114 results about "Semantic consistency" patented technology

Systems and processes are disclosed for handling a multi-part voice command for a virtual assistant. Speech input can be received from a user that includes multiple actionable commands within a single utterance. A text string can be generated from the speech input using a speech transcription process. The text string can be parsed into multiple candidate substrings based on domain keywords, imperative verbs, predetermined substring lengths, or the like. For each candidate substring, a probability can be determined indicating whether the candidate substring corresponds to an actionable command. Such probabilities can be determined based on semantic coherence, similarity to user request templates, querying services to determine manageability, or the like. If the probabilities exceed a threshold, the user intent of each substring can be determined, processes associated with the user intents can be executed, and an acknowledgment can be provided to the user.

A method for managing electronic records is provided. Each electronic record includes a data file, a plurality of data files, a portion of a data file, or portions of a plurality of data files. The electronic records include a plurality of record types and data file types. The method includes forming a data file set comprising one or more logically related data files; identifying attributes of each record type in a record type template; identifying specifications of each data file type in a data file type template; and extracting digital components from the data file set. The extracted digital components relate to the attributes in each record type template and the specifications in each data file type template and compose an individual record. An electronic record archive includes record type and data file type templates and a digital component extractor.

The invention provides a method and a system for extracting a Chinese event. The method comprises the following steps of: performing phrasing, word-splitting, entity identification and analysis for syntax and dependence relationship on a text with a to-be-extracted event in turn; marking the words meeting an extracting condition as candidate triggering words, according to internal structures of the words; filtering the triggering words meeting a filtering condition according to the probability, the word class and the internal structures of the words; extracting the triggering words by utilizing the maximum entropy identifying model and obtaining the reliability of each of the triggering words; dividing the triggering words into a consistency processing training set and a consistency processing testing set according to the reliability of each of the triggering words; utilizing a maximum entropy classifier to extract the triggering words from the consistency processing testing set; and utilizing a maximum entropy classifying model to classify the triggering words, thereby obtaining an event set. According to the method and the system provided by the invention, started from the characteristics of Chinese, the internal structures of Chinese words and the semantic consistency of the Chinese words in sections and chapters are comprehensively considered and analyzed, so that the property of extracting the Chinese event is increased.

The invention discloses a zero sample image classification method and system. The method includes the following steps of: S10, inputting training data belonging to a visible category and a category tag thereof for feature extraction; S20, inputting the semantic assistance information of all category tags to obtain the semantic embedded representations of respective tags and semantic difference measure between the tags; S30, establishing a zero sample classification model, performing semantic consistency regularization on the model based on the semantic differences between the categories; S40,iteratively updating model parameters until convergence; and S50, predicting the category tag of a test image. The method and system establish a semantic consistency regularized zero sample classification model so that the output of the model conforms to the semantic neighborhood relation between categories to adapt to the semantic structure of the target category so as to obtain high classification accuracy.

The invention relates to a network image retrieval method based on semantic analysis, which is used for extracting low-level features. Content-based image retrieval is performed on each type of feature to find out a visually-similar network image set. The related text information is used for semantic learning corresponding to each image in the network image set corresponding to each image in the network image set to obtain the semantic expression for the image query. The semantic consistency of the retrieval image set corresponding to various features on the text information is judged, the semantic consistency is used to measure the description capacities of various features, to endow the description capacities with different degree s of confidence. The semantics and semantic consistency of the image query are adopted to perform text-based image retrieval in the image base to obtain the semantic relevance of each image in the image base and the image query; the low-level features are adopted to perform content-based image retrieval on the image base to obtain the visual relevance of the each image in the image base and the image query; the semantics is fused with visual relevance through a linear function to ensure the image for the user to have both semantic and visual relevance.

The invention provides a cross-modal generalized zero sample retrieval method based on a dual learning generative adversarial network. The method comprises: constructing a generative adversarial network based on dual learning; mapping the high-dimensional visual features of different modes to a common low-dimensional semantic embedding space; secondly, constructing multiple constraint mechanisms to perform cyclic consistency constraint, generative adversarial constraint and classifier constraint so as to maintain visual-semantic consistency and generated feature-source feature consistency, andperforming cross-modal retrieval after training of the whole network, so that the model is more powerful in performance in generalization of zero-sample retrieval. Meanwhile, in the whole training process, paired multimedia data pairs on the pixel level do not need to serve as training samples, only paired data on the category are needed, so that the complexity and expensive cost of data set collection are reduced, the retrieval effect is better, and performance improvement is more obvious in the zero-sample generalization retrieval problem.

