36results about How to "Improve labeling ability" patented technology

The invention discloses a semantic role labeling method based on a synergetic neural network, and relates to the fields of semantic role labeling, mode identification and synergetic neural networks, in particular to a method for introducing the principle of the synergetic neural network into shallow semantic analysis. The semantic role labeling method comprises the following steps: extracting characteristics from training language material and testing language material and constructing corresponding semantic characteristic vectors; performing kernel transformation on the semantic characteristic vectors and constructing a prototype pattern and a mode to be tested on the basis; constructing an order parameter and calculating a plurality of candidate roles for each dependent component; constructing a predicate base and combining the candidate roles of all the dependent components corresponding each predicate to get role chains of all the predicates; and optimizing a network parameter, performing dynamic evolution on the synergetic neural network to get an optimal role chain, and outputting the labeling mode. The principle of the synergetic neural network is firstly introduced into the semantic role labeling, and the method can be widely applicable to various natural language processing tasks. The semantic role labeling method has better application prospects and application value.

The invention provides an annotation method, device and system. The method comprises the steps that an annotation signal sent by annotation input equipment is received, coordinate data in the annotation signal is analyzed, and a display region corresponding to the annotation signal is determined based on the coordinate data; signal processing is performed on the annotation signal according to a display strategy of an annotation of the display region; and the processed annotation signal is sent to the display region for display. Through the scheme, the instruction effect of the annotation signal can be improved.

The invention discloses a method for marking picture semantics based on a Gauss mixture model, which belongs to the technical field of image retrieval and automatic image marking. The method comprises the following steps: S1, obtaining a relationship between a low-level visual feature of the image and a semantics concept through monitoring Bayesian learning, and obtaining an image feature set; S2, establishing two Gauss mixture models for each semantics concept by means of an expectation-maximization algorithm, and adding a step of eliminating a noise area; and S3. according to the image feature set, calculating the color posterior probability of the pattern posterior probability of an area layer, arranging the calculated posterior probabilities which belong to all concepts of the image according to a descending order, and obtaining the color ordering value of each concept; similarly, arranging the pattern posterior probabilities and obtaining the pattern ordering value of each concept; and selecting a concept class marking image with a least summation of front R ordering values. According to the method of the invention, the difference between the low-level visual feature of the image and the high-level semantics concept expression is remarkably reduced, thereby effectively settling a semantic gap problem.

The present invention discloses a feature selection based multi-example multi-tag learning method and system. The method comprises: mapping all features of a packet in a known data set into a projection example reference space consisting of all examples in the known data set, and obtaining a feature vector of the packet; removing projection examples corresponding to the features of the packet with an invalidly marked tag in the projection example reference space by adopting a feature selection method based on l2 and 1 norm constraints, and further obtaining a representative projection example reference space; mapping the features of the packet into the representative projection example reference space, and obtaining a new feature vector of the packet; and constructing a linear decision function according to the new feature vector of the packet, and training a classifier based on tag correlation by adopting an optimal algorithm based on the tag correlation. By using the method provided by the present invention, the classifier based on the tag correlation is obtained by learning by utilizing the known database, so that a tag set of unknown samples is predicted and the accuracy of tag prediction is improved.

The invention discloses a sequence labeling method based on a multi-head self-attention mechanism, which comprises the following steps: step 1, local context semantic coding: learning local context semantic representation of words in a text by using BLSTM serialization, step 2, global semantic coding: based on the local context semantic representation of the words coded in the first step, coding global semantic representation of the words through a multi-head self-attention mechanism; step 3, semantic feature fusion: fusing the local context semantic representation encoded in the step 1 and the global semantic representation encoded in the step 2, and taking a fusion result as the input semantic feature of the step 4; step 4, sequence labeling: in order to fully consider the dependency relationship between labels in a sequence labeling task, utilizing CRF to predict the labels; step 5, performing model training; step 6, performing model reasoning. On the basis of the recurrent neural network, a multi-head self-attention mechanism is further introduced to learn global semantic representation of words, and therefore the sequence labeling effect is improved.

