64 results about "Semantic difference" patented technology

Systems and techniques are provided for tracking puts and takes of inventory items by subjects in an area of real space. A plurality of cameras with overlapping fields of view produce respective sequences of images of corresponding fields of view in the real space. In one embodiment, the system includes first image processors, including subject image recognition engines, receiving corresponding sequences of images from the plurality of cameras. The first image processors process images to identify subjects represented in the images in the corresponding sequences of images. The system includes second image processors, including background image recognition engines, receiving corresponding sequences of images from the plurality of cameras. The second image processors mask the identified subjects to generate masked images. Following this, the second image processors process the masked images to identify and classify background changes represented in the images in the corresponding sequences of images.

A system and method for providing computerized, knowledge-based medical diagnostic and treatment advice. The medical advice is provided to the general public over a telephone network. Two new authoring languages, interactive voice response and speech recognition are used to enable expert and general practitioner knowledge to be encoded for access by the public. “Meta” functions for time-density analysis of a number of factors regarding the number of medical complaints per unit of time are an integral part of the system. A semantic discrepancy evaluator routine along with a mental status examination are used to detect the consciousness level of a user of the system. A re-enter feature monitors the user's changing condition over time. A symptom severity analysis helps to respond to the changing conditions. System sensitivity factors may be changed at a global level or other levels to adjust the system advice as necessary.

A multimodal and real-time method for filtering sensitive content, receiving as input a digital video stream, the method including segmenting digital video into video fragments along the video timeline; extracting features containing significant information from the digital video input on sensitive media; reducing the semantic difference between each of the low-level video features, and the high-level sensitive concept; classifying the video fragments, generating a high-level label (positive or negative), with a confidence score for each fragment representation; performing high-level fusion to properly match the possible high-level labels and confidence scores for each fragment; and predicting the sensitive time by combining the labels of the fragments along the video timeline, indicating the moments when the content becomes sensitive.

The invention provides a system and method for automatically indexing and retrieving multimedia content. The method may include separating a multimedia data stream into audio, visual and text components, segmenting the audio, visual and text components based on semantic differences, identifying at least one target speaker using the audio and visual components, identifying a topic of the multimedia event using the segmented text and topic category models, generating a summary of the multimedia event based on the audio, visual and text components, the identified topic and the identified target speaker, and generating a multimedia description of the multimedia event based on the identified target speaker, the identified topic, and the generated summary.

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 provides a method for using a virtual machine to improve the performance in software-testing based on fault injection. Based on the EAI (environment-application interaction) model put forward by WENLIANG DU, Syracuse University (US), the fault injection is carried out on the interaction point of the application program and the environment thereof in the invention to disturb the environment and further test the software vulnerability, thus reducing the semantic difference between the injected fault and the actual fault, reducing the number of test cases and respectively achieving the two functional parts of the test tool at the host and guest of the virtual machine; and based on the shared files between the virtual machine and the guest, and the backup and recovery mechanism of the virtual machine, the invention can improve the robustness and flexibility of the test tool. The tool generated by the method comprises an application program configuration file (10), a fault test case generator (5), a security analysis module (6), a graphical user interface (7), an environment recovery module (9), a fault injector (11) and a data collector (12), wherein the tool further comprises software (13) to be tested and shared memory auxiliary modules (14), (15) and (16) of the virtual machine for sharing the data related to the tests. The method of the invention is capable of effectively detecting and simulating the security breaches in the software and improving the security of the software.

The invention discloses an entity relationship prediction method and a prediction system based on knowledge representation learning. The method comprises four modules, namely, knowledge preparation, knowledge representation model construction, knowledge representation model training and entity relationship prediction. The knowledge preparation module completes the data preparation and builds the knowledge map. The knowledge representation model construction module completes the construction of the model, which eliminates the semantic differences among different types of entities through projection operation; the training module of knowledge representation model forms the final knowledge representation model based on the parameters of the iterative training knowledge representation model ofknowledge map. Entity relationship prediction module can predict the possible relationship between any given two entities. The method of the invention predicts entity relationship based on knowledgemap, projects different types of entities to the same semantic space through a spatial projection algorithm, and performs calculation operation, thereby achieving high reliability of prediction results.

The invention provides a convolution neural network-based music recommending system and method. The system comprises a music user modeling module for collecting historical behavior data of a music user and constructing a preference model of the music user; a music feature modeling module for obtaining a regression model; and a recommendation algorithm module for finding music objects matched withthe preference of the music user through the regression model and recommending the music objects to the music user. According to the system provided by the invention, deep learning is applied to the recommending system, semantic differences between song features and audio signals are effectively compensated and the problems such as "cold start" and the like in collaborative filtering are avoided at the same time, so that the accuracy of the recommending system is increased; and the contradiction between low training efficiency and a high timeliness requirement is solved by adopting a convolution neural network and the historical behavior information of the user and audio acoustic features are added to the model, so that the recommendation results are more in line with the preference requirements of the user and the user experience of the recommending system is improved.

