53 results about "Word model" patented technology

The invention discloses an intelligent test paper composition and scoring method and system based on a knowledge graph. The method comprises the following steps: S1, constructing a student knowledge graph state sub-graph; s2, constructing a question bank knowledge graph; s3, on the basis of a node embedding modeling method of an improved DeepWalk algorithm, performing joint training and low-dimensional representation learning on knowledge point nodes by utilizing a continuous bag-of-words model of multi-head context aggregation, and constructing a topic adaptation degree scoring index; s4, outputting an optimal test paper scheme set; s5, collecting a current answer record when the student completes the test paper; and S6, extracting a semantic feature vector and a behavior feature vector, constructing a fusion score feature vector, and generating scores of test paper subjective question answers. According to the invention, accurate test paper composition and multi-dimensional subjective question intelligent scoring for individual knowledge states of students are realized, the method is suitable for personalized learning evaluation and teaching feedback scenes in an education evaluation platform, and the method has the advantages of high personalization, fine feedback and automatic scoring.

The invention provides a satellite denial high-altitude hovering positioning method based on visual perception and inertial measurement, which is applied to an unmanned aerial vehicle control system, and the unmanned aerial vehicle control system comprises a front-end processing module and a rear-end processing module, the front-end processing module executes the following steps that ORB feature detection and extraction are carried out on a ground environment image shot by the downward-looking binocular camera, and space coordinate information of feature points is obtained; constructing a key frame dictionary database based on a bag-of-words model, and completing loopback detection through similarity calculation of feature points; initial pose estimation is realized based on a PNP algorithm and inertial measurement unit IMU pre-integration, and an initial estimation pose sequence is generated; wherein the rear-end optimization module executes the following steps: performing real-time optimization on an initial estimation pose sequence by adopting a sliding window optimization algorithm; and solving the maximum posterior probability estimation through an L-M optimization algorithm, and determining the high-altitude hovering positioning pose of the target. And reliable guarantee is provided for stable operation of the unmanned aerial vehicle in a high-altitude complex environment.

The present application provides a text clustering method, apparatus, electronic device and computer-readable storage medium. Based on anchor words corresponding to the full-text semantics of each training text, anchor word model features corresponding to the anchor words are obtained to avoid the introduction of additional noisy features. Then, a first clustering result and a second clustering result of each training text are obtained based on the anchor word model features, and a self-training objective function of multiple training texts is determined based on each first clustering result and each second clustering result, as well as a self-training target value of the self-training objective function. Finally, the text clustering model is updated based on the self-training target value until the text clustering model converges, and the converged text clustering model is applied to text clustering. After continuous training until the convergence of the text clustering model, the accuracy and stability of the text clustering model are continuously improved, and the accuracy of text clustering is avoided from being affected by additional noisy features.

Embodiments of the present disclosure disclose a method, device and medium for optimizing data query based on quantum technology. In the method, query text data of a user at a current time and in a historical time window is collected, and a bag-of-words model is constructed after preprocessing the query text data. Feature vectors in the bag-of-words model are quantum state encoded, and a dominant feature vector of the query text data is extracted through a quantum principal component analysis algorithm. Based on the dominant feature vector, a plurality of data items that satisfy a similarity condition with the dominant feature vector are retrieved from a target database to generate a preloaded data set. The preloaded data set is loaded into a head position of an LRU cache linked list in order of similarity with the dominant feature vector. Embodiments of the present disclosure enable faster extraction of data features and identification of data features when facing large amounts of data and high-dimensional data, and faster finding of data with similar features.

The application relates to the field of artificial intelligence and discloses a text connotation quality evaluation method for improving the accuracy of text connotation quality evaluation and improving model training efficiency. The text connotation quality evaluation method comprises the following steps: obtaining an initial text from a preset medical record text; processing the initial text through a natural language processing algorithm to obtain a target text; performing text coding on the target text through a preset bag-of-words model and a preset automatic coding model to obtain first text features; performing feature extraction on the target text to obtain second text features, wherein the second text features comprise text complexity features, text grammar style features and medical semantic features; performing evaluation processing on the first text features and the second text features through a trained logistic regression model to obtain an evaluation result, wherein the evaluation result is used for identifying the connotation quality grade of the preset medical record text. The application also relates to the blockchain technology, and the target text is stored in the blockchain.