The invention discloses a semi-automatic word segmentation corpus labeling and training device, which aims to overcome the defects of the corpora used during the word segmentation corpus labeling and training process. The device of the invention is realized through the following technical schemes of using a text corpus annotation preparation module for managing the to-be-annotated corpora and the segmented word corpora; based on a plurality of word segmentation algorithms, such as the bidirectional maximum matching word segmentation based on an integrated dictionary, CRF, JIEBA, etc., submitting the word segmentation annotation work of the raw corpus to a semi-automatic corpus word segmentation annotation module; creating the segmented word tagging tasks, selecting a labeling applicable algorithm model, carrying out the automatic annotations, on the basis of automatic labeling result fusion, feeding back a training model corpus and a labeling model generated by the text corpus labeling preparation module to the feedback model learning training module; selecting and carrying out model learning training, calling a unified training model interface to generate a core dictionary, updating a word segmentation training model table, establishing a labeling algorithm comprehensive evaluation model to evaluate a model labeling effect, so that a new word segmentation labeling task is completed.

The invention discloses an automatic image marking method based on work correlation. A training set T comprises l images, n marking words are marked on each image of the training set T, the training set T is provided with a corresponding vision lemma, and the image to be marked is I. The method includes the steps that a semantic vector of each marking word w is calculated according to a formula, the marking word w is represented by the vector form w=<v1, v2,......, vm>, ci is an associated word in a context, and the associated words in the context total m; semantic similarity of the marking words is calculated according to a formula, and vector module calculated is achieved as is shown in the specification; p(A) is calculated according to the formula, wherein A is a marking word group in w1, w2,......wn, and n is the number of the marking word groups; the conditional probability p(I / wi) is calculated according to a formula; the marking word group A of the image I to be marked is calculated according to the formula A=arg maxAp(I / A) p(A).

The invention provides a Russian semantic role labeling method, system and device and a storage medium. Aiming at the characteristics of Russian, the relation between predicates and arguments is reasonably utilized, the semantic role annotation of the Russian can be well realized, the semantic role annotation accuracy is improved, and higher annotation performance is obtained. The method comprisesthe following steps: 1, preprocessing corpora, extracting classification characteristics, and converting the classification characteristics into characteristic vectors; 2, constructing classificationmodels based on neural networks of different architectures, and inputting classification features into the classification models for training to obtain trained classification models; 3, based on a voting fusion mechanism, fusing the trained classification models according to the principle that a small number obeys a majority to obtain a fusion model; 4, inputting the preprocessed corpus into a fusion model, identifying a semantic role, attaching a prediction label, and performing performance evaluation on an obtained semantic role prediction result.

The invention discloses an automatic labeling method for multi-view images. The automatic labeling method for multi-view images comprises the following steps: (1) setting semantic labels of labeled images and various visual features as multiple views, inputting the multiple views into a multi-view sparse model and carrying out training learning to obtain various view dictionaries and various viewweight value factors; (2) inputting multiple visual features of images to be labeled; (3) reconstituting the images to be labeled sparsely by the various view dictionaries and the various view weightvalue factors, and calculating to obtain sparse reconstituted coefficients of label views; (4) multiplying the label view dictionaries with the sparse reconstituted coefficients of the label views toobtain the scores of the semantic labels of the images to be labeled; and (5) arranging the scores from high to low, and labeling the images to be labeled by using top five semantic labels. The automatic image labeling performance of a computer is improved, and precision ratio and recall ratio of automatic labeling are improved.