The invention discloses a Web service oriented home network system architecture which comprises a physical architecture and a logical architecture. The Web service oriented home network system architecture is characterized in that a home area network is constructed by the aid of IEEE (institute of electrical and electronic engineers) 802.3 and IEEE 802.11 protocols in the architecture, the problem of semantic difference is solved by the aid of a home gateway, the home gateway is divided into a broadband router and a central server when the home gateway is designed, a home gateway software system is deployed on the central server, a virtual equipment technology is adopted in the home gateway software system of the central server and provides an operation interface and an API (application program interface) for end users and third-party application programs, and design detail, implementation algorithm and coding operation for realizing functions is completed. The Web service oriented home network system architecture has the advantages that an equipment interconnecting function which is the most important in various functions of an existing home network is realized, the problem of semantic difference of equipment is solved, and the system architecture is sufficiently flexible and expandable.

The invention provides a method for selecting positive sample data in source domain data to extend training data of a target domain by fusing semantic difference and label difference of sentences in the source domain and the target domain, so as to achieve the purpose of enhancing named entity recognition performance of the target domain. Based on a conventional Bi-LSTM+CRF model, in order to fusesemantic differences and label differences of sentences in a source domain and a target domain, semantic difference and label difference are introduced through state representation and reward settingin reinforcement learning; therefore, the trained decision network can select sentences having positive influence on the named entity recognition performance of the target domain in the data of the source domain, expand the training data of the target domain, solve the problem of insufficient training data of the target domain, and improve the named entity recognition performance of the target domain at the same time.

The invention provides a system and method for automatically indexing and retrieving multimedia content. The method may include separating a multimedia data stream into audio, visual and text components, segmenting the audio, visual and text components based on semantic differences, identifying at least one target speaker using the audio and visual components, identifying a topic of the multimedia event using the segmented text and topic category models, generating a summary of the multimedia event based on the audio, visual and text components, the identified topic and the identified target speaker, and generating a multimedia description of the multimedia event based on the identified target speaker, the identified topic, and the generated summary.

The invention discloses a computer network defensive strategy conversion-oriented semantic similarity detection system, which comprises a defensive strategy configuration module, a strategy statement processing module, a node-link configuration module, a lexical and syntax analyzing module, a measure statement processing module, a structural similarity calculating module, a key concept pair matching module, a concept similarity calculating module, a CND strategy and measure body module and a similarity accumulation calculating module. A traditional symbol description-based strategy conversion system can only detect lexical and syntax errors before and after conversion, and hardly detects semantic inconsistency before and after strategy conversion comprehensively and automatically. By employing the semantic similarity processing method, the semantic similarity detection system automatically and effectively measures semantic difference before and after the computer network defensive strategy conversion, provides a basis for accurately deploying network defensive measures for semantics, is mainly applied to a computer network defensive system, deploys defensive measures according to a certain condition based on large-scale network attack, and rapidly and effectively fulfills the aim of large-scale defensive measure deployment.

The invention provides a language translation quality auditing method. The method comprises the following steps of: translating a translated parent language to form a translation comparison library based on numerous conventional reliable translation methods; respectively performing a semantic comparison which takes category identification as a base structure on each translation sentence and each target language to be audited in the translation comparison library with a reference sentence generated according to the parent language; calculating a corresponding semantic difference degree; judging whether the semantic difference degree between the target language and the reference sentence is in the range of the semantic difference degree between each translation sentence and the reference sentence in the translation comparison library and then judging whether the translation is reliable; and proving modification for mistranslation. The auditing method has high automation and standardization degrees and is an ideal translation quality auditing method.

The invention belongs to the technical field of image segmentation, and relates to a medical image segmentation method based on deep learning, which comprises the following steps: firstly, using a novel convolution module in the stages of an encoder and a decoder, and secondly, designing a residual bottleneck structure containing an attention mechanism, and using the residual bottleneck structurein skip layer connection, thereby reducing the semantic difference between the encoder and the decoder, and enabling the neural network to pay more attention to a target area to be segmented in the training process, so that finer semantic features can be extracted. The method is simple, the blurred boundary can be better recognized, the effect of the segmented image is more coherent, the noise interference resistance is high, and the generalization ability is high.

