176 results about "Relation classification" patented technology

The present invention relates to a computer-based system for generating a risk register relating to a named entity. The system comprises a computing device, a risk database accessible by the computing device and having stored therein a set of risk types based on an induced taxonomy of risk types previously derived at least in part upon operation of a machine learning module, an input adapted to receive a set of source data, the set of source data being in electronic form and representing textual content comprising potential risk phrases, a entity-risk relation classifier adapted to identify and extract entity-risk relations from the set of source data, a risk tagger adapted to identify in the set of source data a set of risk candidates (ri) based on the set of risk types, a entity tagger adapted to identify mentions of entity names (ci) in the set of source data, and a risk register aggregator adapted to generate a first risk register based on the set of tuples associated with a first entity.

The invention provides a named entity relation extraction and construction method based on deep learning, which is applied to the technical field of internet information. The method includes, to a specific field, capturing news data within the field from a vertical website and preprocessing the captured news data; segmenting the news data, extracting key words to generate a field lexicon, and segmenting the news data again according to the field lexicon; extracting a seeded lexicon; constructing an entity relation network in an unsupervised manner, extracting sentences containing at least two entities from the news data, extracting verbs of the sentences and corresponding documents, building a word clustering model based on deep learning to the extracted documents, and constructing the entity relation network according to relation between words described by the verbs; and finally defining category of the entity relation and performing relation classification to each entity pair of the entity relation network. Without input of large-scale manpower to mark sample data, dependence to the corpus is reduced and performance of entity relation extraction is high.

The invention relates to a remotely-supervised Dual-Attention relation classification method and system. The method comprises the following steps: aligning entity pair in a knowledge base to news linguistic data through remote supervision, and constructing an entity pair sentence set; performing word-level vector encoding on the sentence through a Bi-LSTM model based on a word-level attention mechanism so as to obtain a semantic feature encoding vector of the sentence; performing encoding and denoising on the semantic feature of the sentence through the Bi-LSTM model based on the sentence-level attention mechanism so as to obtain a sentence set feature encoding vector; and packing the sentence set feature encoding vector and the entity pair translation vector, and performing the relation classification of the entity pair on the obtained packet feature. Through the technical scheme provided by the invention, the noise data of the model training is reduced, the artificial data annotationand the caused error transmission thereof are avoided. The entity alignment is performed by applying the open domain text and the large-scale knowledge library, and the annotation data scale problemof the relation extraction is effectively solved.

The invention discloses an enterprise entity relation extraction method based on a convolutional neural network. The method comprises the steps of a relation corpus building stage, wherein an initial seed relation pair set is built artificially, and by means of an internet search engine and a Bootstrapping technology, relation language materials are generated in an iteration mode, and finally a relation corpus is formed; a relation classification model training stage, wherein term vectors and position embedding are combined to build a sentence vector matrix representation to serve as input of a network, the convolutional neural network is built, the network is trained by means of a back propagation algorithm, and a relation classification model is obtained; an enterprise entity relation extraction stage in a web page, wherein the web page is preprocessed by combining web page text extraction with a named entity identification technology, and then enterprise entity relation extraction is conducted on the preprocessed web page. By means of the method, not only the defects of an artificial feature method can be overcome, but also the enterprise entity relation can be extracted from the web page more accurately and efficiently.

The invention discloses a multi-triad joint extraction method based on knowledge graph embedding, comprising the following steps of: processing an acquired text statement to obtain a text statement matrix; inputting the text statement matrix into a Transformer model to extract semantic information of text statements to obtain semantic feature vectors; applying the semantic feature vectors to an entity recognition sequence labeling task to obtain entity recognition cross entropy loss loss1; applying the semantic feature vector to a relationship classification task, and solving entity recognition cross entropy loss loss2 of relationship classification; constructing an entity word relationship by utilizing an entity labeling prediction matrix and a statement entity word relationship classification matrix, and solving cross entropy loss loss3 of the relationship; calculating a minimized total loss function loss by utilizing an optimization algorithm based on gradient descent of the loss1,the loss2 and the loss3; and obtaining a trained Transformer model according to the text statement to be predicted, inputting the text statement to be predicted into the trained Transformer model to obtain a predicted semantic feature vector of the predicted text statement, and completing a multi-triad joint extraction method.

