368 results about "Semantic annotation" patented technology

System and method for the indexing and retrieval of semantically annotated information units from a collection of semantically annotated indexed information units in response to a query using an ontology-based IR model. The retrieval method comprises: receiving a semantically annotated query with semantic annotations to individuals or classes within a determined populated base ontology; embedding, as a set of weighted-mentions to individuals or classes within the populated base ontology, the semantically annotated query in a semantic representation space of an ontology-based metric space IR model; obtaining the representation in the semantic representation space for every indexed information unit; computing the Hausdorff distance between the space representation of the query and the space representation of all the indexed information units of the collection; retrieving and ranking, the relevant information units based on the computed Hausdorff distance.

The invention discloses a multi-modal information fusion football video event detection and semantic annotation method. The multi-modal information fusion football video event detection and semantic annotation method includes the steps of detecting the event type of Internet match result report text description statements with the potential semantic analytical method; detecting football video intermediate semantic objects, dividing a site area, conducting attack-and-defense transition analysis, and determining boundaries of video event fragments; determining the match starting time according to kick-off circle and whistling detection results, and achieving initial semantic classification of attack-and-defense fragments with the Bayesian network; under the constraint of coarse-grained time information in text descriptions, achieving the football video event semantic annotation according to semantic synchronization text descriptions and video events of texts and the video fragments. By means of the method, the Internet text information and video inherent audio-visual feature analysis are fused for analyzing football videos, accuracy for detecting the video events and the boundaries of the video events is improved, the rich semantic annotation of football video contents is achieved, and a solid foundation is laid for building a video indexing mechanism based on semantics.

The invention discloses a real-time high-performance street-view image semantic segmentation method based on deep learning. The real-time high-performance street-view image semantic segmentation method includes the steps: preparing a street-view image training, verifying and testing data set; carrying out downsampling on images of the data set to reduce the resolution of the images; transforming an existing lightweight classification network to serve as a basic feature extraction network of semantic segmentation; connecting identification hole space pyramid pooling in series after the basic feature extraction network for solving the multi-scale problem of semantic segmentation; stacking a plurality of convolutional layers to form a shallow spatial information storage network; fusing the obtained feature maps by using a feature fusion network to form a prediction result; comparing the output image with a semantic annotation image in the data set, and performing end-to-end training by using a back propagation algorithm to obtain a real-time high-performance street-view image semantic segmentation network model; and inputting the street-view image to be tested into the real-time high-performance street-view image semantic segmentation network model to obtain a semantic segmentation result of the street-view image.

The invention provides a robot semantic SLAM method based on object instance matching, a processor and a robot. The robot semantic SLAM method comprises the steps that acquring an image sequence shotin the operation process of a robot, and conducting feature point extraction, matching and tracking on each frame of image to estimate camera motion; extracting a key frame, performing instance segmentation on the key frame, and obtaining all object instances in each frame of key frame; carrying out feature point extraction on the key frame and calculating feature point descriptors, carrying outfeature extraction and coding on all object instances in the key frame to calculate feature description vectors of the instances, and obtaining instance three-dimensional point clouds at the same time; carrying out feature point matching and instance matching on the feature points and the object instances between the adjacent key frames; and performing local nonlinear optimization on the pose estimation result of the SLAM by fusing the feature point matching and the instance matching to obtain a key frame carrying object instance semantic annotation information, and mapping the key frame intothe instance three-dimensional point cloud to construct a three-dimensional semantic map.

The invention discloses a semantic annotation method for a hyperspectral remote sensing image. The semantic annotation method comprises the following steps of: I, acquiring training data and test data of the hyperspectral remote sensing image through spectral information and an annotated truth value of the hyperspectral remote sensing image; II, constructing a convolutional neural network according to the number of bands of the hyperspectral remote sensing image; III, training the convolutional neural network through the training data to obtain a convolutional neural network model; IV, classifying the test data through the convolutional neural network model to obtain a semantic annotation result; V, constructing a unary potential-energy function of a conditional random field model according to the semantic annotation result; VI, constructing a binary potential-energy function of the conditional random field model in a neighborhood by using an edge constraint model based on an improved mahalanobis distance; VII, carrying out weight adjustment of the unary potential-energy function and the binary potential-energy function on the conditional random field model; VIII, solving the conditional random field model to obtain the semantic annotation result. Through the above steps, the semantic annotation method for the hyperspectral remote sensing image is realized.

