920 results about "Conditional random field" 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 invention provides an image semantic division method based on a depth full convolution network and a condition random field. The image semantic division method comprises the following steps: establishing a depth full convolution semantic division network model; carrying out structured prediction based on a pixel label of a full connection condition random field, and carrying out model training, parameter learning and image semantic division. According to the image semantic division method provided by the invention, expansion convolution and a spatial pyramid pooling module are introduced into the depth full convolution network, and a label predication pattern output by the depth full convolution network is further revised by utilizing the condition random field; the expansion convolution is used for enlarging a receptive field and ensures that the resolution ratio of a feature pattern is not changed; the spatial pyramid pooling module is used for extracting contextual features of different scale regions from a convolution local feature pattern, and a mutual relation between different objects and connection between the objects and features of regions with different scales are provided for the label predication; the full connection condition random field is used for further optimizing the pixel label according to feature similarity of pixel strength and positions, so that a semantic division pattern with a high resolution ratio, an accurate boundary and good space continuity is generated.

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.

The invention discloses a method for extracting a text-oriented field term and term relationship. The method is characterized by comprising the following steps of: firstly, preprocessing original linguistic data to obtain a candidate word set including clauses, participles and part of speech tagging, and filtering noise words; secondly, extracting term characteristics from the original linguisticdata and the Internet, and separating terms from candidate words by combining with a dual-model structure algorithm; thirdly, constructing a term dictionary by adopting an inverted index method, and tagging the terms in a text to be identified by using a longest match algorithm; and finally, carrying out multilevel sign sequence tagging through a conditional random field model according to a multi-dimensional node signing rule to obtain a relationship among the terms in the text to be identified.

A method and apparatus for automatically performing medical image analysis tasks using deep image-to-image network (DI2IN) learning. An input medical image of a patient is received. An output image that provides a result of a target medical image analysis task on the input medical image is automatically generated using a trained deep image-to-image network (DI2IN). The trained DI2IN uses a conditional random field (CRF) energy function to estimate the output image based on the input medical image and uses a trained deep learning network to model unary and pairwise terms of the CRF energy function. The DI2IN may be trained on an image with multiple resolutions. The input image may be split into multiple parts and a separate DI2IN may be trained for each part. Furthermore, the multi-scale and multi-part schemes can be combined to train a multi-scale multi-part DI2IN.

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.

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.

A system for extracting information from text, the system including parsing functionality operative to parse a text using a grammar, the parsing functionality including named entity recognition functionality operative to recognize named entities and recognition probabilities associated therewith and relationship extraction functionality operative to utilize the named entities and the probabilities to determine relationships between the named entities, and storage functionality operative to store outputs of the parsing functionality in a database.

The invention discloses a deep learning-based weakly supervised salient object detection method and system. The method comprises the steps of generating salient images of all training images by utilizing an unsupervised saliency detection method; by taking the salient images and corresponding image-level type labels as noisy supervision information of initial iteration, training a multi-task fullconvolutional neural network, and after the training process is converged, generating a new type activation image and a salient object prediction image; adjusting the type activation image and the salient object prediction image by utilizing a conditional random field model; updating saliency labeling information for next iteration by utilizing a label updating policy; performing the training process by multi-time iteration until a stop condition is met; and performing general training on a data set comprising unknown types of images to obtain a final model. According to the detection method and system, noise information is automatically eliminated in an optimization process, and a good prediction effect can be achieved by only using image-level labeling information, so that a complex andlong-time pixel-level manual labeling process is avoided.