The invention provides an image automatic annotation method and device based on a decision tree. The device comprises an input unit, a pre-processing unit, a segmentation and extraction unit, an annotation probability calculation unit, a spanning tree unit, an inter-word correlation calculation unit and a keyword selection unit. Compared with the prior art, the method of the present invention hasthe advantages that some obtained blurred images are repaired, thus the image semantic automatic annotation technology is applied to a wider scope, the underlying features of the images are more comprehensively extracted, global features and local features are used to reflect the true visual content of the images, the accuracy of image semantics automatic annotation is improved, scale-invariant features of principal component analysis are adopted by the global features, the computational efficiency is improved, especially for high-dimensional images, the unmeasurable nature of massive image sets is solved, an image annotation problem is converted into a classification problem to carry out annotation, and the annotation performance of a traditional model is improved.

The invention provides a contract labeling method and device, and relates to the technical field of artificial intelligence, and the method comprises the steps: extracting at least one contract samplefrom contracts of each service type, and obtaining an initial sample set; constructing and training an initial annotation model based on the contract sample of each service type; obtaining a plurality of pre-stored contracts of each service type, and dividing the contracts into a sample expansion set and a test set; marking contract elements in contracts in the sample expansion set by utilizing an initial marking model; combining the labeled sample expansion set and the initial sample set into a training sample set, and optimizing and training an initial labeling model by utilizing the training sample set to obtain a labeling model; inputting the test set into an annotation model, and obtaining an annotation result of a contract in the test set output by the annotation model; and judgingwhether the annotation model needs to be continuously optimized or not according to the annotation result of the test set. According to the technical scheme provided by the embodiment of the invention, the problem of low marking accuracy of contract elements in the prior art can be solved.

The invention provides a sequence labeling method and device, a storage medium and computer equipment, and a sequence comprises a to-be-labeled word and a labeled word, and the sequence is used for generating a text. The method comprises the steps: obtaining the sequence; identifying context information of words to be labeled in the sequence; according to the context information, determining a second part-of-speech of the to-be-annotated word in combination with the first part-of-speech of the annotated word adjacent to the to-be-annotated word, wherein the second part-of-speech is used for annotating the to-be-annotated word. According to the method, the part-of-speech of the to-be-annotated word is annotated according to the context information of the to-be-annotated word in the sequence, the part-of-speech annotation accuracy can be improved, the sequence annotation effect is improved, and therefore text generation is effectively assisted.

The invention relates to an image annotation method for regional context deep learning, and aims to solve the problem that most of current universal image annotation methods are based on visual features, especially depth features of images, but when semantic keywords needing to be annotated are more and scattered, complex image annotation cannot be effectively solved only by means of the visual features. The invention provides an image annotation method based on thea regional context relation deep learning network. The method comprises the steps of automatically learning a context relationshipbetween regions in an image, fusing the context relationship into feature expression of the regions, accumulating features of the regions to serve as features of the image, and finally learning a multi-class support vector machine to perform semantic recognition on each image and sort contribution values of semantics to obtain annotation keywords. The method is flexible and convenient, can automatically learn the regional context relation of the image while achieving the image labeling, and is higher in practicality.

The invention discloses a sequence labeling method and device, a training method and equipment of a sequence labeling model and a computer readable storage medium. In the scheme, a score layer of a sequence labeling model comprises second score layers which are in one-to-one correspondence with labeling specifications. The invention comprises a first score layer corresponding to all the labeling specifications. Due to the unique design of the score layer in the model, heterogeneous data of multiple annotation specifications can be used as a training set of the model, the scale of training corpora is expanded, and the model can learn the generality of corpora of different annotation specifications, so that the annotation performance of the model under a single annotation specification is improved. Besides, the output result of the model is a binding label sequence, equivalently, the label sequence under various labeling specifications is directly obtained, and conversion of the text between different labeling specifications is facilitated.