The invention discloses an extraction method of a semantic relation between Chinese entities. The extraction method comprises the following steps of: carrying out syntactic analysis on natural statements to determine a complete syntactic tree of the natural statements; extracting a shortest path containing tree between two Chinese entities from the complete syntactic tree; extracting a path verb nearest to a second Chinese entity from the shortest path containing tree; respectively acquiring the semantic information of the two Chinese entities and the path verb; adding the three acquired semantic information into a root node of the shortest path containing tree according to a preset rule to determine the expanded shortest path containing tree to be a natural statement relation tree; and carrying out relation classification on the relation tree by utilizing a prestored classification model. According to the extraction method of the semantic relation between Chinese entities, which is disclosed by the invention, the relation tree contains abundant structured information and lexical semantic information and has better generality and semantic relation extraction overall performance, the dependence degree of a large-scale corpus is relieved, and meanwhile, the calculated amount of the system is lower.

The invention discloses a relation extraction method in combination with clause-level remote supervision and semi-supervised ensemble learning. The method is specifically implemented by the following steps of 1, aligning a relation triple in a knowledge base to a corpus library through remote supervision, and establishing a relation instance set; 2, removing noise data in the relation instance set by using syntactic analysis-based clause identification; 3, extracting morphological features of relation instances, converting the morphological features into distributed representation vectors, and establishing a feature data set; and 4, selecting all positive example data and a small part of negative example data in the feature data set to form a labeled data set, forming an unlabelled data set by the rest of negative example data after label removal, and training a relation classifier by using a semi-supervised ensemble learning algorithm. According to the method, the relation extraction is carried out in combination with the clause identification, the remote supervision and the semi-supervised ensemble learning; and the method has wide application prospects in the fields of automatic question-answering system establishment, massive information processing, knowledge base automatic establishment, search engines, specific text mining and the like.

The invention relates to a method for constructing a health knowledge graph. The method comprises the following steps of directly extracting entities of users, symptoms, diseases, experts, treatment schemes and commodities belonging to generalized representations in structured and semi-structured data from a network data source by utilizing an html label and a regular expression; extracting entities belonging to the six summarized representations from the unstructured data by using a conditional random field algorithm; using Bi-pairs of entities extracted in the same context The LSTM algorithmcarries out relation classification and determines a relation between entities; calculating the correlation between the entity names and the entity descriptions and achieving the disambiguation of the entity information; and complementing the knowledge graph relation by using an owl reasoning function of a jena tool, capturing ambiguous triplet by using a criterion, and feeding back the triplet which is judged to be possibly wrong to a domain expert for verification. The method has the beneficial effects that the health knowledge graph of the traditional Chinese medicine theory is constructed, the incomplete relation is automatically complemented by applying the knowledge reasoning technology, and the more perfect health graph is constructed.

The invention discloses an entity relationship extraction method of a concerned associated word, and the method comprises the following steps: S1, inputting a labeled text and a to-be-tested text, carrying out text word segmentation, and obtaining a corresponding real value vector of each word mapping; s2, combining the real value vectors corresponding to all the words in pairs to obtain binary combination feature vectors, calculating weight vectors of the binary combination feature vectors and the relation labels, and obtaining binary word group features of the sentences; s3, inputting the real value vector into a neural network layer, and obtaining a semantic structure and a feature vector representation of the text; and S4, connecting the binary phrase characteristics output in the step2 and the semantic characteristics output in the step 3 in series to serve as the representation of a final text sentence, inputting the representation of the final text sentence into a concern layerof a sentence level, obtaining the weights of the sentence for different relation types, and obtaining a final relation classification result and outputting the final relation classification result.