The invention discloses a geographical science domain named entity recognition method, which is used for recognizing geographical science core term entities and geographical location entities. The method mainly comprises three steps of (1) establishing a geographical science domain dictionary, and using a new word discovery algorithm to identify new words in the geographical science domain in an unsupervised way; (2) training and testing based on a conditional random field (CRF) model and a multichannel convolutional neural network (MCCNN) model; (3) carrying out error correcting and fusion on entities recognized by the models by using a rule-based method. According to the geographical science domain named entity recognition method, the new words of the domain are identified as the dictionary in an unsupervised way by using the new word discovery algorithm, so that the work distinguishing effect is improved. The semantic vectors of the words are learnt from large-scale unmarked data in an unsupervised way, and basic characteristics of the words are synthesized and are taken as the input characteristics of the MCCNN model, so that manual selection and construction of the characteristics are avoided. The predicting results of the two models are fused by means of a custom rule, so that the problem of error marking in a recognition process can be corrected.

The invention provides a customer service information providing method and device, electronic equipment and a storage medium. The method comprises the steps of receiving a Chinese text input by a user; inputting the input Chinese text into a Chinese customer service question-answering model based on a Bi-LSTM (Bidirectional Long Short-Term Memory) model and a CNN (Convolutional Neural Network) model to acquire an answering statement; inputting the input Chinese text into a content extraction and intention classification model based on a Bi-LSTM-CRF (Conditional Random Field) model and an LSTMclassifier to acquire customer intention classification and key information; determining service recommended to a user according to the customer intention classification and the key information; inputting the input Chinese text into a Chinese text emotion analysis model based on the CNN model to acquire a user emotion classification; adjusting the answering statement according to the user emotionclassification; and in combination with the adjusted answering statement and the determined service, providing customer service information to the user. According to the method and device optimizationmodel provided by the invention, the automatic customer service answering is realized.

The invention discloses a generative adversarial network-based pixel-level portrait cutout method and solves the problem that massive data sets with huge making costs are needed to train and optimizea network in the field of machine cutout. The method comprises the steps of presetting a generative network and a judgment network of an adversarial learning mode, wherein the generative network is adeep neural network with a jump connection; inputting a real image containing a portrait to the generative network for outputting a person and scene segmentation image; inputting first and second image pairs to the judgment network for outputting a judgment probability, and determining loss functions of the generative network and the judgment network; according to minimization of the values of theloss functions of the two networks, adjusting configuration parameters of the two networks to finish training of the generative network; and inputting a test image to the trained generative network for generating the person and scene segmentation image, randomizing the generated image, and finally inputting a probability matrix to a conditional random field for further optimization. According tothe method, a training image quantity is reduced in batches; and the efficiency and the segmentation precision are improved.

A method and apparatus for automatically performing medical image analysis tasks using deep image-to-image network (DI2IN) learning. An input medical image of a patient is received. An output image that provides a result of a target medical image analysis task on the input medical image is automatically generated using a trained deep image-to-image network (DI2IN). The trained DI2IN uses a conditional random field (CRF) energy function to estimate the output image based on the input medical image and uses a trained deep learning network to model unary and pairwise terms of the CRF energy function. The DI2IN may be trained on an image with multiple resolutions. The input image may be split into multiple parts and a separate DI2IN may be trained for each part. Furthermore, the multi-scale and multi-part schemes can be combined to train a multi-scale multi-part DI2IN.

The invention relates to an industry comment data fine grain sentiment analysis method. The industry comment data fine grain sentiment analysis method is applied to Internet data analysis and comprises obtaining comment data of e-commerce industry goods and preprocessing the comment data; establishing initial industry sentiment word libraries and computing distribution of words under different sentiment polarities through 1-gram and 2-gram; performing Chinese word segmentation on the comment data; based on the sentiment word libraries established through the 1-gram and the 2-gram, utilizing combined sentiment models to perform word modeling to obtain the probability distribution of the words which belong to different topics under different sentiment distributions; utilizing context information to re-determine the sentiment alignment of sentiment words in sentences; performing named entity identification and extracting comment characteristics through conditional random fields to compute the sentiment alignment of comment words of the comment characteristics. The industry comment data fine grain sentiment analysis method computes the sentiment of the comment words through the two dimensions of topic and sentiment to achieve fine grain sentiment analysis on the industry comment data, thereby achieving high precision and interpretability of analysis results.

