Data classification method and work order classification method

Abstract

The embodiments of this specification provide a data classification method and a work order classification method, wherein the data classification method includes: obtaining a target data feature information set corresponding to a data set to be classified, wherein the target data feature information in the target data feature information set has the same dimension; generating data distance information corresponding to the data set to be classified based on the target data feature information set; determining at least one similar data set and similar data summaries corresponding to each similar data set according to the data distance information. By extracting target data feature information and generating data distance information, the classification problem between data sets is effectively solved. By unifying feature input, generating data distance information and extracting similar data summaries, the bottleneck of traditional models in processing massive data is broken, thereby improving the efficiency of data processing while ensuring accuracy when processing massive data.

Description

Technical Field

[0001] The embodiments of the present specification relate to the field of computer technology, and in particular to a data classification method and a work order classification method. Background Art

[0002] With the development of generative models, it has brought the ability to communicate fluently with humans, complete specific thinking tasks, assist writing, generate images or videos based on descriptions, etc., showing adaptability in solving basic problems. Although the results are not perfect, in the information age of data explosion, the expansion of massive data requires us to improve our ability to mine data value, and cluster summarization has become one of the important data mining methods.

[0003] At present, using a large model to summarize a single text and then synthesize all the summarized texts can solve the problem of data summarization and synthesis to a certain extent, but the current large model algorithm cannot achieve discriminative mining. In addition, the input text and window length of the large model are limited. Compared with data with a length of millions or even hundreds of thousands per month, the input length is still limited. Therefore, a data classification method is needed that can achieve discriminative mining and highly abstract generalization of data while processing massive data. Summary of the invention

[0004] In view of this, the embodiments of this specification provide a data classification method, a work order classification method, and a data classification method applied to the cloud, and a work order classification method applied to the cloud. One or more embodiments of this specification also involve a data classification device, a computing device, a computer-readable storage medium, and a computer program product to solve the technical defects existing in the prior art.

[0005] According to a first aspect of an embodiment of this specification, a data classification method is provided, including:

[0006] Acquire a target data feature information set corresponding to the data set to be classified, wherein the target data feature information in the target data feature information set has the same dimension;

[0007] Generating data distance information corresponding to the to-be-classified data set based on the target data feature information set, wherein the data distance information represents the degree of similarity between each to-be-classified data in the to-be-classified data set;

[0008] At least one data set of the same type and a data summary of the same type corresponding to each data set of the same type are determined in the data set to be classified according to the data distance information.

[0009] According to a second aspect of an embodiment of this specification, a work order classification method is provided, including:

[0010] Acquire a target work order feature information set corresponding to the work order data set, wherein the target work order feature information in the target work order feature information set has the same dimension;

[0011] Generating work order distance information corresponding to the work order data set based on the target work order feature information set, wherein the work order distance information represents the similarity between each work order data in the work order data set;

[0012] At least one similar work order set and similar work order summaries corresponding to each similar work order set are determined in the work order data set according to the work order distance information.

[0013] According to a third aspect of an embodiment of this specification, a data classification method is provided, which is applied to a cloud device, including:

[0014] A data classification request sent by a receiving end-side device, wherein the data classification request carries a set of data to be classified;

[0015] Acquire a target data feature information set corresponding to the data set to be classified, wherein the target data feature information in the target data feature information set has the same dimension;

[0016] Generating data distance information corresponding to the to-be-classified data set based on the target data feature information set, wherein the data distance information represents the degree of similarity between each to-be-classified data in the to-be-classified data set;

[0017] Determine at least one data set of the same type and a data summary of the same type corresponding to each data set of the same type in the data set to be classified according to the data distance information;

[0018] A data classification result is generated based on each similar data set and a similar data summary corresponding to each similar data set, and the data classification result is sent to the terminal side device.

[0019] According to a fourth aspect of an embodiment of this specification, a work order classification method is provided, which is applied to a cloud device, including:

[0020] A work order classification request sent by a receiving end-side device, wherein the data classification request includes a work order data set;

[0021] Acquire a target work order feature information set corresponding to the work order data set, wherein the target work order feature information in the target work order feature information set has the same dimension;

[0022] Generating work order distance information corresponding to the work order data set based on the target work order feature information set, wherein the work order distance information represents the similarity between each work order data in the work order data set;

[0023] Determine at least one similar work order set and similar work order summaries corresponding to each similar work order set in the work order data set according to the work order distance information;

[0024] A work order classification result is generated based on the similar work order summaries corresponding to each similar work order set, and the work order classification result is sent to the terminal side device.