The invention discloses a body-based verification tool of a power grid public information model and a method of the body-based verification tool, wherein a model verification system which is provided for analyzing metadata information from a CIM / XML (Common Information Model / Extensive Makeup Language) power grid model, comparing the metadata information with a uniform standard information model and analyzing compatibility and semantic consistency of the model; and aiming at semantic requirements of model verification, an OWL (Web Ontology Language) main body language is adopted as the description language for integrating the body model, and a body integration verification method based on CIM Schema and CIM Profile is provided. According to the invention, the defect of power grid enterprise information integration bus on semantic specification and the problem of mismatching models occurring during mutual operation process are solved, thereby being beneficial to the specification in semantics of the information bus and convenient to the management of assembly models and the upgrading maintenance of CIM.

The invention discloses a cross-modal retrieval method based on multilayer semantic alignment, which comprises the following steps of: acquiring a remarkable fine-grained region by utilizing a self-attention mechanism, promoting the alignment of entities and relationships among modal data, providing an image text matching strategy based on semantic consistency, extracting semantic tags from a given text data set, and performing global semantic constraint through multi-label prediction, so that more accurate cross-modal association is obtained. Therefore, the problem of cross-modal data semantic gap is solved.

The invention discloses an answer selection method and system for allowing a reading robot to perform reading comprehension, which can improve the accuracy of answer selection while a robot performs reading comprehension. The answer selection method for allowing a reading robot to perform reading comprehension includes: positioning a paragraph associated with a question; adopting a bag of words model to extract a sentence having the most matching words with options as evidence materials; analyzing the semantic consistency of the options and the evidence materials; and selecting an optimal answer to the question according to the consistency analysis result. The answer selection method and system can applied to the reading robot, and can intelligent answer the selected question through automatic analysis and deduction.

The embodiment of the invention discloses a text translation method and device, equipment and a storage medium. The method comprises the steps of obtaining an original language text; translating the original language text by using a text translation model; obtaining a target language text corresponding to the original language text, wherein the text translation model is a translation model obtained by correcting an original translation model according to a text evaluation result of a training translation, the training translation model is an output result after translation of the original translation model, and the text evaluation result is used for evaluating the semantic relationship of the context text in the training translation. According to the technical scheme provided by the embodiment of the invention, the problem that in the prior art, semantic consistency before and after translation of a translated text obtained by independently translating each sentence is poor is solved;according to the technical scheme, the text translation model is effectively corrected, the translation accuracy of the text translation model is improved, and then the front-back semantic consistencyand fluency of translated texts are improved.

The invention relates to a video content description method based on frame selection, comprising the following steps: a feed-forward neural network is used for constructing a screening model, and thescreening model screens the video frame according to the visual richness and semantic consistency of the video frame; constructing a description model for describing the content of the video to be described; training the screening model and the description model with training data; Selecting a description frame in the video to be described through the filtering model; the visual features of the description frame are extracted and the description model is input to obtain the description sentence of the video to be described.

A CNN-based sky talk semantic consistency checking method comprises the steps of picking different repeat statement pairs from a real sky talk record to form positive sample data characterized in semantic consistency; forming negative sample data characterized in semantic inconsistency, and forming a corpus by means of the positive sample data and the negative sample data; conducting word segment on each repeat statement pair in the corpus; converting each word into a word vector, and forming a word vector library by means of the word vector; forming a corresponding matrix for each word; forming two semantic vectors of the repeat statement pairs with the matrix of each repeat statement as the input of a convolution nerve network input layer; calculating cosine similarity of two semantic vectors of the repeat statement pairs; conducting classification by means of the k-nearest neighbor method. By means of the parameter sharing advantage of the convolution nerve network, operation time can be shortened, and overall processing of statements can be conducted when the semantic vectors of statements are calculated.

The invention relates to a semantic consistency verification method of a semantic Web service combination, which belongs to the field of semantic nets and Web services. The invention is characterized in that the method is realized in a way that: establishing a storage unit of a semantic Web service coloring Petri net model, a parameter semantic consistency verification submodule, a QoS (Quality of Services) attribute consistency verification submodule and a consistency analytic result storage submodule on a computer; and analyzing the semantic consistency of the semantic Web service combination by using a coloring Petri net formalized verification method, wherein the semantic consistency comprises parameter semantic consistency and QoS attribute consistency. The invention can verify whether the semantic Web service combination is consistent in semantics and satisfies the demands of users for service non-function attributes.

