70 results about "WordNet" patented technology

The invention belongs to the field of multimedia information retrieval and relates to a method for realizing thesaurus-based query expansion and sort in image retrieval. The method comprises a WordNet-based English word semantic similarity metric algorithm, a HowNet-based Chinese word semantic similarity metric algorithm, an expansion rule-based query expansion word selection and optimization algorithm and a retrieval result evaluation and optimization algorithm. In the method, an image search engine is improved by the relevant text processing method and the relevant semantic network dictionary; and the retrieval result is sorted through semantic expansion, user interaction and improved similarity measurement. Compared with the traditional method, the method has the advantages of high accuracy rate, high integrality and low space-time cost. The method has very important significance for performing high-efficiency image retrieval according to image high-layer semantic information and on the basis of a large-scale image data set, and has wide application value in the field of cross-linguistic and cross-media retrieval.

The invention relates to a microblog-oriented emotion entity searching system. The emotion entity searching system comprises a user interface (1), a query expansion module (2), a query processing module (3), an emotive information mining module (4), an emotive information judging and index building module (5) and a reverse index building module (6). The user interface (1) is used for interaction between a user and the system, and the user can submit a query request through the user interface and obtain a feedback result; the query expansion module (2) is used for carrying out word relation mining on microblog corpus data and building a weighting word relation graph in combination with a WordNet ontology base; the query processing module (3) is used for converting the query request of the user into query key words or query statements and for carrying out query expansion on the basis of the word relation graph built by the query expansion module (2), wherein the query key words or the query statements can be accepted by an index base; the emotive information mining module (4) is used for performing emotion mining on the microblog corpus base and generating a judging rule for emotion entities and emotion polarities; the emotive information judging and index building module (5) is used for judging the emotion entities and emotion polarities, building an emotive information index and storing the emotive information index; the reverse index building module (6) is used for building a reverse index for microblog text information and storing the reverse index. The microblog-oriented emotion entity searching system solves the problems that difficulty exists in microblog emotion entity extraction, emotion polarity analysis, emotion entity search and the like, and a novel intelligent searching product is provided for analyzing and monitoring social networking public opinions.

The invention relates to the field of ontology learning, in particular to an ontology concept and hierarchical relation generation method. The technical scheme includes that a PAM (probability aim model) is applied to extraction of ontology concept and hierarchical relation to improve effect of ontology concept and hierarchical relation learning so as to generate ontology concept more accurately and effectively. The method includes: establishing a PAM-based ontology concept and hierarchy generation model to effectively convert a domain ontology concept learning problem into a domain document set based statistic inference problem, and adopting a Gibbs sampling method to obtain a probability distribution feature vector; performing Wordnet-based semantic similarity calculation, generating a conception according to similarity correlation, and obtaining a set and hierarchical relation of ontology concepts. By the method, concept sets and hierarchical relation among concepts in domain ontology can be obtained more accurately and effectively.

The invention relates to a semantic fuzziness searchable encryption method capable of realizing sorting verifiability. The method comprises the following steps that: firstly, importing a relevancy score and a domain weighting score to obtain a more accurate index structure; then, on the basis of a WordNet dictionary set, carrying out semantic extension on query keywords, integrating semantic similarity with an encryption score to design a dual-factor sorting algorithm, and carrying out accurate sorting on a search result; and finally, applying a Bloom filter and a message authentication code to carry out comprehensive and effective verification on keywords, indexes and ciphertexts. By use of the method, semantic fuzziness search can be realized, search results can be sorted, and a sorting verifiability function is supplemented on the basis of the accuracy and the integrity of a verification result of an existing verifiable scheme.

The invention provides a semantic-based three-dimensional model retrieval system and method. The system comprises a semantic retrieval module and a semantic database. The semantic retrieval module is used to segment and filter semanteme of a retrieved sentence so as to obtain a semantic concept. The method comprises the steps of inquiring the synonymous semantic concept in the WordNet according to the semantic concept, calculating the similarity between the semantic concept and the inquired semantic concept and screening the semantic concept according to the similarity to obtain other semantic concepts associated with the semantic concept; and combining the semantic concept with other associated semantic concepts to retrieve corresponding three-dimensional information in the semantic database to obtain the retrieval result. According to the three-dimensional model retrieval system and method provided by the invention, the semantic gap of high-level semantic information and low-level semantic information is complemented, the retrieval range of the three-dimensional model is expanded, and the retrieval precision is improved.