The invention discloses a text sequence annotating system and method based on a Bi-LSTM (Bidirectional Long Short-Term Memory) and a CRF (Conditional Random Field). The system comprises a learning module and an annotating module, wherein the annotating module comprises a word segmenting module, a corpus annotating module and an adjusting and optimizing module; and the corpus annotating module comprises a part-of-speech annotating module and an entity recognizing module. The method comprises the following steps: preprocessing an obtained corpus; inputting the preprocessed corpus into a preset learning model; adjusting and saving parameters of the learning model; adding a corresponding prediction tag for the corpus respectively according to a sequence classifying result output by the learning model; performing word segmentation on an unknown corpus; initially annotating the unknown corpus being subjected to the word segmentation by using the adjusted learning model; adjusting and optimizing the initially-annotated unknow corpus; and finally annotating the adjusted and optimized corpus. Through adoption of the text sequence annotating system and method, a user can adjust a word library as required; a human-computer interactive adjusting function is realized; a process of automatically annotating in the same field and semi-automatically annotating in different fields is realized; the efficiency is increased; and the cost is lowered.

An Active Random Field is presented, in which a Markov Random Field (MRF) or a Conditional Random Field (CRF) is trained together with a fast inference algorithm using pairs of input and desired output and a benchmark error measure.

The invention discloses a network text named entity recognition method based on neural network probability disambiguation. The method includes: performing word segmentation on an unlabeled corpus, utilizing Word2Vec to extract a word vector, converting a sample corpus into a word feature matrix, windowing, building a deep neural network for training, adding a softmax function into an output layer of the neural network, and performing normalization to acquire a probability matrix of named entity type corresponding to each word; re-windowing the probability matrix, and utilizing a condition random field model for disambiguation to acquire final named entity annotation. A word vector increment learning method without changing structure of the neural network is provided according to the characteristic that network words and new words exist, and a probability disambiguation method is adopted to deal with the problem that network texts are nonstandard in grammatical structure and contain a lot of wrongly written or mispronounced characters, so that the method has high accuracy in network text named entity recognition tasks.

The invention discloses a Chinese word segmentation method based on two-way LSTM, CNN and CRF which improves and optimizes traditional Chinese word segmentation base on deep learning algorithm. The method comprises following specific steps: preprocessing the initial corpus, extracting corpus character feature information and pinyin feature information corresponding to characters; using the convolutional neural network to obtain pinyin feature information vector of the characters; using the word2vec model to obtain the character feature information vector of text; splicing pinyin feature vectors and character feature vectors to obtain context information vectors and put the context information vectors to a bidirectional LSTM neural network; decoding the output of the bidirectional LSTM using the linear chain condition random field to obtain the word segmentation sequence; decoding the word segmentation label sequence to obtain word segmentation results. The invention utilizes the deep neural network to extract text character features and pinyin features and combines the conditional random field decoding, can effectively extract Chinese text features and achieve good effect on Chinese word segmentation tasks.

A method reconstructs at three-dimensional (3D) real-world scene from a single two-dimensional (2D) image by identifying junctions satisfying geometric constraint of the scene based on intersecting lines, vanishing points, and vanishing lines that are orthogonal to each other. Possible layouts of the scene are generated by sampling the 2D image according to the junctions. Then, an energy function is maximized to select an optimal layout from the possible layouts. The energy function use's a conditional random field (CRF) model to evaluate the possible layouts.

The invention discloses a text mining method based on online medical question and answer information. The text mining method comprises the following steps of: extracting disease question and answer information from an obtained original webpage by adopting a network data extracting mode based on DOM and a webpage template; carrying out medical named entity identification in the extracted disease question and answer information by virtue of characteristics of a conditional random field model; and mining a medical entity relationship by virtue of the medical named entity identification. The method can be used for effectively obtaining a potential association relationship among various entities. The method is suitable for mining work of all disease classes, and has certain expandability.