The invention discloses a power grid operation violation dictionary construction and automatic classification method and system, and relates to the technical field of operation violation classification, and the method comprises the steps: carrying out the text preprocessing of power grid operation data; constructing an initial word vector space based on Jieba word segmentation and a word bag model; considering the problem that an initial violation dictionary obtained by initial word segmentation has many semantic similar or repeated word segmentation redundancy, introducing skewness and kurtosis based on a relative frequency matrix to construct a unique index, selecting key segmented words, further constructing a classical violation dictionary, and forming a keyword vector space; and establishing a violation classifier according to the obtained keyword vector space and the corresponding violation type label, and intelligently realizing automatic classification of violation codes, violation types and problem types. Intelligent classification of violation codes, violation types and problem types is completed according to unstructured text data formed by power grid field operation violation description.

The application provides a network security event processing method and device, electronic equipment and storage medium, which belong to the technical field of network security. The method comprises the following steps: in response to receiving a monitoring instruction for a network security event, instructing a large language model to monitor the network security event based on the monitoring instruction; when receiving a monitoring result of the network security event monitored by the large language model, outputting alarm information for the monitoring result; in response to receiving user interaction information input for the alarm information, instructing the large language model to output a response strategy for the network security event based on the user interaction information; inputting the response strategy into a prompt word model to obtain a security response instruction for the network security event; and instructing the large language model to process the network security event based on the security response instruction and output a processing result for the network security event. The processing efficiency of the network security event is improved.

The application discloses a kind of unmanned aerial vehicle monocular vision inertial SLAM loop detection methods based on undergrowth trunk geometry, the method includes the following steps: step 1, trunk region segmentation and three-dimensional point cloud generation;Step 2, trunk clustering detection and position positioning;Step 3, triangular geometric feature construction and screening;Step 4, multidimensional similarity calculation;Step 5, geometric and visual joint decision and loop confirmation.This method extracts stable geometric information of trunk by driving depth data, constructs triangular features with invariance with trunk position as vertex to break through the dependence of visual texture, realizes double-path independent detection and joint decision by combining traditional visual bag-of-words model, balances detection accuracy and real-time through adaptive matching strategy, thereby effectively improves the accuracy, robustness and environmental adaptability of loop detection in undergrowth GNSS-free environment.

The application provides a task allocation model construction method and system for a heterogeneous intelligent agent, and the method comprises the following steps: determining a target heterogeneous intelligent agent in real time, and matching a target performance parameter corresponding to the target heterogeneous intelligent agent in a preset database in real time; creating a corresponding professional term vocabulary according to the target performance parameter based on a preset rule in real time, and converting the professional term vocabulary into a plurality of distributed vectors corresponding thereto through a preset continuous bag-of-words model in real time; and training a task allocation model for the target heterogeneous intelligent agent according to a preset bidirectional long short-term memory network and the plurality of distributed vectors in real time, wherein each distributed vector has uniqueness. The task allocation model can objectively and accurately complete the task allocation of the heterogeneous intelligent agent, and the work efficiency is greatly improved.

The present invention relates to a production line-oriented protocol parsing method and intelligent data collection device. The method establishes a bag-of-words model for known protocol messages, calculates the word weight of each word in the message using a word frequency-reverse file frequency algorithm, mines keywords in the message based on all word weights, calculates the word frequency of each keyword in the known protocol message, and forms a word frequency matrix. The word frequency matrix is ​​used as a feature vector of the known protocol message and inputs it into an online wide-band learning recognition model. The method then preprocesses protocol data packets captured from the network environment where the target protocol message resides, performs a similarity assessment on the unknown protocol content and the known protocol content of the preprocessed protocol data, obtains a field sequence of the unknown protocol content based on the assessment results, and adds the field sequence to an online wide-band learning template library. In this way, data interaction between different production equipment and CNC machine tools on an industrial production line is achieved, achieving cross-device and cross-system data collection on the production line.