The invention discloses a consistency expression method for multi-source heterogeneous big data. In the method, given the characteristic heterogeneity of multi-source heterogeneous data, the multi-source heterogeneous data is projected to a middle-layer redundant feature homogeneous space by use of the semantic complementarity of the multi-source heterogeneous data and based on a subspace learning method; and related descriptions from different sources are coupled together in the homogeneous space. In order to excavate the semantic consistency of homogeneous descriptions in the middle-layer space, the characteristic homogeneous descriptions are projected to a high-layer semantic sharing subspace by use of the priori knowledge so as to eliminate the redundancy and noise information; and a sematic-consistent mode of the multi-source heterogeneous data can be obtained. The method disclosed by the invention is conducive to obtaining an accurate and robust multi-source data evaluation analysis result in the fields such as multimedia analysis, information retrieval and medical diagnosis.

The invention relates to a multi-platform point cloud intelligent processing method for holographic mapping. The method comprises the steps of organizing and scheduling mass point cloud data; carryingout point cloud data quality control and improvement, and achieving automatic correction of point cloud position consistency under the condition of no control point; carrying out high-precision fusion of multi-platform laser point cloud data, which comprises neighbor point cloud search, global matching energy equation construction and bipartite graph minimum cost matching; automatically extracting the ground object targets, namely realizing high-precision extraction and vectorization of the full-type ground object targets through geometric semantic consistency extraction of the ground objecttargets; carrying out the ground object target multi-detail-level model reconstruction based on the Grignard law, which comprises the step of establishing a multi-detail-level three-dimensional modelof the ground object target through a Grignard mathematical model and a topological relation graph. The holographic surveying and mapping product production process based on the multi-platform point cloud data is achieved, the method is easy to operate and implement, the manual workload of data processing can be greatly reduced, the working production efficiency is improved, and the product updating period is prolonged.

The invention discloses a cross-modal retrieval method and system based on pseudo label learning and semantic consistency, and the method comprises the steps: receiving an image data set and a text data set which comprise marked image text pairs and unmarked image data; Learning a projection matrix projected from the image space to the text space, and projecting the unmarked image data to the textspace; calculating the class center of the marked text; Distributing pseudo labels for the image data according to the similarity between the projection data of the unmarked image data and the classcenters of the text data, and taking the text data corresponding to the class center closest to the image data as a corresponding text mode; Taking the marked image data distributed with the pseudo tags and the corresponding text data as a training data set, and learning a projection matrix of images and texts projected to a public semantic space; Performing cross-modal retrieval. Unmarked data isintroduced into the training data set, and a more effective projection matrix can be obtained.

The invention provides a text consistency calculation method and device. The method comprises the steps of obtaining an article to be processed; extracting a plurality of sentences related to the title from the text; for each sentence in the plurality of sentences, inputting the sentence and the title into a preset semantic consistency model, and obtaining a semantic vector corresponding to the sentence; determining a semantic vector corresponding to the article according to the semantic vector corresponding to each sentence; for each sentence combination in the plurality of sentences, inputting the sentence combination into a preset logic consistency model, obtaining a logic label corresponding to the sentence combination, and determining a logic vector corresponding to the article according to the logic label corresponding to each sentence combination; and according to the logic vector and the semantic vector, determining the text consistency value of the article. According to the method, the quality of the text can be accurately evaluated on the basis of the semantic consistency and the logic consistency of the text, meanwhile, a user can be helped to evaluate and assist the writing quality, and the use experience of the user is improved.

The invention discloses a method for verifying a compiler by graph isomorphism. A device comprises a source program to be processed and the compiler, wherein the source program is transformed into a target program by the compiler. A decompiler is built in a selected environment, the target program is decompiled into a comparison program by the decompiler, semantic consistency between the comparison program and the source program is verified, and isomorphism judgment is performed after relevant node information of control flow graphs of the source program and the comparison program are respectively built on the basis, so that whether the compiler is secure or not is judged. Security of the compiler is verified by judging whether the control flow graphs of the comparison program and the source program are isomorphic or not. By the method, a decompiled code and a source code are directly semantically compared, incomplete verification caused by information loss of the source code in the verifying process is avoided, the control flow graphs do not need to be traversed again in the algorithm, and the graph isomorphism judgment algorithm is simple and easy to implement.

