8results about How to "Improve labeling accuracy" patented technology

The invention provides a data annotation system and method based on a large language model, a medium and a terminal, and the system comprises a data access layer which is used for collecting to-be-annotated data, and carrying out the preprocessing of the to-be-annotated data, so as to enable the to-be-annotated data to be converted into a unified data format; the intelligent annotation engine layer is used for retrieving context information of a vertical field for the received to-be-annotated data in the unified data format, splicing the retrieved context information and the to-be-annotated data and then inputting the spliced information and the to-be-annotated data into a plurality of pre-trained large language models; the large language model is used for generating a labeling result of the to-be-labeled data; and the quality control layer is used for monitoring the quality of the labeling results generated by the plurality of large language models and carrying out feedback analysis according to the monitored labeling quality problem. According to the method and the device, the automation of the whole process from the acquisition of the to-be-labeled data to the generation of the labeling result can be realized, the vertical domain knowledge is integrated, the labeling result is strictly controlled, and the process is continuously subjected to feedback optimization.

The application provides a model updating method and device for image labeling and electronic equipment. The method automatically labels original image data through a pre-labeling model to obtain a labeling data set. The pre-labeling model is a model to be updated. At least part of the labeling data in the labeling data set is subjected to a correction operation. Correction operation data corresponding to the at least part of the labeling data is determined. A correction difference pair is formed based on the at least part of the labeling data and the corresponding correction operation data. If the number of the correction difference pairs is greater than or equal to a preset number threshold, the pre-labeling model is fine-tuned according to the correction difference pairs to obtain an updated pre-labeling model. The original pre-labeling model is replaced by the updated pre-labeling model. The original pre-labeling model is replaced by the updated pre-labeling model to label image data, which can provide more accurate, more efficient and more stable image labeling, reduce the workload of manual auditing, and reduce the labor cost and time cost.

The application discloses a kind of graphic annotation automation processing method and system, specifically related to image clustering and automatic annotation field, including the entropy value and gradient variance of the pixel region of input image are calculated to generate complexity index, according to complexity index adjustment region division scale, obtain division region;Color histogram, texture direction and edge curvature are calculated in division region, form region feature, and complexity index and region feature are combined to generate feature matrix;The application generates feature matrix by calculating region complexity index and extracting multi-dimensional features, inputs clustering model to identify boundary cluster and jump region, combined with weight map and segmented weighting generates multi-class and single-class annotation, to solve the problem of semantic confusion, unstable boundary and error accumulation.

This invention belongs to the field of data processing and artificial intelligence technology, specifically a method and system for automatic classification of medical device patents based on deep learning. The method includes the following steps: acquiring textual data of medical device patents and performing preprocessing; introducing medical device classification standards as an external knowledge base, and calibrating the labels of the preprocessed text data to form a standard dataset; for categories with insufficient sample numbers in the standard dataset, generating synthetic data based on the patent text using a generative model; constructing a deep learning classification model that integrates global and local semantic features; training the deep learning classification model using the enhanced dataset, and inputting the medical device patent text to be classified into the trained model for classification prediction, outputting its corresponding medical device classification standard category. This invention solves the problems of mismatch between existing patent classification systems and industry regulatory standards, and low classification accuracy.

The application discloses a model training method, a category prediction method, a device, a computing device and a computer readable storage medium, to solve the problem of low category prediction accuracy of the model. The training method comprises: training a category prediction first model by using an original sample object set; determining the matching degree between the label of the sample object and the predicted category of the sample object by using the trained first model; selecting sample objects meeting a preset condition from the original sample object set; and training a category prediction second model by using the selected sample objects. Since the sample objects used for training the category prediction second model are sample objects with relatively high label annotation accuracy, compared with the category prediction model trained directly based on the dirty data set, the category prediction model trained by the method provided in the application embodiment has higher accuracy.

The application provides a kind of semi-automatic labeling method and system of ultrasonic video, comprising: step S1, target marking is carried out to target frame image in ultrasonic video data, and first marking information is obtained;Step S2, according to first marking information, target identification marking is carried out to front and rear frame image, and then second marking information corresponding to all frame images is obtained;Step S3, when the same target exists different marking, all markings corresponding to the repeatedly marked target are arbitrated;Step S4, according to the arbitration result, update second marking information.Have beneficial effect: the application carries out target tracking marking to front and rear frame image by the marking information of the target of target frame image, can mark target more accurately, and can automatically update the marking of front and rear frame;At the same time, arbitration and update are carried out to repeated marking, determine the final marking result, can avoid labeling error and reduce labeling redundancy, can further improve labeling accuracy.

The application provides an extended training data set, and a method for classification using a classification model trained by the extended training data set. The method for extending the training data set includes generating merged data based on original data and auxiliary data, the auxiliary data being associated with an application scenario of the classification model and coming from a source different from a source of the original data; and updating a category of the merged data using the classification model to generate a first extended data set for training the classification model. The scheme of the application can expand the number of labeled data samples and improve the labeling accuracy of the data, so that the classification performance of the trained classification model is optimized.

The invention relates to the technical field of artificial intelligence, in particular to an automatic entity labeling method perpendicular to industry policies and regulations. The method comprises the following steps: analyzing a document format, judging whether the document is a text document, dynamically splitting the document into text blocks based on semantic integrity, vectorizing the text, identifying entities and attributes, vectorizing the entities, clustering based on vector similarity and the like. According to the method, through full-process automation of semantic-driven text partitioning, LLM entity recognition and disambiguation, knowledge network construction and graph fusion, the tedious links of reading-understanding-labeling-rechecking in traditional manual labeling are replaced, the requirements for professional labeling personnel and quality inspection personnel are reduced through the automatic process, and in addition, the automatic labeling efficiency is improved. Through cooperation of a dynamic partitioning strategy (reducing LLM single processing capacity), vector retrieval (quickly positioning similar entities) and map fusion (incremental updating knowledge), the labeling period is greatly shortened.