The invention discloses a bag-of-word model loopback detection method based on a depth image, and the method specifically comprises the following steps: 1, carrying out the registration of a laser point cloud and a visual image, and calibrating the external parameters of a laser radar and a camera; 2, extracting linear edges of all object contours in the field of view, and converting all point cloud coordinates to a camera coordinate system by aligning edge features in a point cloud image of the laser radar and a visible light image of the camera to realize depth projection to obtain an initial depth image; 3, complementing the initial depth image; generating a complemented complete depth image; and step 4, for the complemented complete depth image, detecting a loop by using a bag-of-word model loop detection mechanism. The loopback detection precision is higher.

The invention relates to a text auxiliary analysis and evaluation method. The method comprises the following steps: S1, data preparation: preprocessing a text, including stop word removal, punctuation mark processing, special character processing and case and small case conversion; s2, establishing a bag-of-words model, converting a text into a word frequency vector, and weighting by using TF-IDF; s3, word embedding: converting each word into a high-dimensional vector by respectively using Word2Vec and GloVe, and then performing result comparison; s4, performing feature selection, and performing screening according to importance or correlation of features to reduce noise and overfitting; s5, selecting a learning model, and importing the features for training; and S6, predicting new text data by using the trained model. According to the method, preprocessing, feature extraction, model training and evaluation are carried out on text data through a series of steps, and finally a model capable of being used for text auxiliary analysis and evaluation is obtained.

The present invention relates to an intelligent expansion similar word model system and method thereof, which operates in a database system host and includes: a text analysis unit, which combines a plurality of keyword acoustic models and an interference sound keyword test set into a keyword forward test module; a candidate word generation unit, which generates a plurality of candidate word temporary acoustic models; a recognition rate processing unit, which generates a first candidate word acoustic model; a false call rate processing unit, which generates a second candidate word acoustic model; and an adjustment unit, which combines a plurality of keyword acoustic models with the second candidate word acoustic model into a similar word acoustic model.

The invention relates to a visual repositioning method and electronic equipment. The method comprises the following steps: acquiring a current frame image of a vehicle for repositioning; determining at least one candidate image from each mapping image of a visual map according to the global bag-of-word vector of the current frame image and the global bag-of-word vector of each mapping image of the visual map; determining a distribution word bag vector and a depth word bag vector of the current frame image; according to the distribution word bag vector and the depth word bag vector of the current frame image and the distribution word bag vector and the depth word bag vector of the candidate image, determining the real-time pose of the vehicle; wherein the global bag-of-words vectors represent matching conditions of all feature points in the corresponding image and feature points in the bag-of-words model; the distribution word bag vectors represent matching conditions of feature points of different areas in the corresponding image and feature points in the word bag model, and the depth word bag vectors represent depth information of the feature points of different areas in the corresponding image. According to the invention, the repositioning precision is improved.

This invention provides a method for constructing a hierarchical classification model for government procurement items. The method first constructs and encodes a hierarchical structure for the procurement items. Then, it performs text cleaning and word segmentation on the procurement project names, using a continuous bag-of-words model to train the corpus to obtain word vectors, or directly obtaining word vectors using a BERT Chinese pre-trained model, thus obtaining an initial text representation. Finally, the text representation and label representation are fed into the model for training. The three-layer network contained in the HA-BiGRU model can effectively solve the above problems. The text encoding layer with a BiGRU network as the encoder can further extract semantic information from the context; the hierarchical attention recurrent network layer can model the dependencies between layers and enhance the association between text and labels through a text-label attention module; the hybrid prediction layer integrates the local and global losses of hierarchical labels, optimizing the model by reducing the total loss.