The invention discloses a Chinese domain term recognition method based on mutual information and a conditional random field model. The Chinese domain term recognition method includes the following steps: (1) gathering domain text corpus and marking all the punctuations, spaces, numbers, ASSCII (American Standard Code for Information Interchange) characters and characters except Chinese characters in the corpus; (2) setting character strings and computing the mutual information values of the character strings, (3) computing the left comentropy and the right comentropy of every character string, (4) defining character string evaluation function, setting evaluation function threshold, computing the evaluation function values of every character string, determining that every character string is a word, comparing in sequence the evaluation function value of the former character with the evaluation function value of the latter character in the character string and segmenting character meaning character strings one by one, (5) utilizing conditional random fields to train a conditional random field model and recognizing domain terms with the conditional random field model. When the Chinese domain term recognition method is used to recognize terms, the data sparsity of legitimate terms is overcome, the amount of calculation of conditional random fields is reduced, and the accuracy of the Chinese domain term recognition is improved.

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

The invention provides an image salient target detection method combined with deep learning. The method is based on an improved RFCN deep convolution neural network of cross-level feature fusion, anda network model comprises two parts of basic feature extraction and cross-level feature fusion. The method comprises: firstly, using an improved deep convolution network model to extract features of an input image, and using a cross-level fusion framework for feature fusion, to generate a high-level semantic feature preliminary saliency map; then, fusing the preliminary saliency map with image bottom-layer features to perform saliency propagation and obtain structure information; finally, using a conditional random field (CRF) to optimize a saliency propagation result to obtain a final saliency map. In a PR curve graph obtain by the method, F value and MAE effect are better than those obtained by other nine algorithms. The method can improve integrity of salient target detection, and has characteristics of less background noise and high algorithm robustness.

The invention discloses a retinal blood vessel segmentation method based on the deep-learning adaptive weight, comprising: performing sample expansion on retinal blood vessel images and grouping the samples; constructing a full-convolution neural network for blood vessel segmentation; pre-training the network through the training samples; performing the global adaptive weight segmentation on the retinal blood vessel images; obtaining the initial model parameters for retinal blood vessel segmentation; adding a conditional random field layer to the last network layer and optimizing the network; using the rotation test method to input the testing samples to the network; and obtaining the retinal blood vessel segmentation result. The blood vessel segmentation full-convolution neural network structure and the adaptive weight method proposed by the invention can realize the human-eye level image segmentation and can test on two internationally published retinal image databases DRIVE and CHASE_DB1 with the average accuracy of 96.00 % and 95.17% respectively, both higher than the latest algorithm.

The invention discloses an efficient CNN-CRF network-based retina image segmentation method. According to the method, aiming at an image space information constraint problem, a full convolutional neural network and a conditional random field are combined; and aiming at a retina image blood vessel segmentation problem, a whole image is designed and an end-to-end deep learning and segmentation model is trained. Through predicting image pixels through the full convolutional neural network and segmenting semantic meanings of the conditional random field, a retina blood vessel image segmentation result is finally obtained. Compared with a pixel-by-pixel segmentation method, the method is capable of segmenting a complete image through a forward operation, so that the processing effect of the method is higher than that of the current technological level; and the method can be widely applied to the retina diagnosis fields of diabetes, hypertension and glaucoma, and provide powerful theoretical and technological support for the pathological diagnosis of retina images.

The present invention is an automated image analysis framework for cervical cancerous lesion detection. The present invention uses domain-specific diagnostic features in a probabilistic manner using conditional random fields. In addition, the present invention discloses a novel window-based performance assessment scheme for two-dimensional image analysis, which addresses the intrinsic problem of image misalignment. As a domain-specific anatomical feature, image regions corresponding to different tissue types are extracted from cervical images taken before and after the application of acetic acid during a clinical exam. The unique optical properties of each tissue type and the diagnostic relationships between neighboring regions are incorporated in the conditional random field model. The output provides information about both the tissue severity and the location of cancerous tissue in an image.