The present invention proposes an efficient labeling method combining laser point cloud and image, which performs initial external parameter automatic calibration through planar checkerboard target image data and laser point cloud data, realizes pre-labeling through automatic segmentation algorithm, and combines a small amount of manual intervention to check and correct The image labeling information is further refined, and the 3D laser point cloud corresponding to the image labeling object is determined by back projection, and then the accurate 3D point cloud of the target to be marked is obtained by re-segmentation clustering and growth, and finally through the precisely matched 3D point cloud Cloud and image calibration objects are further optimized for external parameters; the efficient labeling method of the joint laser point cloud and image of the present invention does not require a lot of manual intervention, reduces the difficulty of laser point cloud labeling, improves labeling efficiency, and has higher labeling precision. Not only can the point-by-point category information of the laser point cloud be obtained, but also new labeling data such as joint labeling information of image and laser point cloud object level can be obtained.

The invention discloses a sample acquisition and rapid labeling method with a relatively fixed target state, and the method comprises the following steps: firstly, arranging a camera for shooting a target; secondly, connecting a picture acquisition device in communication way with each camera and obtaining pictures shot by each camera; then, enabling the picture acquisition equipment to configure all target information, camera information, an associated target, a camera and camera preset position information; then, enabling the picture acquisition equipment to start an automatic snapshot tool to generate a sample picture and a pre-annotation file; and finally, manually checking the automatically captured sample pictures, carrying out batch secondary labeling on the pictures of which the target states are changed, and finally completing sample collection and rapid labeling. According to the method, the problems of difficulty in sample collection and time-consuming and labor-consuming sample labeling in sample training in the field of deep learning image recognition are solved, the samples can be quickly acquired and labeled, and the efficiency of image recognition research or engineering implementation is greatly improved.

The invention discloses a traditional Chinese medicine literature content analysis method and device. The method comprises the following steps: preprocessing an obtained classical Chinese text to obtain unsupervised pre-training data to pre-train a selected large-scale language model Bert; combining the pre-trained model Bert with a conditional random field model to obtain a sequence labeling model; training the obtained sequence labeling model by using the labeled traditional Chinese medicine literature content analysis data; segmenting each paragraph of the to-be-analyzed traditional Chinese medicine literature into clauses, inputting the clauses into the sequence labeling model to obtain a coding sequence of each clause, and generating a probability distribution sequence of a tag to which the corresponding clause belongs according to the coding sequence of the clause; inputting the probability distribution sequence of the clauses into a conditional random field model to obtain the probability that the sequence of the clauses is labeled as different tag sequences; and selecting the tag sequence with the maximum probability as a prediction result, combining adjacent clauses predicted as the same tag, and connecting paragraphs of the literature to obtain a content analysis result of the traditional Chinese medicine literature.

The embodiment of the invention provides a map labeling method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining pixel coordinates of a contact in a target map container; calculating actual latitude and longitude coordinates of the contact in a target map according to the pixel coordinates; and based on the actual latitude and longitude coordinates, selecting a preset identifier to render the contact. According to the invention, the pixel coordinates of the contact in the target map container can be determined, the pixel coordinates are converted into the actual latitude and longitude coordinates, and the contact is rendered based on the actual latitude and longitude coordinates and the preset identifier, so that the target marker can be rendered on the target map, and the user can conveniently mark the target mark object which needs to be marked on the target map according to own requirements. In addition, the labeling effect of labeling the target labeling object on the target map is improved.

The invention discloses a semantic role labeling method based on a synergetic neural network, and relates to the fields of semantic role labeling, mode identification and synergetic neural networks, in particular to a method for introducing the principle of the synergetic neural network into shallow semantic analysis. The semantic role labeling method comprises the following steps: extracting characteristics from training language material and testing language material and constructing corresponding semantic characteristic vectors; performing kernel transformation on the semantic characteristic vectors and constructing a prototype pattern and a mode to be tested on the basis; constructing an order parameter and calculating a plurality of candidate roles for each dependent component; constructing a predicate base and combining the candidate roles of all the dependent components corresponding each predicate to get role chains of all the predicates; and optimizing a network parameter, performing dynamic evolution on the synergetic neural network to get an optimal role chain, and outputting the labeling mode. The principle of the synergetic neural network is firstly introduced into the semantic role labeling, and the method can be widely applicable to various natural language processing tasks. The semantic role labeling method has better application prospects and application value.