The application is suitable for the field of automatic driving technology, and provides a dynamic scene autonomous positioning and mapping method fusing a bag-of-words model, comprising the following steps: recording environment data; constructing a point cloud map and aligning the map to a geodetic coordinate system, selecting a visual key frame and saving the corresponding pose; removing dynamic features in the visual key frame to generate a bag-of-words vector, constructing a mapping relationship between the bag-of-words vector and the pose, and saving the bag-of-words model; in the positioning process, preprocessing the constructed point cloud map, generating a bag-of-words vector by using static features of the camera collected image, searching in the constructed bag-of-words model, and mapping out the initial position of the vehicle; after completing the positioning initialization, performing registration, providing a registration initial value by fusing the bag-of-words model and GPS / IMU, and outputting the pose information of the vehicle in real time. The application solves the problem that the SLAM framework cannot obtain absolute pose and the problem that the autonomous parking positioning information is missing due to the shielding of signals such as GPS.

This invention relates to a fine-grained webpage identification method for complex network environments. First, webpage access traffic is segmented into five-tuples, and the segmented traffic is grouped using the SNI information contained in the data packets. Then, the data packets in specific groups are reassembled into TLS fragments, and their length information is extracted. During multiple accesses to the same webpage, ADUs representing webpage characteristics are repeatedly requested. By statistically analyzing the frequent items in the TLS fragment length information, feature vectors characterizing the webpage are constructed; these feature vectors are considered webpage fingerprints. This invention utilizes clustering to correct numerical fluctuations in feature vectors and designs a bag-of-words model to correct sequential fluctuations in feature vectors. Finally, the corrected feature vectors are input into a machine learning model for training, resulting in a classifier capable of accurately identifying webpages. This method can obtain stable fine-grained webpage fingerprints in complex network environments, enabling accurate identification of harmful webpage access behavior.

The invention discloses a subway construction risk classification method based on LDA topic modeling, and relates to the technical field of engineering risk management. Comprising the steps of exporting a hidden danger text record from a construction unit safety patrol system, obtaining initial text content, carrying out regularization processing to obtain cleaned text content, then constructing a dictionary, inputting the dictionary into a word segmentation tool, obtaining a word segmentation result, constructing a word bag model, calculating TF-IDF weight in combination with the word segmentation result, forming a corpus vector set, and determining LDA hyper-parameters. Inputting the corpus vector set and the LDA hyper-parameters into an LDA model for training to obtain a trained LDA model, inputting a to-be-detected text into the LDA model to obtain topic keywords and text topic distribution, and performing risk classification to form a structured risk text database. Structured and standardized classification of hidden danger information can be realized, and the accuracy and comprehensiveness of risk identification are improved.

The invention provides an analysis-free log anomaly detection method and device based on space-time semantic features, and the method comprises the steps: converting labeled log data into structured vector representation based on a preset bag-of-words model, and carrying out the clustering and grouping of the vector representation through a clustering algorithm, and calculating the feature value of each word in each group by adopting word frequency-inverse document frequency to form semantic feature representation of each log data, so that the extracted features can pay attention to local features. To-be-analyzed log data is matched to a target cluster based on the clustering algorithm, and semantic feature representation of the to-be-analyzed log data is calculated; and finally, calculating a normalized compression distance between the to-be-analyzed semantic feature representation and the labeled semantic feature representation, carrying out classification based on a KNN model, and determining a label of the to-be-analyzed log data. The method does not depend on a complex model structure and a training process, does not need additional training parameters, and has better generalization and applicability.

The present invention discloses a tree species identification method based on bag-of-visual-words representation and state-space modeling, which relates to the technical field of tree species cross-section microscopic image recognition and is characterized in that it includes the following steps: S1: data set construction; S2: local feature extraction using a bag-of-visual-words model; S3: global feature extraction using a state-space model; S4: multi-level feature fusion; S5: data oversampling and classifier training; S6: tree species identification. The technical problem to be solved by the present invention is to provide a tree species identification method based on bag-of-visual-words representation and state-space modeling, which introduces a state-space model to model the feature sequence of tree species images, mines their long-term dependencies, and extracts global feature information. By fusing local and global features, a multi-level feature representation with strong discriminative ability is constructed. The synthetic minority class oversampling technology is used to balance the samples to improve the discriminative ability of the support vector machine classifier.