The invention relates to an evaluation object extraction method based on a domain dictionary and semantic roles and belongs to the field of natural language processing application technologies. The evaluation object extraction method based on the domain dictionary and the semantic role comprises the following steps that firstly, according to the information of the part of speech, dependency information and semantic role information, the domain dictionary DL of evaluation objects is established; secondly, the characteristics in the four aspects of words, dependency, relative positions and the semantic roles are fully extracted, model training and prediction are carried out on the DL and the characteristics through conditional random fields (CRFs), and then the extraction of the evaluation objects is completed. Compared with the prior art, according to the characteristics that the structures of Chinese sentences, especially Chinese sentences of microblogs and forum evaluation information are flexible and diverse, the constructive methods are variable, and the number of the characteristics of the sentences is small, the syntax of different levels and the semantic information are fully utilized, the advantages of the evaluation object extraction method based on rules and machine learning are also utilized, the evaluation object with a high confidence coefficient is found from a corpus automatically, rapidly and accurately, and the accuracy of extraction of the evaluation objects of the Chinese sentences is improved.

The invention relates to a knowledge graph construction method based on deep learning. The method comprises the steps that a target text statement is given, and a two-way long short term memory recurrent neural network model and a conditional random field model are used to recognize target entities in the target text statement; a context sensitive two-way long short term memory recurrent neural network model and a feedforward neural network model are used to extract the relation between every two target entities; a vector space model is used to normalize the target entities, and the normalized target entities are mapped to a concept; and a knowledge graph is constructed according to the target entities, the relation between the target entities and the concept. According to the method, the deep learning technology is applied to construction of the knowledge graph, entity recognition models of a two-way recurrent neural network and a conditional random field are adopted to recognize the target entities in the target text statement, feature engineering in the entity recognition process and the relation extraction process is reduced, the burden and trouble brought by manual design and feature adjustment are relieved, and knowledge in a text is mined precisely.

The invention provides an airborne laser point cloud classification method based on a high-order conditional random field. The airborne laser point cloud classification method specifically comprises the following steps: (1) point cloud segmentation based on DBSCAN clustering; (2) point cloud over-segmentation based on the K-means cluster; (3) construction of a point set adjacency relation based onthe Meanshift clustering; and (4) construction of a point cloud classification method of a high-order conditional random field based on the multi-level point set. The method has the advantages that:(1) a multi-layer clustering point set structure construction method is provided, and a connection relation between point sets is constructed by introducing a Meanshift point set cluster constrained by category labels, so that the categories of the point sets can be classified more accurately; (2) a multi-level point set of the non-linear point cloud number can be adaptively constructed, and information such as the structure and the shape of a point cloud target can be more completely represented; and (3) a CRF model is constructed by taking the point set as a first-order item, and higher efficiency and a classification effect are achieved, so that a higher framework is integrated, and a better effect is obtained.

An image segmentation method based on a hierarchical higher order conditional random field model is provided, comprising: firstly, extracting a multi-class texture feature for a target image, and constructing a one-variable potential function and a pairing potential function of a pixel level; then, acquiring super pixel fragments of different granularities by using a unsupervised segmentation algorithm; designing a one-variable potential function and a pairing potential function of a super pixel level corresponding to each granularity layer; constructing a hierarchical higher order conditional random field model; learning a parameter of the hierarchical higher order conditional random field model in a supervision manner by using a manual marked sample; and finally, acquiring a final segmentation marking result for a to-be-tested image by means of model reasoning. The hierarchical higher order conditional random field model used in the present invention fuses multi-feature texture information and multi-layer super pixel segmentation information of an image, and can effectively improve boundary segmentation accuracy of a multi-target object in the image.