The invention relates to a multi-feature image tag sorting method based on WordNet semantic similarity. The multi-feature image tag sorting method includes the steps of establishing and training a specimen bank, extracting significant region graphs of images in the specimen bank, training an SVM classifier, preprocessing image tags to be tested, judging the types of the images to be tested, and sorting the image tags to be tested. The multi-feature image tag sorting method integrates correlation, visuality, multi-feature and the like, different features of an entire scene image are considered, and different features of a saliency map of an object image are also considered. Before the image tags are sorted, improvement is conducted on the problems such as incorrectness of the image tags and non-comprehensiveness of the tags, and the correlation between the image tags and image content and the correctness and the comprehensiveness of the image tags are improved; according to the multi-feature image tag sorting method, the similarity between visual features of the images is considered, and the semantic similarity between tag texts is also considered, so that sorting of the image tags is more accurate.

The invention discloses a tag recommendation method capable of fusing multi-information-source coupling tensor decomposition. The method comprises the following steps that: firstly, while a tensor constructed by tag-item-user triple information is subjected to CP (Canonical Polyadic) decomposition, adding three auxiliary information matrixes, including tag and tag, tag and item and tag and user, to participate in joint decomposition, considering a cooccurrence relationship among tags and the semantic similarity of the tag in WordNet while the tag and the tag similarity matrix are constructed,and taking two similar linear sets as final similarity measurement among tags; then, after a problematic loss function is constructed, adopting an ADMM (Alternating Direction Method of Multipliers) algorithm to carry out parameter optimization on the target function; and finally, according to a decomposition completion prediction tensor, more accurately recommending a Top-N tag to the (user, item). By use of the method, the isomeric information of the tag is fused, and the method is applied to each socialized annotation system and exhibits universality.

The invention relates to a topic phrase extraction method. The topic phrase extraction method includes preprocessing documents, seeking a document-topic set, a full text lexical chain set and a noun phrase set, seeking a central word set, seeking a candidate topic phrase set, and seeking a topic phrase set. The topic phrase extraction method has the advantages that topic phrases are extracted through combination between an LDA (latent Dirichlet allocation) model and a lexical chain, a knowledge base WordNet with complete semantic information outside a corpus can be utilized, a strong lexical chain can be acquired through semantic relevance calculation and strong chain rule filtration, and accordingly, the ambiguity of topic words is reduced greatly; the topic phrases are extracted according a central word extraction method and by N-P rule combination and deduplication steps, and topics are expressed by the topic phrases with rich semantic information, so that the problems such as low granularity and recognition degree of the topic words are solved, topic extraction accuracy and recall rate can be guaranteed, topic drifting is reduced, and needs of practical applications can be wellmet.

The invention discloses a construction and instantiation method of the domain ontology. The construction and instantiation method of the domain ontology is characterized in that a domain is selected, and core items of the domain and the relation among the core items are obtained through Wikipedia; comparison and screening are carried out on the relation among the core items and the WordNet inquiry mode, a corresponding mode between the relation among the core items and the WordNet inquiry mode is confirmed, and the ontology relation of the domain ontology needing to be constructed is confirmed according to the corresponding mode; the domain ontology of the domain is constructed through the core items and the ontology relation; properties of the selected domain are confirmed and added to the domain ontology; an example of the selected domain is used as an entity of the domain ontology, and instantiation is carried out; the instantiated domain ontology is stored in an AG database. The construction and instantiation method of the domain ontology constructs the ontology by relying on Wikipedia and WordNet resources shared by all mankind, and is suitable for construction of most of specific domain ontology, and construction speed is quite high. The construction and instantiation method of the domain ontology enables the domain ontology to be instantiated, and enables the constructed domain ontology to be strong in readability, easy to maintain, and more convenient to use.