The invention relates to the technical field of text processing, in particular to an intelligent power data classification method and system based on a hybrid coding mode and a textRCNN model. The method comprises the following steps: data preprocessing: cleaning and standardizing original power data; multiple codes are generated, multiple coding modes (including a bag-of-words model, TF-IDF coding, word embedding and one-hot coding) are used for coding the preprocessed data, and corresponding vector representation is generated; vector splicing: the vectors generated by different coding modes are spliced into a new comprehensive vector, and compared with a traditional coding mode, the characteristics are richer; constructing a TextRCNN model and training the TextRCNN model, and training the TextRCNN model by using the spliced comprehensive vector as input; after training is completed, a comprehensive vector is generated from the data needing to be classified in the mode, and a final classification result is obtained through model processing; the method shows higher accuracy and robustness in the task of power data classification.

The invention discloses a map construction and relocation method in a complex dynamic scene, a medium and a product, and relates to the field of map construction and relocation, and the method comprises the steps: obtaining an original image and a to-be-located image; based on the original image, a semantic segmentation result of the original image is obtained by using a deep learning model; determining angular points based on the original image; based on the angular points and the original image, extracting descriptors; constructing a map based on the semantic segmentation result, the angular points and the descriptors; based on the map, adopting a map optimization method to obtain an optimized map; obtaining clustering result information and key frame information of the image features based on the optimized map by adopting a DBoW3 library; constructing a dictionary based on clustering result information of the image features, and storing key frame information in a database; and based on the optimized map, dictionary and database, the to-be-positioned image is repositioned by adopting a bag-of-word model matching method, and accurate map construction and accurate and rapid positioning can be realized in a complex dynamic scene.

Embodiments of the present application relate to a positioning method and device of a mobile device, an electronic device and a storage medium. A key frame is acquired by a vision camera installed on the mobile device; mapping information and a feature descriptor of the key frame are determined; a bag-of-words model and a key frame image database are obtained according to the feature descriptor of the key frame; a current image frame is acquired, and the current image frame is converted into a current frame bag-of-words vector based on the bag-of-words model; a candidate key frame with the highest matching score with the current image frame is determined from the key frame image database according to the current frame bag-of-words vector, and a target key frame is determined according to the candidate key frame; a target node index corresponding to the target key frame is determined according to the mapping information, and pose information of the mobile device is calculated according to the target node index, so as to position the mobile device; that is, the prior pose of the mobile device is provided by the bag-of-words model, and the positioning accuracy is improved.

The invention discloses a saliency-point-line-based back-end optimization and loopback detection method in automatic driving, and the method comprises the steps: S1, obtaining an rBRIEF descriptor of a point feature and an LBD descriptor of a line feature from an image, and carrying out the key frame screening of the features from a time constraint or a key frame number constraint; s2, constructing a bag-of-words model based on saliency point lines, then determining the number and depth of branches constructed by a bag-of-words tree, and performing clustering by using K-means + + to generate a dictionary file; s3, constructing a vocabulary vector packet representing each key frame according to a point-line feature weight adjusting method of the saliency region, and performing similarity calculation according to a similarity calculation method of weighting point-line features by the saliency region; and S4, when the similarity of the two image frames is greater than a preset threshold value, carrying out loopback detection. According to the method, the line features are fused in the word bag model based on the point features, a similarity calculation method for weighting the point and line features by the saliency region is provided, the calculation efficiency is improved, and wrong matching of similar scenes is reduced.

The present invention discloses a domain word expansion method and system based on a pre-trained model. The method is based on a word model pre-trained with massive external general corpus, designs and superimposes an internal flow text word model fine-tuning algorithm, and is supplemented by a natural language processing tool flow to achieve word understanding of a small amount of industry-specific text and improve the efficiency of artificial experience word library construction, effectively improving the recognition ability of non-operating income flow. In this way, keywords related to non-operating income can be automatically mined from the flow data, achieving the expansion of operating flow keywords, improving the recognition of non-operating flow, and having high accuracy and generalization ability, thereby effectively portraying the user's operating ability.