An apparatus and a method for automated semantic annotation of a process model having model elements named by natural language expressions, wherein said apparatus comprises at least one semantic pattern analyser which analyses the textual structure of each natural language expression on the basis of predefined semantic pattern descriptions to establish a semantic linkage between each model element to classes and instances of a reference process ontology for generating a semantically annotated process model.

The invention provides a mass video semantic annotation method based on Spark. The method is mainly based on elastic distributed storage of mass video under a Hadoop big data cluster environment and adopts a Spark computation mode to conduct video annotation. The method mainly comprises the following contents: a video segmentation method based on a fractal theory and realization thereof on Spark; a video feature extraction method based on Spark and a visual word forming method based on a meta-learning strategy; a video annotation generation method based on Spark. Compared with the traditional single machine computation, parallel computation or distributed computation, the mass video semantic annotation method based on Spark can improve the computation speed by more than a hundred times and has the advantages of complete annotation content information, low error rate and the like.

The invention provides an image classification method, an image classification device, an image retrieval method and an image retrieval device, wherein the image classification method concretely comprises the following steps that physical characteristics of images to be classified are extracted; the images to be classified are subjected to semantic annotation to obtain corresponding annotation words; by aiming at the annotation words of the images to be classified, the annotation words are matched with semantic words in a semantic network, and in addition, semantic characterization polybasic groups are generated according to the semantic characterization corresponding to the successfully matched semantic words; semantic words and a plurality of corresponding semantic characterizations are stored in the semantic network, and the semantic characterizations are described by physical characteristics; and characteristic vectors consisting of the physical characteristics of the images to be classified and the semantic characterization polybasic groups are input into an image classifier, and corresponding classification results are output, wherein the image classifier is a classifier obtained according to the image sample training under each image type, and the physical characteristics and the arity of the semantic characterization polybasic groups are identical in the training and classification process. The methods and the devices provided by the invention have the advantage that the image classification accuracy can be improved.

The invention discloses an understanding method of a non-parametric RGB-D scene based on a probabilistic graphical model. The method comprises the steps of carrying out global feature matching between a marked image and an image marked in a training seat, and building a retrieval set of a similar image of an image to be marked; cutting the image to be marked and the image in the similar image retrieval set, so as to generate super-pixels, and carrying out characteristic extraction on the super-pixels extracted; calculating the proportions of all categories in the training set, building a dictionary of rare categories, and taking the training set and the retrieval set of the similar images as a label source of the image to be marked; carrying out characteristics matching on each super-pixel of the image to be marked and all super-pixels in an image label source; and building a probabilistic graphical model, converting the maximum posterior probability into a minimal energy function by using a Markov random field, and resolving the semantic annotation of each super-pixel of the image to be marked obtained by solving the problem through a graph cutting method. According to the method provided by the invention, the overall and local geometric information can be integrated, and the understanding performance of the RGB-D scene can be improved.

The invention discloses an automatic image annotation method based on deep learning and canonical correlation analysis. The method includes: using a depth Boltzmann machine to extract the high-level feature vectors of images and annotation words, selecting multiple Bernoulli distribution to fit annotation word samples, and selecting Gaussian distribution to fit image features; performing canonical correlation analysis on the high-level features of the images and the annotation words; calculating the Mahalanobis distance between to-be-annotated images and training set images in canonical variable space, and performing weighted calculation according to the distance to obtain high-level annotation word features; generating image annotation words through mean field estimation. The depth Boltzmann machine comprises I-DBM and T-DBM which are respectively used for extracting the high-level feature vectors of the images and the annotation words. Each of the I-DBM and the T-DBM sequentially comprises a visible layer, a first hidden unit layer and a second hidden unit layer from bottom to top. By the method, the problem of 'semantic gap' during image semantic annotation can be solved effectively, and annotation accuracy is increased.