The invention discloses an image description method, and the method comprises the steps: extracting global image features in a picture through a VGG convolution neural network, and extracting local image features in picture by using Faster R-CNN network, fusing global image features and local image features through a global-local feature fusion algorithm to acquire the image fusion features; processing image fusion features through a bidirectional long-short-term memory network with an attention mechanism; and generating a preliminary image description statement, performing WordNet word vectorsimilarity calculation based on the image target information obtained during local image feature extraction and nouns in the preliminary image description statement, correcting the image descriptionstatement, and generating a final image description statement. According to the method, the influence of useless information is reduced, the expression of key information is enhanced, the fault tolerance and generalization ability of the model are enhanced, and the accuracy of statement description is improved.

The invention discloses a program readability analysis method based on a WordNet, and belongs to the field of software engineering. The program readability analysis method based on the WordNet comprises the following steps that (1) each method body is traversed according to a package name-class name-method name project structure, and is labeled in a package name-class name-method name mode; (2) warning annotations and TODO annotations are removed, wherein each warning annotation usually appears in a testing case and is used for warning a programmer that a certain serious consequence can happen, each TODO annotation indicates a piece of work, and the programmer thinks that the work should be done, but the work is not done at present for some reasons; (3) the annotations are divided into short annotations and long annotations; (4) the long annotations are preprocessed, and subject terms of the long annotations are extracted through the WordNet; (5) subject terms of codes are extracted from code segments mapped by the annotations through the WordNet; (6) a program readability analysis tree is generated for a whole project, and final analysis data are formed. According to the program readability analysis method based on the WordNet, a later-period maintainer can conveniently comprehend an unfamiliar project, and the program readability analysis method based on the WordNet can be used for maintenance and development of software.

A text classification method based on WordNet and latent semantic analysis relates to the field of a computer. The text classification method based on WordNet and latent semantic analysis considers synonyms, hypernyms and hyponyms of words in a text and word frequencies of the synonyms, the hypernyms and the hyponyms are increased according to the similarity, so that influence of synonymy of a plurality of words on classification is reduced. Different from a common method of carrying out feature extraction on a feature matrix by a single method, the text classification method based on WordNet and latent semantic analysis obtains a plurality of feature matrices by regulating a WordNet invocation parameters and uses a genetic algorithm (GA) to assist latent semantic analysis (LSA) to complete feature extraction together so as to obtain better feature matrices, thereby improving a classification effect.

The invention discloses an information gain-based English social media account number classification method. The method comprises the following steps of: S1, carrying out data preprocessing to obtain feature words of an account number; S2, feature selection: selecting the feature words of the account number by utilizing an information gain method so as to obtain a feature word with class representativeness; S3, feature diffusion: finding a synonym of the feature word by utilizing wordnet and artificially adding certain keywords in a field category to diffuse the feature word; S4, classification model construction: carrying out processing by adoption of a machine learning technology so as to an account number classification model; and S5, classifying unknown social media account numbers. According to the method, a common text classification method is applied to English social media account number classification, so that users can rapidly search account numbers in a certain field category from massive accounts and then obtain related effective information of the field category.

The invention discloses a conversational entity search-oriented query expansion method. The method comprises the steps of preprocessing a Wikipedia document set to obtain a data set D; traversing eachdocument d in the data set D to obtain a keyword reverse index Ik, an entity reverse index Ie and entity class mapping IE; extracting semantic information in WordNet and entity related information inYAGO3; according to a historical query result, constructing a historical result cache queue L; according to a current query Q of a user, retrieving Ik, Ie and IE to obtain pseudo-relevance feedback QPRF, and analyzing the historical result in L by utilizing the semantic information and the entity related information to obtain historical feedback Qh; based on the current query Q, the pseudo-relevance feedback QPRF and the historical feedback Qh, generating an extended query QE, and re-retrieving and sorting Ik, Ie and IE by utilizing QE to obtain a final result set R; and updating the cache queue L by utilizing R. The query process is shortened while the query overhead is reduced.