[0025] According to a fifth aspect of an embodiment of this specification, a computing device is provided, including:

[0026] Memory and processor;

[0027] The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and steps of the work order classification method applied to cloud devices are implemented.

[0028] According to the sixth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions. When the instructions are executed by a processor, the steps of the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and work order classification method applied to cloud devices are implemented.

[0029] According to the seventh aspect of the embodiments of this specification, a computer program product is provided, including a computer program / instruction, which, when executed by a processor, implements the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and steps of the work order classification method applied to cloud devices.

[0030] The solution of the embodiment of this specification is applied to generate data distance information based on the target data feature information set and extract the summary of the same data set, which effectively breaks through the bottleneck of the traditional model in processing large-scale heterogeneous data. This process improves data processing efficiency through cluster analysis, ensures that high accuracy can be maintained when processing massive data, and speeds up the extraction and summarization of key information, providing a more efficient and accurate means for big data classification and summary generation. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 is a flow chart of a data classification method provided by an embodiment of this specification;

[0032] Figure 2is a schematic diagram of a data classification process provided by an embodiment of this specification;

[0033] Figure 3 is a flow chart of a work order classification method provided by an embodiment of this specification;

[0034] Figure 4 is a flow chart of a data classification method applied to a cloud device provided by an embodiment of this specification;

[0035] Figure 5 It is a flow chart of a work order classification method applied to a cloud device provided by an embodiment of this specification;

[0036] Figure 6 is an architectural diagram of a data classification system provided by an embodiment of this specification;

[0037] Figure 7 is a processing flow chart of an image classification method provided by an embodiment of this specification;

[0038] Figure 8 It is a structural schematic diagram of a data classification device provided by an embodiment of this specification;

[0039] Fig. 9 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0040] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0041] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0042] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0043] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0044] In one or more embodiments of this specification, a large model refers to a deep learning model with large-scale model parameters, which usually contains hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than 10 trillion model parameters. A large model can also be called a foundation model / foundation model. The large model is pre-trained with large-scale unlabeled corpus to produce a pre-trained model with more than 100 million parameters. This model can adapt to a wide range of downstream tasks, and the model has good generalization ability, such as a large-scale language model (LLM), a multi-modal pre-training model, etc.

[0045] When the big model is used in practice, only a small number of samples are needed to fine-tune the pre-trained model and it can be applied to different tasks. The big model can be widely used in natural language processing (NLP), computer vision and other fields. Specifically, it can be applied to computer vision tasks such as visual question answering (VQA), image caption (IC), image generation, as well as natural language processing tasks such as text-based sentiment classification, text summary generation, and machine translation. The main application scenarios of the big model include digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc.

[0046] First, the terms involved in one or more embodiments of this specification are explained.

[0047] Convolutional Neural Network (CNN): It is a special neural network structure composed of convolutional layers, pooling layers and fully connected layers. The convolutional layer is mainly used to extract local features of the data. It slides and convolves the data through a series of convolution kernels to capture the spatial features in the image or signal. The pooling layer downsamples the feature map output by the convolutional layer to reduce the data dimension and improve computational efficiency while maintaining the significance of the features. The fully connected layer maps the extracted features to the target category or regression value to achieve the classification or prediction task at the end.

[0048] Recurrent Neural Network (RNN): It is a neural network structure used to process sequence data. It can retain contextual information during processing and is suitable for time series, text, audio and other data. It remembers the previous input state through the cyclic connection of the hidden layer and passes the information step by step to the subsequent step.

[0049] Long Short-Term Memory (LSTM): solves the problem of gradient vanishing in long sequence data of traditional RNN (Recurrent Neural Network). It introduces memory units and gating mechanisms, which can dynamically update or maintain information according to the importance of data, thereby capturing long-term dependencies. Through input gates, forget gates, and output gates, key information can be saved more effectively, and it is widely used in tasks such as natural language processing and time series prediction.