The invention discloses a man-machine interaction question-answering method and system based on complex intention intelligent recognition, and the method comprises the steps: obtaining an original question sentence of a user, carrying out the sentence segmentation and part-of-speech tagging, and obtaining the part-of-speech information of each component word of the question sentence; performing dependency syntax analysis on the question sentence to obtain a dependency syntax tree; carrying out industry entity identification to obtain industry entities and the number, and extracting a core dependency tree to simplify questions; carrying out industry question relation classification on the questions, carrying out Chinese multi-intention question rewriting, and then carrying out knowledge retrieval on the questions; and selecting and generating answers for knowledge retrieval results, and returning the answers to the user. According to the method and system, multi-intention complex questions can be effectively simplified in any industrial scene, the intention of the user can be accurately understood, the industrial knowledge can be more naturally fed back to the user, the user can more accurately and quickly obtain the required industrial knowledge, the user experience is improved, and the method and system are particularly suitable for man-machine interaction intelligent questions and answers in the medical industry.

The invention provides a Chinese relationship extraction method, which comprises the following steps of S1, data preprocessing: performing pre-training processing of multi-granularity information on atext of input data to extract distributed vectors of three levels of characters, words and word meanings in the text; S2, feature coding: taking a bidirectional long-short-term memory network as a basic framework, obtaining hidden state vectors of the characters and hidden state vectors of the words through the distributed vectors of the three levels of the characters, the words and the word meanings, and then obtaining final hidden state vectors of the character level; and S3, relationship classification: learning the final hidden state vector of the word level, and fusing the hidden state vector of the word level into a sentence-level hidden state vector by adopting an attention mechanism of the word level. The problems of word segmentation ambiguity and polysemy ambiguity are effectively solved, the performance of the model on a relation extraction task is greatly improved, and the accuracy and robustness of Chinese relation extraction are improved.

The invention provides a sequence text information-combined knowledge graph expression learning method and a knowledge graph expression learning device. According to the method, not only the ternary relation group information between entities is utilized, but also the sequence text information containing the entities in a designated corpus is fully utilized. An energy equation is constructed, so that the entities have different expression vectors in the structured ternary relation group information and the non-structured text information. Meanwhile, a marginal-based evaluation function is minimized, and the expression of structure-based entity vectors, text-based entity vectors and relation vectors is learned. Therefore, the expression learning effect of a knowledge graph is remarkably improved. According to the method and the device, the learned knowledge graph expression fully utilizes the sequence text information of entities contained in the corpus. Therefore, the higher accuracy can be obtained in the tasks such as ternary group relation classification, ternary group head and tail entity prediction and the like. The good practicability is achieved, and the expression performance of the knowledge graph is improved.

The present invention provides an entity and relation joint learning method based on an attention model, which is characterized by the following steps: first, inputting a clinical text sequence into an Embedding layer in the form of [CLS] Sequence [SEP] to obtain the initial vector representation H0 of each word; then, inputting the H0 into the first N-K layer of a multi-head self-attention mechanism, and outputting the context representation information Hm of each word; then, inputting the Hm into the remaining K layers to obtain a word vector representation H<m><task> corresponding to an entity recognition and relation extraction task; finally, inputting the H<m><task>, learning by a matrix MASKtask, entity recognition, or relationship classification downstream task layer, and then outputting entity and relation information. Experimental results show that the method of the present invention is significantly superior to other methods in entity recognition, relation extraction, and joint learning, indicates the effectiveness of the method.

The invention discloses an application program recommendation method and device used for describing the wide correlation among application programs and improving the effect of recommending the application programs to users. The application program recommendation method comprises the steps that whether the correlation exists among the application programs is determined according to comment information and description information of the application programs; if the correlation exists among the application programs, the classification relation among the application programs is determined from a relation classification system orienting the application programs according to the preset relation classification feature among the application programs; the application programs are recommended to the users according to the classification relation among the application programs.