The invention provides a conversation error correction method and device and electronic equipment, and the method comprises the steps: carrying out the hidden layer weighting of a sentence through a word granularity language model, and obtaining a hidden layer weighting combination vector of the sentence; then, carrying out grammar diagnosis on the hidden layer weighted combination vector of the sentence; obtaining a grammar diagnosis result of the sentence; according to the embodiment of the invention, the word granularity language model is obtained by training a large amount of unlabeled data; the purpose of the invention is to enhance the grammatical diagnosis correctness of a grammatical diagnosis model with less annotation data, and the learning training word granularity language model can effectively improve the annotation effect and generalization ability of the grammatical diagnosis model, and has feasibility and usability in practice.

The invention relates to the technical field of bioinformation, and discloses a marking method and system for protein functions. Therefore, protein marking performance is improved, expensive cost of a bioexperiment method and poor efficiency are solved. The method comprises following steps: estimating the first possibility of a certain function in a to-be-inquired protein according to a first-stage structure neighborhood and a second-stage structure neighborhood; estimating the second possibility of the certain function in the to-be-inquired protein according to all homologous sequences; inputting a PSSM matrix of the to-be-inquired protein into an SVM prediction model to obtain the third possibility of the certain function in the to-be-inquired protein; converting the distribution of the function corresponding to other species according to the gene co-expression fraction into the fourth possibility of the function occurring in a target species in the to-be-inquired protein; and mixing the first possibility, the second possibility, the third possibility and the fourth possibility to estimate the comprehensive possibility of the function in the to-be-inquired protein.

A system for automatically generating an animation scene comprises a data preprocessing unit, a knowledge graph construction unit, a text analysis unit and a scene generation unit, performs statistical classification on a natural environment and an artificial environment, selects a scene most conforming to a plot and defines constraint rules at different times according to the natural environment and the artificial environment, the change of the same scene at different time is displayed in this way; the orientation relation of different entities is defined, the scene is refined in sequence, the whole background is generated by splicing all the entities, the problem that a material library does not completely meet plot requirements is solved, and the scheme of automatically generating the scene under the condition that materials are missing and incomplete can be achieved.

The invention provides a method and system for generating an ancient Chinese annotation model. The method comprises the following steps: S1, training and forming a language model capable of representing context semantics; s2, constructing a multi-task joint learning model; and S3, simultaneously training the language model and the multi-task joint learning model, wherein the language model and the multi-task joint learning model form an ancient Chinese annotation model. According to the invention, punctuations, quotations, book names and entities can be marked at the same time, the F1 index on tasks such as automatic punctuation, book name and entity identification reaches more than 90%, and an ancient Chinese information marking effect similar to that of manual work is realized.

The invention discloses a sample acquisition and fast labeling method with a relatively fixed target state, comprising the following steps: firstly, arranging cameras for shooting the target; secondly, a picture acquisition device is communicated and connected with each camera, and obtains pictures taken by each camera; , the picture acquisition device configures all target information, camera information, associated targets and cameras and camera preset position information; then, the picture acquisition device starts the automatic capture tool to generate sample pictures and pre-labeled files; finally, manually check the automatically captured sample pictures , batch secondary labeling of pictures whose target state has changed, and finally complete sample collection and rapid labeling. The invention solves the problems of difficulty in sample collection and time-consuming and laborious sample labeling in sample training in the field of deep learning image recognition, can quickly acquire and label samples, and greatly improves the efficiency of image recognition research or engineering implementation.

