431 results about "Character vector" patented technology

The invention discloses a text named entity recognition method based on Bi-LSTM, CNN and CRF. The method includes the following steps: (1) using a convolutional nerve network to encode and convert information on text word character level to a character vector; (2) combining the character vector and word vector into a combination which, as an input, is transmitted to a bidirectional LSTM neural network to build a model for contextual information of every word; and (3) in the output end of the LSTM neural network, utilizing continuous conditional random fields to carry out label decoding to a whole sentence, and mark the entities in the sentence. The invention is an end-to-end model without the need of data pre-processing in the un-marked corpus with the exception of the pre-trained word vector, therefore the invention can be widely applied for statement marking of different languages and fields.

The embodiment of the invention discloses a named entity identification method, device and equipment and a computer readable storage medium. The method comprises the steps that character vectors and word vectors of a to-be-identified text are acquired, and weighted sum is carried out on the character vectors and the word vectors to obtain a weighted sum result; the weighted sum result is input into a target two-way LSTM model for processing to obtain a text character sequence; and the text character sequence is input into a target CRF model for processing to obtain a named entity identification result of the to-be-identified text. After the character vectors and word vectors of the to-be-identified text are acquired, weighted sum is carried out on the character vectors and the word vectors, dynamic weight information is used better, a relationship among words of a context is fully considered by adopting a two-way LSTM model, two-way information is fully adopted, then processing is carried out combined with a CRF model, thereby improving the accuracy rate of the named entity identification.

The present invention provides a text implication relation recognition method which fuses multi-granularity information, and proposes a modeling method which fuses multi-granularity information fusionand interaction between words and words, words and words, words and sentences. The invention firstly uses convolution neural network and Highway network layer in character vector layer to establish word vector model based on character level, and splices with word vector pre-trained by GloVe; Then the sentence modeling layer uses two-way long-short time memory network to model the word vector of fused word granularity, and then interacts and matches the text pairs through the sentence matching layer to fuse the attention mechanism, finally obtains the category through the integration classification layer; After the model is established, the model is trained and tested to obtain the text implication recognition and classification results of the test samples. This hierarchical structure method which combines the multi-granularity information of words, words and sentences combines the advantages of shallow feature location and deep feature learning in the model, so as to further improve the accuracy of text implication relationship recognition.

The invention discloses a named entities recognition method based on bidirectional LSTM and CRF. The named entities recognition method based on the bidirectional LSTM and CRF is improved and optimizedbased on the traditional named entities recognition algorithm in the prior art. The named entities recognition method based on the bidirectional LSTM and CRF comprises the following steps: (1) preprocessing a text, extracting phrase information and character information of the text; (2) coding the text character information by means of the bidirectional LSTM neural network to convert the text character information into character vectors; (3) using the glove model to code the text phrase information into word vectors; (4) combining the character vectors and the word vectors into a context information vector and putting the context information vector into the bidirectional LSTM neural network; and (5) decoding the output of the bidirectional LSTM with a linear chain condition random field to obtain a text annotation entity. The invention uses a deep neural network to extract text features and decodes the textual features with the condition random field, therefore, the text feature information can be effectively extracted and good effects can be achieved in the entity recognition tasks of different languages.

Transducers of different distributions detect the existence of targets of environment first and a test information composition system mixes the results detected by transducers to follow up the state and pick-up the character of the target. The state information composition system and the character dinformation composition system composite the traced states and identification results of character vector separately by each transducer to get relative accurate state parameter and identified results and information composition makes the entire systematic composition to provide favorable decision for next step, to increase systematic detecting probability, estimated accuracy and identified probability and more resonable algorithm.

The invention relates to an eccentric pump vibration accidence fusion diagnose method and relative vibration signal collector, wherein said invention is characterized in that: it uses the eccentric pump vibration signal collector to collect the normal state, the quantity imbalance, asymmetry rotate and loose base of eccentric pump; uses wavelet decomposition and reconstruction to extract the character of vibration signal; and inputs the character vectors into sub fuzzy neural networks I and II; to be treated and replace the relation factor matched with sensor signal function; the whole fuzzy neural network comprises data fuzzy layer, input layer, hidden layer and output layer; uses D-S theory to obtain the fused signal function distribution, realize the fusion diagnose on normal state, quantity imbalance, asymmetry rotate and loose base. The invention has simple structure and high effect.