Embodiments of the present application provide a relationship extraction method and device, the method comprising obtaining a first entity set and a first text feature from a corpus corresponding to a first target field, and mapping the first entity set and the first text feature into first vector data through a bag-of-words model, then inputting the first vector data into M classification models respectively to obtain M prediction results, and determining a final relationship of each first entity pair in the first entity set from N candidate relationships according to the M prediction results, each prediction result comprising a prediction probability of an inter-entity relationship of each first entity pair in the first entity set being respectively one of the N candidate relationships, and each classification model being obtained by training an initial model according to a second entity vector with a second relationship vector label and a second text vector of a corresponding source knowledge base. In this way, using transfer learning as a relationship extraction method, the knowledge of the source field is migrated to the target field, and the required professional annotation manpower can be reduced.

The present invention discloses a method, system and platform for optimizing the recommendation of enterprise collective procurement plans. User procurement behavior big data is collected through terminals, and after feature analysis, similar procurement behavior data is integrated to form a historical procurement plan. Plan text data is generated based on the historical procurement plan, and the bag-of-words model is combined to perform text segmentation, unified vocabulary generation and text vector conversion, and a relationship table of object keywords is constructed by vector similarity comparison. Based on the relationship table, a knowledge graph is constructed in combination with procurement object attribute data, and rule reasoning and relationship expansion are performed through the AMIE+ algorithm. During the interaction cycle, user data is collected in real time, real-time object keywords and their semantic weights are analyzed, and a user behavior matrix and an associated behavior matrix are constructed. The effective associated matrix is screened out by similarity calculation between matrices, and the associated recommended procurement objects are parsed and marked to generate procurement recommendation information.

The invention belongs to the technical field of voice recognition, and particularly relates to a voice wake-up word user-defined setting method, system and device and voice wake-up device.The method comprises the steps that wake-up word text information input by a user is received, the user is prompted to record and store voice information corresponding to the text information, and the voice information is stored; data analysis and optimization are carried out on the recorded voice information, then multi-dimensional acoustic feature extraction is carried out, a user specific voiceprint model is generated, the voiceprint model is verified, and the voiceprint model is stored after being verified to be qualified; and prompting the user to perform wake-up inspection on the setting result, and outputting corresponding prompt content according to the inspection result. According to the method, rapid modeling and efficient recognition of the user-defined wake-up word can be realized; the voiceprint binding is realized by generating the specific voiceprint model of the user, so that the awakening accuracy is improved, and false triggering is reduced; meanwhile, the user-defined wake-up word model can be quickly generated without collecting a large number of real person voice samples, the research and development cost is remarkably reduced, and the development period is shortened.

This invention relates to a heuristic-based method for generating fault patterns in the aviation field, comprising: S1, extracting features from aviation fault text data using a bag-of-words model and the term frequency-inverse text frequency index method; S2, performing clustering category analysis of aviation fault text data based on k-means heuristics; S3, detecting outliers, extracting faults, and concatenating faults in the aviation fault text data to obtain fault pattern names; and S4, processing the obtained fault description text in real time to generate aviation fault patterns. This invention completes text data clustering category analysis using a bag-of-words model, the term frequency-inverse text frequency index method, and the k-means heuristic method. Furthermore, it obtains fault pattern names through outlier detection, fault extraction, and concatenation, enabling real-time detection of fault text data. By periodically re-clustering, it can discover new fault patterns by utilizing existing prior knowledge, thereby improving the accuracy and effectiveness of fault pattern detection.

The present invention discloses a high-resolution DEM semantic object recognition method using a bag-of-words model, which belongs to the field of geospatial data processing technology and specifically includes the following steps: step 1, landform feature vector expression, step 2, landform bag-of-words model construction, step 3, target landform semantic mapping, and step 4, landform object classification and recognition. In the present invention, landform variables are extracted from a high-resolution DEM data set to generate a landform feature vector, and then an external open data source is used to enrich landform semantic information to form a landform bag-of-words and generate a landform weighted feature vector. Secondly, a mapping is formed between target landform feature elements and high-level concepts in the bag-of-words. Finally, the weighted features of the target data and the weighted features of the training data are compared to realize landform recognition. By integrating landform variables, based on regional features and high-level landform descriptions, automatic landform recognition is realized, significantly improving the recognition accuracy and efficiency of derived landform objects.