[0050] In this specification, a data classification method, a work order classification method, and a data classification method applied to the cloud, and a work order classification method applied to the cloud are provided. One or more embodiments of this specification also involve a data classification device, a computing device, a computer-readable storage medium, and a computer program product, which are described in detail one by one in the following embodiments.

[0051] With the rapid development of generative models, more and more attention is focused on generative large models. Such models can not only achieve fluent conversations with humans and perform artificially defined "thinking" tasks, but also complete text or help humans write, and even generate images and videos based on language descriptions, thereby completing many tasks that require humans to concentrate for a long time or a lot of energy to achieve. Although the current effect still needs to be improved, the generative large model has shown good adaptability in the basic problem from 0 to 1.

[0052] With the explosive growth of data brought about by informatization, various services have also gradually become informatized, further promoting the massive expansion of data. However, if data cannot create value, it will cause a waste of resources, so the mining of data value becomes particularly important. The clustering summary method we proposed is an important means to achieve data mining. It is difficult to cluster using a generative large model. Although the large model is good at summarizing and synthesizing, it is difficult to achieve differentiated mining due to the characteristic of the algorithm selecting the "same distribution".

[0053] See also Figure 1 , Figure 1 A flow chart of a data classification method provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0054] Step 102: Obtain a target data feature information set corresponding to the data set to be classified, wherein the target data feature information in the target data feature information set has the same dimension.

[0055] In practical applications, the data set to be classified is a set of data items that need to be classified, such as images, texts, or work order records from different sources; the target data feature information set is feature information extracted from the data set to be classified, which is used to express the characteristics of each data item.

[0056] The data set to be classified can be understood as a data set that needs to be classified or grouped in the system. These data sets are usually diverse and heterogeneous. Taking text classification as an example, the data set to be classified may contain documents or records from different sources, such as work order records, social media posts, or customer feedback. These data sets may have significant differences in content, length, format, etc. The system needs to uniformly analyze and classify these data from different sources in order to extract the data features of each category and generate relevant summary or clustering information.

[0057] The target data feature information set can be understood as an information set extracted from the data set to be classified through feature extraction technology to represent the specific features of each data item. For image classification tasks, the target data feature information set may include the texture, color distribution or edge features of the image; for text data, it may be extracted keywords, syntactic structure or emotional tendency. By establishing the target data feature information set, the system can more clearly capture the characteristics of each data item, thereby generating data distance information in subsequent calculations, further promoting the clustering and accurate classification of similar data.

[0058] It should be noted that obtaining the target data feature information set corresponding to the data set to be classified can be understood as extracting the feature information of each data in the data set in a variety of ways; the specific way is to determine the target data as the initial feature information from the data set to be classified, and process the initial feature information through a text prediction model to obtain the target data feature information containing extended semantics; it is also possible to use a deep learning model to directly extract features from the data, such as extracting shape, texture and other feature information from image data through a convolutional neural network; it is also possible to use embedding technology to convert text or speech data into a low-dimensional vector representation to capture the key semantic or emotional features of the data. This specification does not impose any restrictions on this.

[0059] Extracting the target data feature information set corresponding to the data set to be classified can reduce the data dimension and improve processing efficiency while maintaining the essential information of the data.

[0060] Furthermore, obtaining a target data feature information set corresponding to the data set to be classified includes:

[0061] Determine target data to be classified, wherein the target data to be classified is any one of the data to be classified in the data set to be classified;

[0062] Acquire initial feature information corresponding to the target data to be classified, wherein the initial feature information represents the features of the target data to be classified itself in the text dimension;

[0063] Processing the initial feature information based on a text prediction model to obtain target data feature information, wherein the target data feature information represents extended features of the target data to be classified in a text dimension;

[0064] A target data feature information set is obtained according to the target data feature information corresponding to each to-be-classified data in the to-be-classified data set.

[0065] In practical applications, the initial feature information is the basic features of the data to be classified in its original state. The text prediction model is a model used to process and predict text content, which can analyze the input text and generate corresponding extended features.

[0066] Initial feature information can be understood as the basic features of the data to be classified without processing. These features directly reflect the information of the data in the original dimension. It usually includes word frequency, text length, basic syntactic structure, etc., which are used to describe the original state of the text data. For example, when classifying customer work orders, the initial feature information includes the keywords, sentence length, and grammatical structure of each work order. These features will be used as basic data for further feature extraction and expansion in subsequent processing.

