.

The present invention discloses a log anomaly detection device based on component subsequence correlation perception, which belongs to the field of intelligent operation and maintenance and deep learning technology. The device comprises the following steps: dividing a portion of a log sequence S as a sample, obtaining a time feature embedding representation t and a semantic feature embedding representation V, and obtaining a time feature embedding representation t. c and semantic feature embedding representation V c Based on the proposed method, the system undergoes feature extraction, sequence dependency capture module, latent relation encoding module, graph convolution module, subsequence fusion module, embedding representation concatenation module, and log template prediction module to obtain the probability distribution of the log template in the next log message. All log templates are sorted and the prediction result is the top-k ranked log template set. The prediction result of the next real log template and sample is used to determine whether the log sequence has an anomaly. This effectively captures anomaly information and improves detection efficiency and accuracy.

The invention discloses an automatic framework design for BIM space recognition and classification, which adopts a node feature enhanced self-supervised graph neural network to bride the semantic gap between BIM and graph learning and capture geometric and spatial semantics. The method comprises the following steps: (1) analyzing a functional space representation form and summarizing types, and representing a building space as nodes and a space relationship as edges; (2) before graph propagation, coding a common wall, corridor, corner and open space connection topological relation into weighted edges, and injecting node features through a learnable fusion layer; and (3) training on three self-built building space layout map knowledge bases containing 12 space types and 4 relation characteristics to realize automatic identification of functional space types. According to the method, the space recognition efficiency and the automation level are improved, the high-confidence-coefficient function label corresponding to the IFC standard is generated, design review, facility management and evacuation path optimization can be accelerated, and an extensible basis is provided for ontology driving graph reasoning in multiple building scenes.

This invention discloses a method and system for constructing relationships between water conservancy objects based on dynamic weights and relational coding, belonging to the field of water conservancy information technology. The method includes: systematically classifying and defining water conservancy objects; constructing a water conservancy-specific relational coding system based on the definitions of water conservancy objects and relationships; performing relational calculations based on a generalized water system map and a business attribution map, and introducing a multi-factor dynamic weight adjustment mechanism; using a spatiotemporal graph database for integrated storage; and achieving iterative optimization of the parameters of the water conservancy object relationship construction system through an online learning and feedback mechanism. This invention achieves knowledge retrieval and matching through a water conservancy-specific relational coding system, introduces a multi-factor dynamic weight model to enable relational calculations to adapt to different hydrological conditions and business scenarios, realizes domain knowledge-driven intelligent reasoning based on a generalized water system map and a business attribution map, and constructs an online learning and feedback mechanism to enable the system to continuously optimize from historical data and real-time monitoring.

The application discloses a rumor detection method based on a graph attention network, belongs to the technical field of image processing, and combines a BERT pre-training model and a graph attention network CNN, wherein the relationship between information is constructed, rumor detection is assisted by means of propagation information, a BERT pre-training model is used for processing data in a tweet coding representation part, and then the feature representation of the CNN is used to complete mapping from word embedding to semantic space; local tweet relationship coding learns the combined representation of each source tweet and related forwarding from each source tweet corresponding forwarding by improving a graph attention network GATv2; global relationship coding indicates how to encode global structure to node representation; and a rumor detection module learns a classification function to predict the label of an original tweet. The original data is processed by using the BERT pre-training method, the information of context is fully considered, the context information is fused by a bidirectional language model, and the representation capability of the model is improved.

The application provides a group chat-oriented dialogue structure analysis model training method, analysis method and device. After obtaining the text representation of each message by using a BERT model, the attention representation of a target message speaker is mined by combining the defined relationship encoding between each speaker through a GRU. According to the reply relationship of the previous message, the corresponding thread is constructed, the target message is connected to each thread, and the content semantic representation of each thread is obtained by inputting the GRU. The discourse structure representation of the target message and the previous message is obtained by a multilayer perceptron. The true reply object of the target message is jointly evaluated, analyzed and judged by combining the attention representation, the content semantic representation and the discourse structure representation, and the accuracy of dialogue structure analysis is greatly improved.

The invention discloses a real-time voice-driven human body posture generation method based on a variational auto-encoder, and belongs to the technical field of digital human interaction, and the method comprises the steps: S1, carrying out the preprocessing and alignment of multi-modal data, and constructing a three-dimensional training sample comprising audio features, emotion labels and posture data; s2, constructing a cross-modal generation model based on a variational auto-encoder, wherein the cross-modal generation model comprises an encoder and a decoder; s3, designing a multi-objective loss function to train and optimize the cross-modal generation model; and S4, inputting the preprocessed audio to be processed into the trained cross-modal generation model, and obtaining a human body posture sequence with a coherent time sequence through real-time reasoning. A variational auto-encoder is applied to a voice-human body posture cross-modal generation scene, the traditional application boundary of the variational auto-encoder is broken, the complex mapping relation between audio features and human body postures is encoded into low-dimensional probability distribution through the probability modeling capability of a potential space of the variational auto-encoder, and a new technical path is provided for posture generation.