The invention relates to a neural network relation classification method and an implementation system thereof, which fuse discrimination degree information, comprising: (1) data preprocessing; (2) training word vector; (3) extracting position vector; Combining the word vector of each word with the high-dimensional position feature vector, the joint feature is obtained. (4) computing the semantic representation of sentences; The context information and semantic information of entities are encoded by bi-directional LSTM. (5) calculating a discrimination vector; After the two entity vectors are subtracted from each other and cascaded with a Bi-LSTM unit encode that input; (6) inputting the outputs of (4) and (5) to the CNN, outputting the feature vector fused with the discrimination information, and inputting the feature vector to the classifier for classification; (5) adopting a Loss function to train the model. The invention does not need to manually extract any features, and the jointmodel does not need to preprocess the data by means of other natural language processing tools, the algorithm is simple and clear, and the effect is best at present.

The invention relates to a specific target emotion classification task based on a graph neural network. The method comprises the following steps: acquiring a data set and initializing a BERT model; obtaining a one-dimensional feature vector of each target word through a BERT model; inputting the feature vector of the target word into a graph convolutional neural network model; constructing a network topological graph, calculating an adjacency matrix, obtaining three features of nodes in the network topological graph in three modes according to the adjacency matrix, introducing relation classification tasks,wherein the whole model is divided into two stages and two tasks in classification, and the two tasks are emotion polarity classification of target subjects and relation classificationbetween the target subjects respectively. According to the method, the graph neural network is adopted to compose a plurality of subjects appearing in sentences and process a plurality of targets at the same time, so that the cognitive law of judging emotion polarity by human beings is better met, the effect of a model is ensured, meanwhile, a relationship classification task is introduced for auxiliary classification, and the classification accuracy is further improved.

The invention discloses an entity relationship joint extraction method based on span and knowledge enhancement, and belongs to the technical field of information extraction and natural language processing. The method comprises the following steps: firstly, constructing a sample data set and labeling the data set; carrying out entity identification and relationship classification and specifically,for the labeled data, mapping words in a high-dimensional discrete space to a low-dimensional continuous space vector by a pre-training language model; carrying out span identification, filtering andrelationship classification by a span-based model; converting relationship classification into graph classification by utilizing a graph-based model, and introducing a syntactic dependency relationship so as to assist relationship judgment and classification; and performing joint training on an output result of the span-based model and an output result of the graph-based model, and identifying entities contained in the data and relationships among the entities. Syntactic information such as the dependency relationship is introduced into the end-to-end neural network model, so that the overlapping relationship is effectively identified, and the joint extraction accuracy of the entity relationship is improved.

An online user relation measurement and classification method based on a three-dimensional relation strength model includes the step of setting up the three-dimensional relation strength model, the step of setting up a visualized model of relation strength, the step of measuring the social network user relation and the step of social user classification and recommendation. Social relations of users in a social network are abstracted into three basic dimensions, namely, the structure dimension, the psychological dimension and the tool dimension, and the dimensions are continuous variables from zero to one. Vectors of the three dimensions are mapped to three basic colors, then the problem of the relation strength can be converted into the color problem, and therefore visualized analysis on the relation strength can be conducted. According to the different relation classifications, the relations between the users are more accurately classified, and therefore basis is provided for subsequent processing, and the user classification and recommendation method which is more practical, more convenient to implement and more accurate can be provided for social network service platforms such as the SNS.

The invention discloses a user relation recognition method and device and belongs to the field of network technology. According to the method, a user relation classification model obtained through training by use of property information of users and interaction information among the users is utilized to perform relation recognition on to-be-recognized users. Since model training is performed basedon the property information of the users, the interaction information among the users, etc., the user relation classification model can represent accurate relations among the users, and accordingly the purpose of recognizing the relations among the users is achieved. Moreover, the user relations obtained through recognition are high in accuracy and beneficial for pertinently making application strategies for the relations among the users subsequently, and a powerful basis is provided for commercial operation, etc.