The invention discloses a continuous speech recognition system based on a wordnet language model. The system comprises a speech input module, a feature extraction module, an acoustic model layer search module, a lexical layer search module and a syntax layer search module. The input end of an acoustic model layer is connected with an acoustic model. The input end of the lexical layer search module is connected with a lexical information library. The input end of the syntax layer search module is connected with a syntax information library. According to the invention, a speech recognition process is divided into the acoustic model layer, a lexical layer and a syntax layer; sub words are searched through the acoustic model layer to acquire a candidate sub word sequence; entries are searched according to the lexical information and the language model of words on the lexical layer to acquire a candidate entry sequence; syntax layer searching is carried out according to grammar, semantic information and the language model of other sentences to acquire a final recognition result; and a speech signal is processed step by step in the manner, which is helpful to improve the recognition accuracy and the recognition efficiency of the speech signal.

The invention discloses an entity relationship classification method of an unstructured text based on WordNet and IDF. The method comprises the following specific steps of (1) obtaining a text training set and performing preprocessing to obtain sentence matrix representation; (2) expanding external semantic information of sentences by utilizing WordNet; (3) expanding internal semantic informationof the sentences by utilizing IDF; (4) calculating position information of words in the sentences, and updating matrix representation of the sentences; and (5) inputting a sentence matrix in the step(4) into a segmented convolutional neural network to obtain feature vectors of the sentences; (6) inputting the feature vectors into a classifier, and calculating a loss function; (7) if the accuracyof the current round of training is improved by more than 0.1% compared with the accuracy of the previous round or reaches an upper limit of iteration, finishing the training of the classification method; otherwise, updating hyper-parameters in the step (5), and continuing the training process. According to the method, semantic features of entities and relations can be accurately expressed, so that the problem of overlarge noise of training set data is relieved, and the classification accuracy is improved.

The invention belongs to the field of computer cryptography, in particular to a fuzzy multi-keyword searchable encryption method based on a synonym set. The technical scheme includes: organizing and coding the synonym set WordNet, which makes the keywords with similar semantics have similar coding, and can calculate the similarity degree of the keywords from the vector; Fuzzy multi-keyword searchable encryption algorithm, through which multiple keywords can be queried at one time and files with similar meanings to keywords can be queried. The method solves the problem that when the data is stored on the cloud server in the form of ciphertext, the powerful computing power of the cloud server is utilized for keyword retrieval, and the privacy of any user is not disclosed to the server. Thisnot only makes the privacy of users be effectively protected, but also greatly improves the retrieval efficiency with the help of the server.

The invention relates to the technical field of mode recognition, in particular to a small sample image classification method based on a component supervision network, and aims to solve the problem that feature extraction is not adaptive in the field of small sample image classification at present. According to the small sample image classification method based on the component supervision network, a new scheme, namely the component supervision network, for improving small sample image classification performance is provided, component information of samples is collected and multiple labels are generated by referring to a hierarchical dictionary WordNet commonly used in natural language processing, and an auxiliary task based on component supervision is constructed; a standard classification task, a component supervision auxiliary task and a self-supervision auxiliary task are used for assisting in training a feature extractor, the trained feature extractor is used for extracting features of new-class data, and a linear classifier is used for completing the final classification task so as to improve the adaptability of the feature extractor.

The invention discloses a statement similarity judgment method and judgment system, and relates to the technical field of natural language semantic similarity calculation. Improvements are performed on a modeling layer, a multi-semantic embedding layer, a semantic importance calculation layer, a semantic alignment layer and an output layer, a multi-granularity level similarity matrix is calculatedby utilizing the multi-semantic matrix, and true semantic alignment of the two sentences is discovered according to the matrix. The fact that different semantics have different importance is considered, and semantic importance calculation is provided. The proposed model does not need sparse features, WordNet and other external resources, is successfully trained in a short time, and obtains a competitive result on a similarity calculation task. Visual analysis shows that the model has good performance and interpretability.