This invention provides an identification method and a device for countenance based on the video, which applies the ASM profile pick-up algorithm in the pick-up of character vectors and picks up the man-face image based on the position of the eyes to generate a normalized character face from the chin and picks up the most effective character in the character face to identify the countenance, which can eliminate the influence of illumination to make the right and left gray value almost the same with the variance.

The invention provides a named entity identification method capable of combining an attention mechanism and multi-target cooperative training. The method comprises the following steps that: (1) carrying out a preprocessing operation on training data, and through character hierarchy mapping, obtaining the character vector representation of a sentence; (2) inputting the character vector representation obtained in (1) into a bidirectional LSTM (Long Short Term Memory) network, and obtaining the character vector representation of each word; (3) through word hierarchy mapping, obtaining the word vector representation of each sentence; (4) through the attention mechanism, splitting the word vector representation obtained in (3) with the character vector representation obtained in (1), and transmitting into the bidirectional LSTM network to obtain the semantic characteristic vector of the sentence; and (5) aiming at the semantic characteristic vector obtained in (4), carrying out entity annotation on each word by a conditional random field, and decoding to obtain an entity tag.

This invention discloses a method for media-crossing searches based on content relativity, which applies the typical relativity analysis to analyze the content characters of different mode media data, maps a visual sense character vector of image data and an auditory character vector of audio data in a low dimension isomorphic sub-space simultaneously by a sub-space mapping algorithm, measures the relativities among different mode data based on a general distance function and modifies the topological structure of a multi-mode data set in the sub-space to increase the cross media search efficiency effectively.

The invention discloses a multi-task named entity recognition and confrontation training method for medical field. The method includes the following steps of (1) collecting and processing data sets, so that each row is composed of a word and a label; (2) using a convolutional neural network to encode the information at the word character level, obtaining character vectors, and then stitching withword vectors to form input feature vectors; (3) constructing a sharing layer, and using a bidirection long-short-term memory nerve network to conduct modeling on input feature vectors of each word ina sentence to learn the common features of each task; (4) constructing a task layer, and conducting model on the input feature vectors and the output information in (3) through a bidirection long-short-term network to learn private features of each task; (5) using conditional random fields to decode labels of the outputs of (3) and (4); (6) using the information of the sharing layer to train a confrontation network to reduce the private features mixed into the sharing layer. According to the method, multi-task learning is performed on the data sets of multiple disease domains, confrontation training is introduced to make the features of the sharing layer and task layer more independent, and the task of training multiple named entity recognition simultaneously in a specific domain is accomplished quickly and efficiently.

The invention discloses a prosodic hierarchy model training method for text-to-speech and a text-to-speech method and a text-to-speech device by a prosodic hierarchy model. The training method includes: training massive non-tagged corpus data to obtain character vectors of individual characters; obtaining textual features and tagging corresponding to the trained data according to the character vectors and prosodic tagging data; based on a deep neural network and a bidirectional LSTM (long-short term memory) neural network, training the prosodic hierarchy model according to the textual features and tagging of the trained data. The training method has the advantages that when a character granularity based dictionary in the training method is compared with a traditional word granularity based dictionary, entry scale is decreased effectively, requirements of the model and resource files on computing resources and memory spaces are lowered, and usability of a prosodic prediction model in embedded intelligent devices is guaranteed while performance of the prosodic prediction model is improved.

A method for automatically identifying status based on sequence measure character in human body physiological image includes preprocessing and normalizing collected physiological image, picking up sequence measure character in normalized image by multi-pole differential filter, coding sequence measure character to form two-valued character vector, calculating Hamming distance between character vectors in two frames of physiological images and judging whether two frame of physiological image are coming from one and the same person or not according to calculated Hamming distance.

The invention belongs to the technical field of the natural language processing of computers, in particular to a construction and utilization method for a context-aware dynamic word or character vector on the basis of deep learning. The dynamic construction method for the context-aware dynamic word or character vector on the basis of the deep learning comprises the following steps of: in massive texts, through an unsupervised learning method, simultaneously learning a global feature vector of a word or character and the feature vector representation of the global feature vector when a specific context appears, and combining the global feature vector with the context feature vector, and dynamically generating word or character vector representation. By use of the method, the word or character vector dynamically constructed on the basis of the context can be applied to a natural language processing system. The method is mainly used for solving a problem that the word or character vector expresses different meanings in different contexts, i.e. the problem that one word or one character has multiple meanings can be solved. The dynamic word or character vector can be used for obviously improving the performance of various natural language processing tasks of different languages, wherein the tasks comprise Chinese word segmentation, part-of-speech tagging, naming recognition, grammatical analysis, semantic role tagging, sentiment analysis, text classification, machine translation and the like.