The invention discloses a sensitive word recognition method and device, equipment, a storage medium and a program product, and the method comprises the steps: determining a word set of a to-be-recognized text based on a pre-generated domain dictionary library, each word in the word set comprising head position information and tail position information; performing character construction component splitting on each word in the word set to obtain a character construction component corresponding to each word; word vectors corresponding to the words and word vectors corresponding to the word components and word component vectors corresponding to the word components are obtained; generating input vectors of the words based on the word vectors of the words and the word construction component vectors; inputting the head position information, the tail position information and the input vector of each word in the word set into a pre-generated sequence labeling model, and determining a labeling result of each word by the sequence labeling model based on the head position information, the tail position information and the input vector; and recognizing the sensitive word according to the labeling result of each word so as to improve the recognition accuracy of the sensitive word.

The invention belongs to the field of computer application and computational vision, and relates to an image automatic tagging algorithm based on semantic scene classification. This method uses the method based on non-negative matrix decomposition to detect the semantic scene information of the label, maps the training set samples to the corresponding scenes in a probabilistic manner, and uses the scene information of the samples to train the scene classifier based on the extreme learning machine and the differential evolution algorithm. Finally, the scene classifier is used to quickly map the samples to be labeled to a sample subset related to its scene, and the KNN-based algorithm is used to complete the labeling in this sample subset. The invention not only narrows the scope of searching for the nearest neighbor samples and improves the efficiency of the algorithm, but also enables the KNN algorithm to label semantically related sample sets, thereby reducing noise interference and improving the labeling effect. The number of scenes in this method is far less than the number of labels, so it solves the problem that the method based on model learning is not suitable for data sets with a large number of labels.

The invention relates to a meta-learning-based child-bearing case automatic labeling system and method, and the system comprises a preprocessing module which is used for processing a received child-bearing case text to obtain a child-bearing case statement; a problem description statement recognition module which is used for receiving the childbearing case statement, performing recognition by calling a first model, and generating a to-be-labeled statement; an influence factor classification module which is used for receiving to-be-labeled statements, calling a second model for classification, and obtaining influence factor category information to which the statements belong; and a specific annotation category classification module which is used for receiving the to-be-annotated statement with the influence factor category information, calling a classifier corresponding to the influence factor category, and generating a specific annotation category of the to-be-annotated statement. According to the method, rapid and automatic labeling can be performed on the childbearing cases, and the labeling efficiency and labeling accuracy of childbearing case data can be improved conveniently.

The embodiment of the invention discloses a data annotation method which comprises the steps that image data and point cloud data are acquired, annotation targets in the image data comprise annotation targets in a first annotation target set and annotation targets in a second annotation target set, and the annotation targets in the first annotation target set are annotation targets in the point cloud data; determining a three-dimensional envelope of each labeled target in the first labeled target set in the laser radar coordinate system; according to the conversion relation between the camera coordinate system and the laser radar coordinate system and the three-dimensional envelope, obtaining a two-dimensional labeling result of the first labeling target set in the pixel coordinate system; determining a two-dimensional labeling result of the second labeling target set in the pixel coordinate system; determining a three-dimensional labeling result of the second labeling target set in the laser point cloud coordinate system according to a conversion relation between the pixel coordinate system and the laser point cloud coordinate system and the two-dimensional labeling result of the second labeling target set; the method and the device are used for improving the labeling capability of joint labeling on a long-distance detection target.

The invention discloses an automatic video annotation method based on automatic classification and keyword marking. The automatic video annotation method comprises the following steps of: S1, carrying out preprocessing on a video classification feature; S2, extracting the global feature and the local feature of a video, wherein the global feature is used for training an SVM (Support Vector Machine) model to enable the SVM model to identify different types; and the local feature is used for establishing a multi-feature index model with multiple features which correspond to keywords; and S3, for un-annotated videos from a user, extracting the global feature and the local feature and then enabling the SVM model to identify specific types of the videos by adopting the global feature; and retrieving relevant keywords in the multi-feature index model for annotating by using the local feature; and S4, ordering an annotation result according to a certain weight and then returning the annotation result to the user. According to the automatic video annotation method, the marking performance of the video is improved.