The invention discloses a method and system for discovering a time-space dependency rule of a cave mural, and relates to the field of time sequence knowledge graph reasoning in a knowledge graph, and the method comprises the steps: obtaining cave mural data needed by initial training, and inputting the data into an initial time sequence knowledge graph reasoning model for training; respectively inputting the trained text data into a relation encoder and an entity encoder to obtain relation embedding and entity embedding, and inputting time information into a time vector generator to obtain a corresponding time vector; performing visual feature extraction on image data in the obtained cave wall painting data; performing feature fusion on relation embedding, entity embedding, time vectors and visual features to generate comprehensive feature vectors; inputting the comprehensive feature vector into a decoder, and calculating a confidence score; and calculating loss according to the confidence score, and further training the time sequence knowledge graph inference model. According to the invention, the reasoning accuracy of the model on the time sequence knowledge graph is improved.

A technique performs tasks completed by knowledge graphs in a both scalable and resource efficient manner. In some implementations, the techniques identify a source entity having a source-target relationship that connects the source entity to a target entity to be determined. The technique also identifies a source-entity data item that provides a piece of source-entity text related to the source entity. The techniques map source-entity data items to source-entity encoding information using a machine-trained encoder model. The technique then predicts an identification of the target entity based on the source-entity encoding information and based on predicate encoding information encoding the source-target relationship. In some implementations, the techniques also predict the target entity based on consideration of one or more neighbor entities connected to the source entity and their respective source-to-neighbor relationships. The technique also allows for the delivery of knowledge across knowledge graph training phases.

This invention discloses a network data exchange method and system based on data storage, relating to the field of network data exchange technology. The method includes: constructing a field dependency graph by extracting dependencies between data fields in a source data set, and performing hierarchical labeling based on node connectivity; decomposing the data into fragment sets and dividing them into parallel groups according to the field dependency graph; establishing a bidirectional index of physical storage addresses and logical access identifiers, and encoding the topology relationships to generate a storage mapping table; decoding and obtaining the dependency topology relationships and setting access preconditions to generate a synchronization sequence; reading data fragments in the same group in parallel according to the synchronization sequence, and reading data fragments in different groups sequentially; and reassembling the data based on the dependency graph to generate a target set. This invention improves data exchange efficiency and optimizes storage resource utilization.

The invention discloses a human body posture estimation method based on a double-layer hidden space, which comprises the following steps of: firstly, training an encoder by using a vector quantization variational auto-encoder, encoding a joint dependency relationship between a human body posture and human body joint points into discrete mark sequences with different scales, and obtaining a trained codebook and a decoder; then freezing the obtained codebooks and the decoder, extracting image features from the preprocessed image through a neural network, firstly performing image feature preprocessing, then performing nearest neighbor search algorithm quantization and splicing quantization on the image features by using two classification heads and combining the two codebooks to obtain two features, and inputting the two features into the decoder to reconstruct a posture; and outputting a human body posture estimation result. According to the method, the ambiguity problem caused by shielding is solved, and the detection precision of the positions of the shielded joint points is improved.

The present invention discloses a document-level relation extraction method based on long-tail data distribution, belonging to the fields of information extraction and machine learning. It includes document preprocessing, document encoding, relation encoding, data augmentation, and relation prediction. In terms of data augmentation, for the set of labeled triple vectors, the present invention randomly selects or presets the relation types that need to be augmented, designs a mask vector, and perturbs the pooled context representation in the original triple vectors to be augmented with data to generate new triple vectors; it can effectively improve the accuracy of the document-level relation extraction model in predicting tail relation types. At the same time, compared with the traditional text-based data augmentation method, the present invention does not require an additional text encoding process, improving the computational efficiency of model training. In addition, the contrastive learning pre-training framework proposed by the present invention can effectively improve the accuracy of document-level relation extraction in the long-tail data distribution scenario.

The invention discloses an artificial intelligence-based edge computing load balancing method, which belongs to the technical field of edge computing, and comprises the following steps of: obtaining an edge computing task queue, sequentially taking out a plurality of target edge computing tasks from the edge computing task queue according to a fixed number, and then aiming at the plurality of target edge computing tasks, carrying out load balancing on the target edge computing tasks; carrying out allocation relation coding on the edge computing server by adopting an intelligent coding method to obtain a plurality of task allocation codes, and carrying out load balance evaluation by adopting an artificial intelligence algorithm to determine the load balance corresponding to the task allocation codes; and finally, according to the load balance corresponding to the task allocation codes, determining a current optimal code, optimizing other task allocation codes according to the current optimal code, obtaining a final task allocation code, and on the basis of the final task allocation code, performing task allocation. And the target edge computing task is allocated to the edge computing server for processing, so that load balancing of edge computing is realized, overload of the edge computing server can be effectively avoided, and the overall data processing efficiency is improved.