The embodiment of the present invention provides a named entity identification method and apparatus. The method includes obtaining a text to be recognize; obtaining a character vector corresponding toeach character in the text to be recognized by embedding the text input character into the model, and obtaining a phonetic vector corresponding to each character by embedding the text input characterinto the phonetic embedding model; obtaining combination vectors by combining each character vector and corresponding phonetic vectors, and inputting the combination vectors of all characters into BiLSTM for semantic coding to obtain the corresponding semantic information features of the text to be recognized; obtaining a corresponding entity label sequence in the text to be recognized accordingto the semantic information features. The apparatus is used for performing the method described above. The embodiment of the invention obtains the character vector and the phonetic vector of the textto be recognized respectively according to the character embedding model and the phonetic vector model, and combines the character vector and the phonetic vector and inputs the character vector and the phonetic vector into the BiLSTM for recognition, so the deficiency of the vector representation of the character can be well compensated, and the recognition accuracy is greatly improved.

The invention discloses a fused attention model-based Chinese text classification method. The method comprises the following steps of: respectively segmenting a text into a corresponding word set anda corresponding character set through word segmentation preprocessing and character segmentation preprocessing, and training a word vector and a character vector corresponding to the text by adoptionof a feature embedding method according to the obtained word set and character set; respectively carrying out semantic encoding on the word vector and the character vector by taking a bidirectional gate circulation unit neural network as an encoder, and obtaining a word attention vector and a character attention vector in the text by adoption of a word vector attention mechanism and a character vector attention mechanism; obtaining a fused attention vector; and predicting a category of the text through a softmax classifier. The method is capable of solving the problem that more redundant features exist in the classification process as existing Chinese text classification methods neglects character feature information of texts, the extracted texts are single in features, all the pieces of semantic information of the texts are difficult to cover and features having obvious contribution to the classification are not focused.

The invention discloses a text classification method based on feature information of characters and terms. The method comprises the steps that a neural network model is utilized to perform character and term vector joint pre-training, and initial term vector expression of the terms and initial character vector expression of Chinese characters are obtained; a short text is expressed to be a matrixcomposed of term vectors of all terms in the short text, a convolutional neural network is utilized to perform feature extraction, and term layer features are obtained; the short text is expressed tobe a matrix composed of character vectors of all Chinese characters in the short text, the convolutional neural network is utilized to perform feature extraction, and Chinese character layer featuresare obtained; the term layer features and the Chinese character layer features are connected, and feature vector expression of the short text is obtained; and a full-connection layer is utilized to classify the short text, a stochastic gradient descent method is adopted to perform model training, and a classification model is obtained. Through the method, character expression features and term expression features can be extracted, the problem that the short text has insufficient semantic information is relieved, the semantic information of the short text is fully mined, and classification of the short text is more accurate.

The embodiment of the invention discloses a character detection method and apparatus. The character detection method includes the steps: performing feature extraction of a plurality of abstraction layers on an image to be detected by means of a feature extraction network; predicting the probability that each pixel point in the image to be detected is a character pixel point, and when each pixel point is a character pixel point, the positional information of a bounding box of a character, opposite to the pixel point, by means of a character detection network; based on the prediction result of the character detection network, determining the positional information of the bounding box of each candidate character; inputting the extracted features into a character mapping network, and performing transformation on a feature diagram output from the character mapping network to generate a character vector; determining the neighbour candidate character of each candidate character in the image to be detected, and connecting the each candidate character with the related neighbour candidate character into a character set; and according to the positional information of the bounding box of eachcandidate character in the character set, determining the character region of the image to be detected. The character detection method improves accuracy of detection of irregular characters.

The invention discloses a URL (Uniform Resource Locator) classification-based phishing website detection method which comprises the following steps: conducting inductive analysis to the URL character of a phishing website, and vectorizing the URL characters; and according to the URL character vector, training the original data to obtain an initial classification model. The method has strong applicability and high efficiency.