The invention provides a word vector correction method based on semantic relation constraint and a computing system. The method comprises the following steps: a synonym and antonym constraint set with a symmetric relation and a direct hypernym / hypernym constraint set with an asymmetric relation are extracted as an external knowledge source update word embedding vector from WordNet and Roget semantic dictionaries; and the similarity of the pair of vocabularies can be calculated by randomly inputting two vocabularies of which the similarity needs to be calculated. The vocabulary semantic similarity calculation system comprises an input unit, an initialization unit, a calculation unit and an output unit. According to the method, the existing neural network word embedding vector is updated based on the lexical semantic relationship constraint provided by the external knowledge source so as to be used for lexical semantic similarity calculation. The updating speed and the word embedding vector semantic updating effect are obviously better than those in the prior art, and the accuracy rate in calculating the word semantic similarity is higher.

The invention provides a cross-linguistic plagiarism detection method based on multiple features. The method comprises the steps of 1, corpus building; 2, translation feature building, wherein according to the europeanized phenomenon and the translation body problem which generally occur in translated articles, translation feature building is conducted, by means of feature selection, the featuresare cleaned and filtered to obtain the effective features, and noneffective features or the features with unapparent effects are filtered out; 3, feature selection, wherein the effective features areselected from the multiple features for classifier training, and then whether or not the cross-linguistic plagiarism problem exists in a certain article or multiple articles is classified; 4, based onplagiarism detection corresponding to the features, for Chinese features, accurate English feature corresponding is conducted, and according to the translation features and the structural features, plagiarism results are correspondingly filtered and generated, and through WordNet, final confirmation is conducted on the plagiarism results. By means of the method, the cross-linguistic plagiarism problem can be solved according to the multiple kinds of features mined from translation.

The invention discloses a structured functional knowledge completion method for implicit conflict detection. The method comprises the following steps: firstly, obtaining paraphrases in an online dictionary according to a given group of implicit attribute words; for each implicit attribute word Ii belongs to I, obtaining a part-of-speech label of a paraphrasing sentence in the online dictionary through a lexical parser, and taking all nouns as candidate attribute words Ci of Ii; dividing all words cij belonging to Ci in the candidate attribute words into two classes, one class being attributesof implicit attribute words, the other class being non-attribute words, obtaining an accurate classification result through an iterative classification algorithm by utilizing an inter-word relationship in WordNet, and finally complementing missing attributes in the structured functional knowledge. The extracted structured functional knowledge can be used for implicit conflict judgment in the smarthome system, the automation degree of the implicit conflict detection system is improved, and the detection accuracy is greatly improved.

The invention provides an application method of a knowledge graph in zero-order learning. The method comprises the following steps: firstly, training by utilizing a ResNet deep convolutional neural network model to obtain visual features of an image; then, utilizing a wordnet knowledge graph to construct a relational graph between categories; calculating the weight relationship according to the distance between the nodes; then, optimizing nodes in the relation graph by utilizing a GraphSAGE (Graphical Spatial Absorption Gradient Error) algorithm; mapping the semantic features of the optimizedclass nodes to a dimension space which is the same as the visual features by using a graph convolutional neural network; and finally, searching a category closest to the Euclidean distance of the visual features, and taking the category as a judged category. According to the method, the knowledge graph is used in the zero-order learning task, the relation graph between the categories is constructed, more priori knowledge is added, the relation between the categories is utilized, and the GraphSAGE algorithm is introduced, so that the nodes in the constructed relation graph can be optimized, andthe description of the nodes is more perfect. And the final classification result also has better performance.

A method for semantic service registration and query based on WordNet is disclosed. The method includes the following steps: (1) semantic service registration: a service provider registers a service and uploads the Web Service Description Language (WSDL) document corresponding to the service, and a system parses the WSDL document to form a service description tree, then constructs a WordNet ontology tree according to the input of the service, performs a semantic annotation on the input / output of the service to form a Web Service Semantic Description Document (WSDL-S), and finally stores it in a register library; (2) semantic service discovery: a service requester inputs the information of service type, semantic information of the service input / output and other user-defined information to the register library to retrieve the services meeting the requirements; and (3) similarity sorting: the services meeting a certain threshold are sorted in descending order. The method has the advantages of the combination of WordNet ontology library and the semantic description language of WSDL-S, and definite semantic meaning.