The invention provides a method and device for constructing a legal text syntax tree based on a deep neutral network. The method comprises the following steps: acquiring a training text of the legal service; performing word vector sparse representation on the training text; performing semantic annotation on the training text through the DNN according to the word vector sparse representation; performing dependency description on the training text according to the semantic annotation to construct the legal text syntax tree. The invention provides a systematic method to construct the syntax tree about technical language understanding in the legal service field, and the global-local training is used for solving the semantic bias problem in the legal language understanding.

The invention belongs to the technical field of intelligent education question-answer, and particularly relates to a mixed automatic question-answer method based on education knowledge graphs and texts. The mixed automatic question-answer method comprises the following steps: constructing a basic education knowledge graph by constructing a basic education body, semantic annotation and informationextraction; constructing a general template of the question according to the keywords in combination with a regular expression; establishing a full-text search engine, and preprocessing mass texts; training test question and answer pairs as a training set until a deep text matching model converges; identifying the user questions to obtain a subject list, and endowing the subject list with confidence; carrying out template matching to obtain a predicate list, and giving confidence to the predicate list; inquiring a knowledge graph according to the subject and predicate lists to obtain an answerlist, and endowing confidence coefficients; obtaining keywords by using a part-of-speech tagging method, performing coarse-fine granularity matching to obtain answers, and sorting the answers; if thehighest confidence of the answer based on the educational knowledge graph exceeds a threshold value, returning the answer; or, returning the answer with the highest sorting based on the text.

The invention relates to a structure feature based on Chinese painting and calligraphy seal image automatic extracting method, comprising the steps as follows: 1) chromatography of an input image; 2) image segmentation based on the connected region; 3) image region filtration based on the seal structure feature; and 4) seal image extraction in the input image. In the method, Chinese painting and calligraphy chromatography analysis model based on machine learning is adopted to realize the image segmentation of painting and calligraphy works based on granularity level of the homogeneous region of the color layers. On this basis, a painting and calligraphy seal image extracting method based on seal outline template and internal structure complexity is provided, thereby realizing a content-based Chinese painting and calligraphy image retrieval and semantic annotation system with the seal image as the key information so as to lay a foundation for the retrieval, appraisement, classification and the like of the Chinese painting and calligraphy works.

The invention discloses an intelligent interaction method and device, and a storage medium. The intelligent interaction method includes the steps: according to at least one of the position of a user,the business scene and the language habits of the user, performing word segmentation on the information input by the user to obtain a plurality of words; after respectively performing semantic annotation analysis on the plurality of words, identifying at least one keyword; according to the importance and/or semantics corresponding to the position of the at least one keyword in the input information, determining the weight and/or semantic annotation of the at least one keyword; according to the determination result, performing new semantic analysis on the input information, and obtaining the semantic result; and inquiring the information matching the semantic result, and returning the inquiry result to the user. The intelligent interaction method can improve accuracy of semantic identification so as to improve reliability of intelligent interaction.

One feature of the present invention uses the parsing capabilities of a structured language model in the information extraction process. During training, the structured language model is first initialized with syntactically annotated training data. The model is then trained by generating parses on semantically annotated training data enforcing annotated constituent boundaries. The syntactic labels in the parse trees generated by the parser are then replaced with joint syntactic and semantic labels. The model is then trained by generating parses on the semantically annotated training data enforcing the semantic tags or labels found in the training data. The trained model can then be used to extract information from test data using the parses generated by the model.

The invention provides an image semantic annotation method based on super pixel segmentation. The method comprises the steps that a feature block extracted based on image super-pixel segmentation is input into a convolution neural network; a feature vector trained by the convolution neural network is expanded and weighted; and finally, a conditional random field model is constructed for semantic class annotation prediction. According to the technical scheme of the invention, the method takes a super-pixel block as a research object, which simplifies the complexity of the feature block extracted based on image super-pixel segmentation, and improves the calculation efficiency of semantic annotation; in addition, the multi-layer super-pixel block is used for semantic analysis, and the annotation results are integrated; and the accuracy and the robustness of semantic annotation are improved.