The invention discloses a video semantic segmentation method and device based on prediction for feature propagation, and the method comprises the steps: predicting the semantic difference of video frames according to a superficial neural network, and obtaining a plurality of key frames and a plurality of non-key frames in the video frames; Obtaining high-order semantic features of a plurality of key frames according to the picture semantic segmentation network, and predicting the high-order semantic features of a plurality of non-key frames according to sequential information of the high-ordersemantic features; And classifying the high-order semantic features of the plurality of key frames and the plurality of non-key high-order semantic features, and performing sampling to a preset sizeto generate a video semantic segmentation result. According to the method, hypotheses do not need to be made for high-order and low-order features, video semantic segmentation is obtained through prediction and fine adjustment, and the time complexity of the algorithm can be reduced on the premise that the video segmentation accuracy is guaranteed.

The invention discloses a method for mapping between a CIM (common information model) and a relational database. The method comprises the steps as follows: building meta-model descriptions of a class and an object respectively to remove semantic differences related with the class and the object from the meta-model systems of the class and the object; and mapping meta-models to a relational database so as to build the database tables of the two meta-models, namely a pattern relation table collection (called as a pattern library) formed by mapping the class meta-model and an object relation table collection (called as an object library) formed by mapping the object meta-model. The database built by the method has limited number of tables; and as the meta-models are universal, the situation means that the table struture is stable and the arbitrary expansion requirement and the referential integrity constraint on an infomation model in a CIM appliation can be met.

The invention discloses a CIM (common informatoin model) mapping method which comprises the steps as follows: building meta-model descriptions of a class and an object respectively to remove semantic differences related wtih the class and the object from the meta-model systems of the class and the object; and mapping meta-models to a relational database so as to build the database tables of the two meta-models, namely a pattern relation table collection (called as a pattern library) formed by mapping the class meta-model and an object relation table collection (called as an object library) formed by mapping the object meta-model. The database built by the method has limited number of tables; and as the meta-models are universal, the situation means that the table struture is stable and the arbitrary expansion requirement and the referential integrity constraint on an infomation model in a CIM appliation can be met.

The invention provides a method and system to manage and resolve “impedance mismatches” between and among standards and among each standard's attending messaging protocols. The invention provides a messaging service manager whose implementation accommodates the occurrence of “translations” between and among various messaging service providers. ServiceManager™, a software-based messaging services manager, provides for integration of various heterogeneous web services in a single environment both at the level of explicit conversational differences and implicit special case semantics. The invention provides places in its implementation for such “translations” to occur between and among various messaging service providers. ServiceManager™ addresses the need for a solution to the unexpected alteration to a request / response set providing places in its implementation for such “special case” situations to be defined and handled. The invention provides a messaging service manager whose implementation accommodates the definition and handling of such “special case” situations.

The invention provides a method for selecting positive sample data in source domain data to extend training data of a target domain by fusing semantic difference and label difference of sentences in the source domain and the target domain, so as to achieve the purpose of enhancing named entity recognition performance of the target domain. Based on a conventional Bi-LSTM + CRF model, in order to fuse semantic differences and label differences of sentences in a source domain and a target domain, semantic difference and label difference are introduced through state representation and reward setting in reinforcement learning; therefore, the trained decision network can select sentences having positive influence on the named entity recognition performance of the target domain in the data of thesource domain, expand the training data of the target domain, solve the problem of insufficient training data of the target domain, and improve the named entity recognition performance of the targetdomain at the same time.

The invention discloses a cross-modal retrieval method, system and device based on dual coding and combination. The method is a cross-modal retrieval algorithm based on dual coding and dual joint embedded learning. The method comprises the steps of extracting and encoding various features of a video through a neural network, performing multi-layer encoding on character features, learning and training two network models combined with video text embedding, and obtaining a text-to-video retrieval result or a video-to-text retrieval result through the two models. By using the method provided by the invention, the semantic difference between the video features and the text described by the natural language can be reduced, the potential information and the relationship between the video and thetext are captured, learned and optimized in a targeted and complementary manner, and finally, the retrieval accuracy between the video and the text is improved. The method can be widely applied to thetechnical field of video processing.

The invention provides a hardware Trojan horse detection method based on information flow security verification. The method includes constructing an information flow model for integrated circuit hardware design; defining information flow security attributes required to be met by hardware design; adopting an EDA formal verification tool for information flow safety verification, detecting potentialTrojans in hardware design by capturing the situation that information flow safety attributes violate, and achieving accurate positioning of Trojan design and triggering mechanism revealing of the Trojan design by means of the reverse tracking capacity provided by the information flow model. The method can effectively avoid additional design expenditure caused by semantic difference between the hardware design language and the verification language, and improves the efficiency of security verification and Trojan horse detection.