The invention provides a character relation classifying method and device. The method comprises the following steps: acquiring a labeled vector set, a unlabeled vector set and a candidate relation test set; expanding the labeled vector set according to the labeled vector set and the unlabeled vector through robust path similarity measurement; carrying out training study on the expanded labeled vector set by a nonlinear semi-supervised classifying function to obtain a character relation classifying model; and carrying out character relation classification on the candidate relation test set by the character relation classifying model. The labeled vector set is expanded by the robust path similarity measurement, influence of noises and isolated points is eliminated by the effect of robustness, manually annotated corpus is reduced, consumption and waste of manpower resources are reduced, the bottleneck problem of manually labeled corpus when a large amount of labeling is required is solved, a regularized semi-supervised classification algorithm is applied to character relation extraction, and the character relation classification efficiency is improved.

The invention belongs to the technical field of movement behavior and social relations analysis, in particular to an offline social relations classification method based on user movement behavior and a device. The social relations classification method based on user movement behavior comprises the steps that individual behavior record of users are acquired from a user movement behavior log database, a user behavior matrix and a space and time interaction matrix between every two users are constructed and two interaction characteristics evaluating user social similarity and interaction regularity are extracted; spatial temporal entropy and regularity are obtained; random processing is conducted on the user individual movement behavior through a null hypothesis, a null model of a user individual behavior matrix and a time and space interaction matrix between the users is constructed, and according to the null model and the preset probability, the spatial temporal entropy random threshold value and regularity random threshold value are determined; by comparing the value relationship between the user interaction matrix in the spatial temporal entropy and the regularity and random threshold value of the user interaction matrix, the offline social relations between the two users are determined. The social relations classification method based on user movement behavior and the device are capable of being used for distinguishing offline acquaintance relations from stranger relations, digging out the familiar stranger relations between the two and improving the accuracy and the practicability of offline social relations classification.

The invention provides an entity relationship extraction method and device, electronic equipment and a storage medium, and relates to the technical field of natural language processing. The method comprises the steps of obtaining a text of a to-be-extracted entity relationship, and processing the text through an entity and relationship joint extraction model to obtain a first entity relationship triple; determining words in the text and labels corresponding to the words, matching the words and the labels according to the relation rule set, and determining a second entity relation triple; constructing entity pairs according to the words and the labels, processing the text through a relationship classification model, and determining third entity relationship triples for the entity pairs; anddetermining an entity relationship extraction result according to the first entity relationship triad, the second entity relationship triad and the third entity relationship triad. According to the invention, the entity relationship extraction accuracy can be improved.

The invention relates to the technical field of natural language processing, in particular to a question-answering method for entity relationship extraction based on transfer learning. The acquisitionof a relationship classification result comprises the steps: obtaining and preprocessing a source domain text data set and a target domain text data set; inputting the preprocessed data into a skip-gram model for training to obtain word vectors of the source domain text data and the target domain text data, obtaining position vectors of the source domain text data and the target domain text data,and cascading the position vectors with the word vectors to obtain joint feature vectors of the source domain text data and the target domain text data; inputting the joint feature vector of the source domain text data into a BiLSTM network for pre-training to obtain network parameters in the pre-training process and context information and semantic features of the source domain text data; and inputting the joint feature vector of the target domain text data into a BiLSTM _ CNN fusion model for retraining to obtain a high-dimensional feature vector of the target domain text data, sending thehigh-dimensional feature vector into a classifier, and outputting the relationship classification result. Question-answering accuracy can be improved.

The embodiment of the invention discloses an information extraction method based on artificial intelligence, a storage medium and a related device, and is applied to the technical field of informationprocessing of artificial intelligence. For each sentence of a short text in a to-be-processed document, the information extraction method firstly determines relationship attributes contained in the short text according to a preset relationship classification model, then extracts entity pairs corresponding to the relationship attributes according to a preset entity extraction model, and can form triple information corresponding to each sentence of the short text. Practice proves that triple information of the to-be-processed document can be accurately obtained in the mode that the relation attributes and the entity pairs corresponding to the relation attributes are determined through different machine learning models, and transfer learning can be carried out due to the fact that the machine learning models are adopted for information extraction.