The invention relates to an automatic image semantic annotation method based on scale learning and correlated label dissemination, which comprises the following steps: firstly, the global and partial feature descriptor of each image is extracted after the image library is read; the feature descriptor is sent to a model based on a structured support vector machine for learning the distance scale between the images, actually the Mahalanobis distance; a model about the internal relation between key words is built; the learned Mahalanobis distance is embedded in a built label dissemination model so as to obtain the confidence degree score of each key word belonging to the image to be labeled; and a threshold value is set for the confidence degree score of each key word, and the key words of which the scores are higher than the threshold value are distributed to the images to be labeled, thereby completing labeling. The learning algorithm model based on the structured support vector machine can effectively solve the measuring problem of similarity between the images, the internal relation between the key words is fully excavated through the embedded-type correlated label dissemination model, and the accuracy of the image annotation and image retrieval is effectively improved.

The invention discloses an image marking method based on a multi-view and semi-supervised learning mechanism. The method comprises the following steps that a multi-view classifier learning process based on irrelevant view characteristics is carried out; a multi-view classifier optimization process based on marking samples and pseudo marking samples with higher reliability is carried out; a multi-view marking process based on a maximum entropy voting principle and the marking correlation is carried out. The performance of the multi-view and semi-supervised image marking method provided by the invention is obviously superior to the performance of other schemes, the method has the main ideal that firstly, mutually irrelevant views are utilized for training a plurality of independent classifiers, then, initial marking samples and pseudo marking samples are utilized for optimizing the view classifier, and finally, proper semantic annotation is allocated to each unmarked image on the basis of the maximum entropy voting principle and the correlation between all marks.

The invention relates to an automatic driving scene key target detection and extraction method based on deep learning, and is characterized in that the method carries out key target detection, semantic annotation and target frame selection in a driving scene based on an improved single multi-frame detector, and the training process of the improved single multi-frame detector specifically comprisesthe steps: 1) obtaining a training set; 2) generating a prior frame, and matching the prior frame with a real target with class labels and frame labels of corresponding pictures to form positive andnegative samples; And 3) iteratively updating the model parameters of the single multi-frame detector by utilizing the loss function, wherein the iterative updating is realized based on momentum optimization gradient updating. Compared with the prior art, the method has the advantages of high detection precision, capability of detecting a small target and the like.

An adaptive dialogue system and also a computer-implemented method for semantic training of a dialogue system are disclosed. In this connection, semantic annotations are generated automatically on the basis of received speech inputs, the semantic annotations being intended for controlling instruments or for communication with a user. For this purpose, at least one speech input is received in the course of an interaction with a user. A sense content of the speech input is registered and appraised, by the speech input being classified on the basis of a trainable semantic model, in order to make a semantic annotation available for the speech input. Further user information connected with the speech input is taken into account if the registered sense content is appraised erroneously, incompletely and/or as untrustworthy. The sense content of the speech input is learned automatically on the basis of the additional user information.

The invention discloses a semantic comprehension training method and system, and electronic equipment. The method comprises the following steps of carrying out semantic annotation on the correct textof user voice data to generate the correct text with the semantic annotation; inputting the user voice data into a voice identification system for identification so as to obtain an identification text; and randomly inputting the correct text, the correct text with the semantic annotation, and the identification text into a semantic comprehension system so as to carry out unsupervised adaptive training on the semantic comprehension system. In the invention, the correct text is annotated and the identification text does not need to be annotated; and through carrying out unsupervised adaptive learning on the correct text, the correct text with the semantic annotation and the identification text without annotation, the semantic comprehension system which is robust to a voice identification error can be trained and acquired.

Data elements from data sources and having a data value set are linked by using hash functions to determine a dimensionally reduced instance signature for each data element based on all data values associated with that data element to yield a plurality of dimensionally reduced instance signatures of equivalent fixed size such that similarities among the data values in the data value sets across all data elements is maintained among the plurality of instance signatures. Candidate pairs of data elements to link are identified using the plurality of instance signatures in locality sensitive hash functions, and a similarity index is generated for each candidate pair using a pre-determined measure of similarity. Candidate pairs of data elements having a similarity index above a given threshold are linked.