The invention discloses a semantic item representation and disambiguation method based on word statistics and WordNet. A word meaning item set and a synonym set which are sorted in WordNet and widelyaccepted internationally are used as priori knowledge. The invention provides a semantic item vector generation method based on Wikipedia word statistics. Based on Wikipedia as a corpus, a word statistical model is trained to obtain a preliminary word vector. Then, semantic information of word statistics vector dimension words is fully utilized; the word vectors of the WordNet synonyms are combined to form semantic item vectors of the words. Meanwhile, the semantic calculation precision of words in different language environments can be improved, semantic item vectors can be reasonably and accurately used in practical application, and the method can be widely applied to various semantic calculation occasions of natural language processing.

The invention provides a tree network method for testing and updating through WordNet embedding and relates to the intelligent classification of pictures. The method comprises a step of building a tree network, a step of carrying out pre-training, a step of carrying out dynamic pruning based on an SVM, dynamically pruning a node with a too low activation value of characteristic mapping and a subtree of the node in a test process, wherein if the activation value of the node is too low, the probability of a class represented by the node is low, so the probability of a descendant node is low andnegligible, a step of carrying out the acceleration of a model in the test process of using the SVM since the sum of the activation values of characteristic maps has strong linear separability, anda step of carrying out online update based on a branch and transmitting a sample with a high prediction probability back to the tree network for training with a detected image as a training sample.

The invention discloses a crowdsourcing market-oriented task allocation method based on fitness. The method comprises the following steps: problem background definition: giving a problem background and definition for allocating picture classification tasks in the field of crowdsourcing knowledge acquisition; crowdsourcing task difficulty determination: using WordNet for calculating depth information of candidate tags, and calculating influences of all parts with the help of a polynomial function and a sigmoid function; worker ability evaluation: performing Max-Min normalization processing on the accuracy of the workers under each grade to further refine the worker performance; and optimal allocation scheme generation: converting an allocation problem between workers and tasks into a weighted bipartite graph matching model, measuring the fitness between the tasks and the workers by using a weighted Euclidean distance with a penalty factor, and solving the model by using a Kuhn-Munkres (KM) algorithm. According to the invention, the crowdsourcing platform can be helped to improve the task completion effect in the initial task distribution stage, and the crowdsourcing task completion accuracy is ensured by providing a group of effective task allocation schemes.

This paper proposes a new method for the objective evaluation of student work through the identification of original content in writing assignments. Using WordNet as a lexical reference, this process allows instructors to track how key phrases are employed and evolve over the course of a student's writing, and to automatically visualize the point at which the student's language first demonstrates original thought, phrased in their own, original words. After sketching the method for isolating “points of originality,” the paper provides a method for visualizing the resulting information. By visualizing otherwise subjective information in a way that is objectively intelligible, the goal is to provide educators with the ability to monitor student investment in concepts from the course syllabus, and to extend or modify the boundaries of the syllabus in anticipation of pre-existing knowledge or trends in interest.

The present invention relates to a microblog-oriented emotional entity search system. The system comprises the following five modules: 1) a user interface used for the interaction between a system and a user, so that the user can submit a query request via the module and obtain a feedback result; 2) a query expansion module used for conducting word relationship mining on microblog corpus data and establishing a weighted word relationship diagram in combination with a WordNet ontology base; 3) a query processing module used for converting a query request of a user into a query key word and a query sentence which can be accepted by an index database and conducting query expansion based on the word relationship diagram constructed by module 2); 4) an emotional information mining module used for conducting emotional mining on a microblog corpus and generating a determination rule for an emotional entity and an emotional polarity; 5) an emotional information decision and index establishment module used for determining the emotional entity and emotional polarity of microblog data, establishing an emotional information index and storing same; and 6) an inverted index establishment module used for establishing an inverted index for microblog text information and storing same. The present invention solves the difficult problems of the extraction of a microblog emotional entity, the emotional polarity analysis and the search for an emotional entity, etc., thereby providing an intelligent search product for analyzing and monitoring public opinions on a social network.