The invention discloses a method for verifying the semantic consistency of land-air conversation based on improved LSTM-RNN. The method comprises the steps: making a corpus: making a special vocabulary according to the conversation standard and the corpus of the civil aviation and obtaining a one-hot vector of the words: generating a semantic vector of two sentences in each sentence pair; processing the semantic vectors through an average pooling method and inputting the processed semantic vector into an MLP model; learning the degree of correlation between two semantic vectors through the MLP; determining the consistency of semantics through KNN. An original method for processing a sequence based on LSTM-RNN is lower in verification precision even if preventing the gradient of an independent RNN algorithm from disappearing. The method provided by the invention employs the average pooling method for preventing the overfitting phenomenon, innovatively employs the MLP for the computing,and enables the model to be more complete in the learning degree of the correlation. Because a model in deep learning is used for automatically learning the sample features, the method does not need the statistical analysis of mass data.

The invention provides an image semantic segmentation model training method and device, an image semantic segmentation method and device, and a storage medium, and relates to the technical field of computers, and the method comprises the steps: carrying out the judgment of a semantic segmentation image generated by a semantic segmentation model through employing a discriminator model; constructinga loss function corresponding to the discriminator model, wherein the loss function comprises a target domain loss function generated based on the target domain image, the target domain loss functioncomprises at least one of a first semantic loss function generated based on semantic consistency of the image blocks, a second semantic loss function generated based on semantic consistency of the clustering clusters and a third semantic loss function generated based on image space logic construction. According to the method and device and the storage medium, the semantic segmentation model reasoning result of the model on the target domain image is constrained in the form of the regularization item in the training process, cross-domain migration is performed on the image semantic segmentation model, and the efficiency and accuracy of image semantic segmentation model training are improved.

The invention discloses a word vector and context information-based short text topic model. A semantic relationship between words is extracted from word vectors; by explicitly obtaining the semantic relationship, the shortcoming of word co-occurrence deficiency of short text data is made up for; and through data of a training set, the semantic relationship between the words is further filtered, sothat the data set can be better trained. A background topic is added in a generation process; and through the background topic, noise words in a document are modeled. The model is solved by using a Gibbs sampling method in model inference, and the probability of the words with relatively high semantic correlation in related topics is increased by using a sampling strategy of a generalized Polya urn model in the sampling process; and in the way, the semantic consistency of the words in the topics is greatly improved. A series of experiments show that the method provided by the invention can improve the semantic consistency of the topics to a greater extent, and a new method is provided for short text topic modeling.

To provide a workflow tracking apparatus having scalability that can be efficiently operated. A framework of management using a hierarchical structure such as DNS is introduced. A simple decentralization structure leads to a failure in establishing semantic consistencies of Wf definition and meta-information, results in incapability of management. Therefore, there are additionally prepared (1) an element measure that manages, as a predetermined designated standard form, the descriptions of element workflows to be decentralized; (2) an element measure that transforms the descriptions of element workflows to be decentralized into the standard form; and (3) a workflow combining measure that combines the descriptions of element workflows to be decentralized, thereby reproducing the description of the entire workflows. These measures are used to combine the descriptions of element workflows to be decentralized, thereby generating consistent Wf definition and meta-information, thereafter copying replicas of them to a workflow monitor system that manages the domain.

The invention discloses a cross-modal retrieval method and device for multi-modal data, cross-modal retrieval equipment for multi-modal data, and a computer readable storage medium. The method comprises the steps: inputting training sample data of different modals into deep neural networks corresponding to all modals in batches, and obtaining the sample data features of all training sample data; respectively mapping each sample data feature into a common space, and calculating a corresponding loss function according to the intra-class low-rank loss constraint and semantic consistency constraint of each training sample data of different modes of the same class; adjusting network parameters of the deep neural network by using a loss function, and determining a target feature extraction model; and, after target data and to-be-retrieved data of different modes are obtained, calling the target feature extraction model to perform cross-modal retrieval operation, and then obtaining a retrieval sorting result of the to-be-retrieved data corresponding to the target data, so that the target feature extraction model can extract data features with higher quality, thereby improving the accuracyof cross-modal retrieval of the multi-modal data.

The invention relates to a knowledge graph construction method, which comprises the steps of receiving data in real time after graph library cluster configuration is completed; analyzing the ID numberand each key value pair in the data, judging the type of the data, and if the type of the data is an event with a time attribute, creating a new event node by using the ID number, or modifying a corresponding attribute value in the original event node in the picture library by using the key value pair. The knowledge graph constructed through the method is a real-time attribute graph with enterprises, natural persons and events as main bodies, the semantic consistency is guaranteed, the semantics are clear, the reusability is high, a graph library does not need to be created for different scenes, and more information can be constructed in the graph in the connection mode.