The invention relates to an agricultural field relation extracting method based on syntactic-semantic. The method comprises the following steps: firstly extracting texts related to agricultural knowledge from an agricultural professional website; organizing the collected texts, and removing unrelated texts, to obtain the text related to the agricultural field; combining a predetermined relation type, performing artificial labelling on the agricultural field text, and building a corpus of the agricultural field; according to a determined characteristic type, performing feature extraction on thetext, and building a feature vector; and finally, modeling by using a SVM pair classifier, to obtain a relation classification model towards the agricultural field. The method has the better extracting effect.

Embodiments of the invention provide a text-based entity relationship extraction method and device. The method comprises the steps of determining words in a description text to which a to-be-processedentity belongs in a to-be-processed text from a preset corpus as to-be-processed words; obtaining text vector representation according to the affiliation relationship between the to-be-processed entity and the to-be-processed text and the affiliation relationship between the to-be-processed word and the to-be-processed text; inputting the text vector representation into a preset relationship classification model to obtain the relationship between to-be-processed entities. According to the method, the entities to be processed and the description text to which the entities to be processed belong are introduced, so that the obtained text vector representation contains potential information existing in the entities to be processed, and therefore, the text classification accuracy can be improved by applying the method provided by the embodiment of the invention.

The invention discloses an establishment method and system for a bilingual semantic relation classification model. The method comprises the followings: semantic relation classification is performed on the first un-annotated corpus of an un-annotated semantic relation, which corresponds to first language, through a first language classifier, so as to obtain a semantic relation classification result; the first un-annotated corpus is translated into a second un-annotated corpus corresponding to a second language; the corresponding relation of the first un-annotated corpus and the semantic relation classification result is stored into a first training set of the first language classifier; the corresponding relation of the second un-annotated corpus and the semantic relation classification result is stored into a second training set of the second language classifier. After the adoption of the method or the system provided by the invention, the demands for large-scale training corpus can be reduced, and the establishing efficiency for the bilingual semantic relation classification model can be improved.

The invention discloses a bio-medical entity relation classification method combining an attention mechanism and a neural network, which belongs to the technical field of biomedicine and data mining,and is used for solving the problem of bio-medical entity relation classification. According to the key points, the method comprises the steps of S1, performing text processing based on reference analysis; S2, constructing a model input vector based on the attention mechanism; S3, building a bio-medical entity relation classification model based on a bidirectional LSTM; and S4, carrying out bio-medical entity relation classification by utilizing the relation classification model. According to the method, following-based reference analysis is designed for sentences in biological literatures; then starting from basic unit words forming the sentences, embedding vectors of the words are weighted by utilizing the attention mechanism; weights of keywords having an important influence on bio-medical entity relation classification are highlighted, so that a relation between candidate entities is clearer; and the bio-medical entity relation classification is carried out.

The invention provides a text recognition method and device and a storage medium thereof, which relate to the technical field of relation extraction and classification. The text recognition method comprises the following steps of obtaining the text to be processed; inputting the text to be processed into a text recognition model based on a convolution neural network and an attention mechanism, andobtaining the entity relationship of the text to be processed outputted from the text recognition model. This text recognition method improves the accuracy of relation classification and has higher efficiency of relation extraction by extracting the entity relation of text based on the convolution neural network and attention mechanism of text recognition model.

The invention provides a remote supervision relation extraction method with entity perception based on a PCNN model. The method specifically comprises the following steps: combining word embedding with head entity and tail entity embedding and relative position embedding by using a multi-head self-attention mechanism to generate enhanced word semantic representation of a perceptible entity, whichcan capture semantic dependence between each word and an entity pair; introducing a global door, and combining the enhanced word representation perceived by each entity in the input sentence with theaverage value of the enhanced word representations to form a final word representation input by the PCNN, and in addition, in order to determine a key sentence segment in which the most important relationship classification information appears. According to the method, another gate mechanism is introduced, and different weights are allocated to each sentence segment, so the effect of key sentencesegments in the PCNN is highlighted. Experiments show that the remote supervision relationship extraction method provided by the invention can improve the prediction capability of the remote supervision relationship in the sentence.