The invention relates to a face model training module, relative method, and a face real-time identify system, and method, wherein in the identification, first using face sample image, to supply one face model supporting vector machine to each user; via collecting the video image input by camera, searching and checking the face of image, and tracking and identifying the image; then automatically marking the organ character point of face, to pre-treat checked face; calculating the Gabor character of face image after pretreatment; selecting low-dimension character vector from high-dimension Gabor character; inputting the selected low-dimension character vector into face model, to process face recognition, to feedback similarity data of each face model; based on said similarity data, outputting final face identifying result. The invention can improve the right rate of face recognition identification.

The invention belongs to the technical field of neural network, and provides a method and apparatus for increasing generalization capability of a convolutional neural network. The method includes the following steps: reading a group of images from a training set, mapping the group of images to a plurality of image character vectors, dividing the plurality of image characteristic vectors into a plurality of classes based on the types of the images; based on the image characteristic vectors of each class, calculating the intra-class loss function of all the characteristic vectors; based on the image characteristic vectors of each class, calculating the inter-class loss function of all the characteristic vectors; based on the intra-class loss function of all the characteristic vectors, using the back propagation algorithm to update the weight of each node of the convolutional neural network; repeating the above mentioned steps until the convolutional neural network converges on the training set or reaches preset repeating times. According to the invention, the method and the apparatus can save all data in long-tailed distribution, makes full usage of abundant inter-class information of tail data, and increases the generalization capability of the convolutional neural network.

The invention relates to an image quality assessment method, in particular to a non-reference image quality assessment method based on information entropy characters, and belongs to the field of image analyzing. The method comprises the first step of carrying out Contourlet conversion on a distorted image to obtain N*M conversion sub-bands, the second step of dividing each conversion sub-band and the unconverted original distorted image, the third step of calculating null domain information entropy and frequency domain information entropy on each block coefficient matrix, and the fourth step of screening the blocking characters and calculating a mean value to obtain the quality character value of each conversion sub-band. The method of a support vector machine and the method of non-reference image quality assessment are utilized for testing on a test set, and quality prediction and assessment are carried out through quality character vectors corresponding to a disaggregated model, an evaluation model and the test set all of which are obtained through training. The non-reference image quality assessment method has the advantages of being high in subjective consistency, small in time complexity and good in university, can be embedded into application systems related to image quality, and has very high application value.

The invention discloses a text sentiment classification method based on a deep learning combined model. The method comprises the steps of (1) training a word vector and a character vector; (2) performing word segmentation for each sentence of an annotated text, filling to a fixed length to obtain a training dataset I, performing character segmentation on each sentence of the annotated text, filling to the fixed length to obtain a training dataset II; (3) endowing words and characters of the two training datasets with corresponding word vectors and character vectors; (4) training the two training datasets by use of a text CNN (Convolutional Neural Network) and an improved Dynamic CNN model to obtain four base classifiers, wherein a progressive learning method is adopted as a training methodand a focal loss function is adopted as a loss function and (5) combining the four base classifiers to obtain the text sentiment classification method of a combined model. The text sentiment classification method implemented by the invention does not depend on a specific sentiment dictionary, is not limited to a specific field and has high universality and expandability.

The invention relates to a legal document named entity identification method and apparatus, and a computer device. The method comprises the steps of obtaining a legal document to be identified; inputting the legal document to be identified into the deep neural network model for identification to obtain an identification result; wherein the deep neural network model is obtained by training a language model through a plurality of legal document data with labels, a bidirectional recurrent neural network and a conditional random place; wherein the language model is obtained by training a Google Bert model through a plurality of corpora. According to the invention, the deep neural network model is adopted to carry out entity identification; extracting a character vector from a Chinese charactersequence of the legal document to be identified by adopting a language model obtained by training a Google Bert model; and inputting the character vectors into a bidirectional recurrent neural network, inputting the output codes of the bidirectional recurrent neural network into a linear chain conditional random field, and obtaining a recognition result, so that the network for realizing named entity recognition is simple in structure, low in training cost and high in prediction capability.