The invention provides a language description guided video timing sequence positioning method and system. The method comprises the following steps: receiving a video query information group; obtaininga target video clip from the to-be-queried video; extracting character feature information from the query character; extracting target feature information from the target video clip; calculating a target loss value corresponding to the target video clip; when the target loss value is not included in a preset loss value set, calculating an action parameter; and adjusting the boundary position of the target video clip in the to-be-queried video according to the action parameters, and returning to the step of extracting the feature information of the target video clip from the target video clip.According to the language description guided weak supervision video time sequence positioning method provided by the invention, a time sequence boundary is adaptively optimized by means of a reinforcement learning normal form according to a boundary adaptive optimization framework, a cross-modal semantic difference is reduced, and a more accurate result is obtained.

The invention provides an automobile taillight shape design evaluation and prediction method which comprises the following steps: collecting automobile taillight shape perceptual intention vocabularies oriented to consumers based on a semantic difference method and a Like scale; defining the shape category of the automobile taillight based on the similarity and proximity principle in the Grignardpsychology; establishing a stepwise regression equation by taking the shape category of the automobile taillight as an independent variable xj and taking the average value of the perceptual meaning words of the shape of the automobile taillight as a dependent variable; performing principal component analysis; and carrying out correlation analysis and supplement on the data obtained in the processes of the two assessment methods in the above steps, and providing an automobile taillight shape design strategy. The taillight design can be more comprehensively and effectively evaluated, and a taillight design strategy can be predicted and put forward.

A multimodal and real-time method for filtering sensitive content, receiving as input a digital video stream, the method including segmenting digital video into video fragments along the video timeline; extracting features containing significant information from the digital video input on sensitive media; reducing the semantic difference between each of the low-level video features, and the high-level sensitive concept; classifying the video fragments, generating a high-level label (positive or negative), with a confidence score for each fragment representation; performing high-level fusion to properly match the possible high-level labels and confidence scores for each fragment; and predicting the sensitive time by combining the labels of the fragments along the video timeline, indicating the moments when the content becomes sensitive.

The invention discloses a medical query expansion method based on a knowledge graph. According to the query expansion technology in the automatic question-answering system, the semantic difference between question-answering sentences is reduced by supplementing the expansion information into the question sentences, so that the accuracy of the question-answering system is improved. In the field of medical questions and answers, an existing query expansion method does not fully combine a co-occurrence association relationship and a reasoning association relationship among medical terms under different query intentions, so that obtained expansion words are not accurate enough. According to the method, the medical knowledge graph is taken as a knowledge source of the extension words, the candidate extension words are obtained by utilizing reasoning association of the medical terms under different query intentions, and the final extension words are screened in combination with negative medical term recognition and mutual information technologies, so that the accuracy of a medical question and answer system is finally improved.

A method is provided for role-oriented risk analysis in a contract. The method generates, using deep semantic association analysis, a report specifying a set of potential risks relating to explicit and hidden roles of contract parties. The generating step categorizes input statements of the contract into respective obligation / right pairs according to a deep semantic association distribution thereof. Each pair includes a respective obligation and a respective right. The generating step detects deep semantic differences between the respective pairs and a set of reference obligation / right pairs. The generating step identifies the explicit and hidden roles of the involved parties in the respective obligations / rights pairs according to domain-specific use scenarios and multidimensional local and global context clues in the contract. The generating step identifies the set of potential risks by applying a deep semantic role-oriented risk entailment model to the deep semantic differences.

The invention provides a street abnormal event detection method. The method comprises the steps: acquiring a target image and a reference image, wherein the street scene recorded by the reference image does not include abnormal event; inputting the target image and the reference image to a semantic difference extraction network to obtain a semantic difference region of the target image relative to the reference image; and acquiring a detection result according to the semantic difference region, wherein the detection result is used for representing whether the street scene recorded by the target image comprises an abnormal event or not. According to the method, the detection accuracy is improved, the false alarm rate is reduced, and the user experience is improved.

The invention discloses a product appearance knowledge base construction method, which comprises the following steps: collecting and analyzing perceptual intention contours and related product cases of product appearances according to a semantic difference method in perceptual engineering, scoring the perceptual intention contours of products, and extracting representative perceptual intention vocabularies as expected functions through a factor analysis method; clustering and analyzing appearance characteristics of related products as expected structures according to corresponding sample casescores in the public factors; and based on the score condition of the CMF variable element of the case product in the public factor, mapping a relationship between an expected function and an expectedstructure, and constructing a knowledge base system of the product. The product appearance knowledge base construction method has the advantages of improving the design efficiency, guiding the product appearance design, and showing the superiority of the design method.