The invention provides a bidirectional text image generation method and system based on semantic consistency, and the method comprises the steps: obtaining a natural language, inputting the natural language into a text encoder, and extracting a word vector and a sentence vector; inputting sentence vectors into an image generation network, inputting word vectors into a word-level space attention mechanism module, connecting image features generated in each stage in the generation network with the word vectors adjusted by an attention mechanism in series to serve as input of the next stage of the generation network, and performing multi-stage refinement, and finally generating an image; inputting the generated image into an image encoder, and extracting image features; inputting the image features into a long short-term memory network with a sentinel mechanism, and outputting a re-described text; and after the image and the re-description text are optimized by adopting two confrontation loss functions, introducing semantic text reconstruction loss based on cross entropy to further optimize the image until the re-description text of the image is consistent with a natural language, and outputting the image.

The embodiment of the invention discloses a positioning method, and the method comprises the steps: obtaining a to-be-positioned image, carrying out the pose recognition of the to-be-positioned image,and obtaining the initial pose information of electronic equipment during the collection of the to-be-positioned image; determining a semantic consistency weight value of a reference image corresponding to the to-be-positioned image, wherein the reference image is an image which is collected by the electronic equipment and is adjacent to the to-be-positioned image, the semantic consistency weightvalue is used for representing the similarity between the reference semantic information of the reference image and the first semantic information of the first image, and the first image is an imagematched with the reference image in a scene map of a scene where the electronic equipment is located; if the semantic consistency weight value of the reference image is greater than the weight threshold, determining target pose information of the electronic equipment when the to-be-positioned image is acquired based on the related information and the initial pose information of the reference image. The embodiment of the invention further discloses a positioning device, electronic equipment and a computer readable storage medium.

The invention provides an abstract generation method based on a planning mechanism and knowledge graph guidance. The method comprises the steps: directly taking a knowledge graph as the input of a model, extracting attention features from the uninterrupted global semantic features of the knowledge graph in an abstract decoding process, and maintaining the semantic consistency of the context of anabstract; compared with a text generation model adopting a graph neural network as an encoder, the method has the advantages that entity information covered by each sentence segment in the abstract isannotated in a supervised manner, so that a planning mechanism module of the training model can capture semantic features of a generated text before the sentence segments are generated; therefore, the sub-graphs of the knowledge graph are further explicitly extracted as fine-grained guidance, so that the model is guided to generate abstract texts with smoother semantics and more self-consistent logic.

The present invention provides a crowd semantic segmentation method and a crowd semantic segmentation system and belongs to the computer video processing technical field. The crowd semantic segmentation method includes the following steps that: time-domain motion group segmentation is carried out on images according to motion consistency, and space-domain distribution group segmentation is carried out on the images according to distribution uniformity; time-domain and space-domain segmentation result-integrated joint semantic group segmentation is carried out; and semantic consistency-based group state description is carried out. According to the crowd semantic segmentation method and system of the invention, the motion consistency and distribution uniformity of crowds are considered comprehensively in a crowd segmentation process, so that groups obtained after segmentation have stable states in the time domain and the space domain, and the states of the groups can be well described, wherein the states of the groups include motion patterns and density grades.

The disclosure provides a medical big data analysis method and apparatus. The method comprises: a sign symptom expressed by a vector h is received; sub space in which a patient is located in medical big data is located by using the sign symptom as a feature, wherein a matrix D is used for expressing a case set in the big data, D is equal to [D1, D2, ...,DM], the Di expresses ith sub space and thei is larger than or equal to 1 and is less than or equal to M, and the location includes calculation of a formula: h=DX; and semantic consistency of the sub space is analyzed to analyze the probability P1 of location of the patient in the specified sub space. In addition, on the basis of an evidence transfer score on a medical mapping knowledge domain, the probability P2 of location of the patientat a specified node is analyzed. The probability P of location of the patient in specified sub space or at a specified node is determined according to a formula: P=alpha+P1+(1-alpha)*P2, wherein thealpha expresses a harmonic parameter and is larger than 0 and is less than 1. Therefore, the accuracy of the analysis can be improved; and the condition of the patient can be analyzed based on the symptom at first time and thus the patient can be checked in an oriented manner, so that the cost is lowered and the efficiency is improved.