The invention relates to a dynamic text cluster method based on network flow and laminated knowledge base, belonging to information processing and network safety technical field. The inventive method comprises that first assuming the clustered file type is provided with vector character, using TFIDF method to extract and normalize character of single clustered text, using the method that defining meaning distance in the knowledge base to calculate the distance of text and type, adjusting and refreshing keyword and weight of the type of new added file, when present file can not be combined with known types, needs to build new type. And the algorism comprises dynamic character vector extraction, type classification, distance calculation, type combination and new type construction. The invention is characterized in that the cluster process is based on the meaning information provided by the laminated knowledge base but not keyword, the invention can dynamically remove noise data, the similarity is calculated network flow algorism, to confirm optimized match, and the invention can meet real-time refresh cluster of Web text, in particular as non-detect type, without pre-appointing type group.

The invention relates to a motion sensing technique based bank ATM intelligent monitoring method. The method includes: arranging a 3D motion sensing device in an operation zone of an ATM; capturing a human body 3D skeleton image of an operator in real time; extracting key joint point character vectors; inputting the character vectors in to a motion classifier to be classified; determining if a motion is a dangerous violence motion needing to be monitored; and training the motion classifier through a learning method in advance. Through the 3D visual perception technique, the method can detect, analyze, and track a human body motion posture and a figure motion trail, can dynamically capture a motion of an ATM operator in real time, and then predict dangerous violence motions of falling on a ground, grappling and destroying the ATM, and sends an alarm in advance, and in this way, abnormal behaviors can be prevented in the operation zone of the ATM.

A natural language processing method and corresponding apparatus are disclosed. The natural language processing method may include converting words in sentence data, recognized through voice recognition, to corresponding word vectors, and converting characters in the sentence data to corresponding character vectors. The natural language processing method also may include generating a sentence vector based on the word vectors and the character vectors, and determining intent information of the sentence data based on the sentence vector.

The invention discloses a film and television entity identification method based on Bilstm-crf and knowledge graph, which comprises the steps of obtaining a character vector and a part-of-speech vector of a to-be-recognized text, performing weighted summation on the character vector and the part-of-speech vector, and inputting a result into a target bidirectional LSTM model for processing to obtain a text feature sequence; inputting the text feature sequence into a target CRF model for processing to obtain a named entity recognition result of the to-be-recognized text; and inquiring the namedentity recognition result in the film and television knowledge graph to further verify the result. According to the method, entity extraction can be effectively carried out on the movie and televisionsearch text of the user, the movie and television knowledge graph is fully utilized to mine the abstract movie and television search intention of the user, and the use experience of the user is improved. Word vectors trained through the language model are used as bottom layer input of the neural network under the condition of less annotation data, so that the training efficiency is improved, andthe method has a good application prospect and can be widely applied to entity recognition scenes in various fields.

The invention discloses an image retrieving method and device. The method includes that an image to be retrieved is subjected to sub-image division so as to obtain a plurality of sub-images; each appointed sub-image among the plurality of sub-images is subjected to image characteristic extraction to obtain a character vector of each appointed sub-image; according to each image in an image library and on the basis of character vectors of the sub-images in sub-image groups to be matched with the image and character vectors of the appointed sub-images, similarity of the image and the image to be retrieved is determined; and on the basis of the similarity of each image in the image library and the image to be retrieved, a retrieving result of an image corresponding to the image to be retrieved is determined. By means of the image retrieving method and device, effectiveness of the retrieving result and retrieving efficiencies are improved when retrieving is performed on the basis of image contents.

This invention relates to method for identifying and confirming an identity by the soundwave information of mankind including a model training method and a soundwave identification method characterizing in that every speaking person makes up of soundwave model: M=***, the character vector sequence to be identified x'={X1,...XT} matching with the speaking person soundwave model M=*** gets marks (log likelihood) as 1:s(x' / M)=***. This invention can adjust threshold value of an operation point according to different requirements to reach the highest accurate rate.

The invention relates to a named entity recognizing method based on LSTM. The method includes the steps that training linguistic data of named entity recognition is labeled to form labeled linguisticdata; words and characters in the labeled linguistic data are converted into vector quantities; a named entity recognition model based on LSTM is set up by using word and character vector quantities,and parameters of the named entity recognition model are trained; named entity recognition prediction is conducted on the data to be predicted through the trained named entity recognition model. By adopting the vector quantities based on the words and characters, characteristics of the characters and the words can be obtained, and the problem of unregistered words can be avoided; in addition, compared with a traditional pure CRF model algorithm, the named entity recognizing method adopting long short-term memory neural network is capable of absorbing more character and word characteristics tofurther improve entity recognition precision.