The application relates to the technical field of natural language processing, and provides a dynamic knowledge graph prediction method and device, electronic equipment and a storage medium, the method acquires historical event corpus; the historical event corpus is input into a graph construction model to obtain a knowledge graph at a current time. The graph construction model realizes extraction and coding of implied relationships in the historical event corpus through an implied relationship extraction module and a relationship coding module, obtains the knowledge graph at the current time through an entity time sequence representation module, so that the obtained knowledge graph not only contains relationship information of each entity that already exists in the historical event corpus, but also contains the association relationship of each entity implied in the historical event corpus, the accuracy of the knowledge graph is higher, and the subsequent application effect of the knowledge graph is better. Moreover, through continuous updating of the historical event corpus, accurate dynamic prediction of the knowledge graph can be realized. The application has been subsidized by a national key research and development plan project (2019YQ1601).

The present invention discloses an image description generation method based on external triples and abstract relationships. First, the present invention extracts triples in the image description text, constructs an external relationship library and performs feature encoding on the triples. Triples with a text similarity higher than a threshold are clustered into one category. At the same time, the model performs object detection on the image to obtain a set of target visual features and a set of target categories; according to the text similarity, triples similar to the target and the target category are queried in the external relationship library. The model uses the target visual features to predict the objects, attributes, and relationships of the image respectively to generate a scene graph; and uses a convolutional neural network to fuse the visual features and text features to perform feature encoding on the objects, attributes, and relationships. Finally, the encoded features of the objects, attributes, and relationships in the scene graph are fused with the encoded features of the similar relationships and abstract relationships and input into a two-layer LSTM sequence generation model to obtain the final image description. The present invention makes the expression of the description generated by the model richer.

The embodiment of the invention relates to a speech synthesis method and device, computer equipment and a storage medium. The method mainly comprises the following steps of: analyzing sentiment semantic information and an inter-word dependency relationship of text data to be synthesized by adopting a graph sentiment semantic encoder, and performing sentiment semantic enhancement to obtain sentiment semantic features; converting the to-be-synthesized text data into phoneme data; performing grammar relation coding on the text data according to the grammar graph of the to-be-synthesized text data by adopting a graph coder to obtain relation coding characteristics, and performing text coding on the phoneme data by utilizing the relation coding characteristics based on a graph attention mechanism to obtain text coding characteristics; splicing the emotion semantic features and the text coding features to obtain spliced features; performing time length prediction on the splicing features to obtain prediction data; decoding the prediction data to obtain spectrum data to be synthesized; and converting the spectrum data to be synthesized into synthesized audio. By adopting the method, the accuracy and emotion expression of speech synthesis can be improved.

The invention discloses a heterogeneous collaborative recommendation method and system for multi-geometric interactive embedding and graph convolution comparative learning, and the method comprises the steps: constructing a knowledge graph triple and a user-article interactive graph from original data, extracting head entities from the knowledge graph triple and the user-article interactive graph, embedding the head entities, and mapping the head entities to a Euclidean space, a hyperbolic space and a spherical space; after mapping results of different spaces are unified through tangent space projection, fusion of the different spaces is achieved through cross attention; after fusion is completed, a graph attention network is used for executing attention aggregation of message propagation, geometrical relationship coding and structure perception on an interaction graph, and final representations of the user and the article are obtained respectively. And according to the final representations of the users and the articles, generating predicted scores of all the articles by the target user, and sorting the predicted scores to generate a recommendation list to finish article recommendation.

The present invention relates to a method for image prediction classification using a context reasoning network based on Graphormer. The present invention proposes a context reasoning model based on Graphormer and designs a degree centrality relationship encoding. This method can specifically spread context information between regions, thereby improving the efficiency of small object detection. It uses Transformer to model and infer the semantic and spatial layout relationships between nodes in the graph structure, as well as the degree centrality of the nodes themselves, and retains the spatial information as much as possible while extracting the semantic features of small objects, effectively solving the problems of false detection and missed detection of small objects.

The invention relates to a set function method based on an attention mechanism, and the method comprises the following steps: 1, obtaining corresponding data from different sensors, the data source comprising corresponding elements of a 3D point cloud or extraction features corresponding to a picture; step 2, inputting the obtained data into an SAB module and converting the input elements into set codes; and step 3, inputting the set code into a subsequent decoder, and mapping the set code to any dimension by the decoder. According to the method, the requirements of replacement invariance and any input scale are met, and meanwhile, the paired or higher-dimensional element relationship among the set elements is coded to a fixed length through the SAB and the RSAB, so that the information of the set elements is aggregated into coding in a new mode, and the original element relationship is reserved during aggregation.