[0067] A text prediction model can be understood as a large model that can analyze and infer input text data. Its main function is to generate new text features or predict text content by learning patterns and grammatical rules in large amounts of text data. A text prediction model can not only expand the semantic information of a text, but also add more potential meanings to the text through understanding the context. For example, given a piece of text input, a text prediction model can infer the sentiment of the text, the semantics of keywords, and the next possible word or phrase based on the context.

[0068] The text prediction model can generate more complex features based on the initial feature information, thereby optimizing the accuracy of subsequent data classification.

[0069] Further, the initial feature information is processed based on the text prediction model to obtain target data feature information, including:

[0070] Inputting the initial feature information into the text prediction model, and obtaining at least one reference feature information output by the text prediction model;

[0071] The target data feature information is generated based on each reference feature information.

[0072] In practical applications, reference feature information is the intermediate feature output from the text prediction model for further generating target data feature information, including the expansion and enrichment of the initial feature information. Reference feature information can be understood as a set of features obtained after processing the initial feature information through the text prediction model, which is used to further deepen the understanding of the classified data. The reference feature information contains a higher-level semantic representation of the initial features, which can provide more details and potential relationships for classification and clustering tasks. For example, when analyzing customer feedback, the initial feature information may include word frequency and text length, while the reference feature information will extract the sentiment tendency, topic keywords, etc. of the text.

[0073] The target data feature information is generated based on the multiple reference feature information output by the text prediction model, aiming to transform the initial feature information into a feature representation with more semantic depth and multi-dimensionality, so that the system can achieve more accurate results in classification and clustering. The reference feature information output by the text prediction model includes the expansion and correlation analysis of the initial data features, such as subject words, sentiment tendencies, context associations, etc. These feature information can effectively capture the potential patterns and relationships in the original data. However, the reference feature information may still be relatively scattered and cannot directly reflect the overall characteristics. Therefore, by integrating multiple reference feature information into target data feature information, the system can obtain more generalized and representative features. This process not only improves the expressive power of data features, but also enhances the robustness and adaptability in subsequent classification tasks, enabling the system to more comprehensively identify and summarize similar items in the data set to ensure the accuracy and effectiveness of data classification.

[0074] Step 104: Generate data distance information corresponding to the to-be-classified data set based on the target data feature information set, wherein the data distance information represents the degree of similarity between each to-be-classified data in the to-be-classified data set.

[0075] In practical applications, data distance information is a set of numerical values ​​that measures the similarity between data items in the data set to be classified, and is used to represent the relative relationship and difference between different data items.

[0076] Exemplarily, data distance information can be understood as a set of indicators used to quantify the similarity between each data item in the data set to be classified. Usually, data distance information is obtained by calculating the features in the target data feature information set. Data distance information can reflect the relationship between data items and help the classification model identify which data items are more similar, so as to classify similar items into the same category. For example, in image classification tasks, data distance information may be based on features such as pixel values ​​and color distribution, and calculate the distance between images through methods such as Euclidean distance and cosine similarity; in text classification, distance information can be used to evaluate the semantic similarity between texts by using cosine distance on text embedding vectors.

[0077] By constructing a data distance matrix to represent the distance relationship between each data item, data distance information can improve classification efficiency when processing massive data. For each data item, the distance matrix contains its distance from other items in the collection, so that the system can identify similar data by looking up adjacent items in the distance matrix. In clustering tasks, data distance information can be used as a basis for classification, and data items can be aggregated into specific clusters according to the size of the distance, so as to achieve effective classification and summary generation of data.

[0078] Further, generating data distance information corresponding to the to-be-classified data set based on the target data feature information set includes:

[0079] Based on each target data feature information in the target data feature information set, generating a target data feature matrix, wherein the target data feature matrix is ​​used to present the corresponding feature representation of each to-be-classified data in the to-be-classified data combination;

[0080] A data distance matrix is ​​calculated according to the target data feature matrix, and data distance information is obtained based on the data distance matrix, wherein the data distance matrix is ​​used to represent the distance relationship between each data to be classified.

[0081] In practical applications, the target data feature matrix is ​​a matrix formed by arranging the target data features of all data to be classified according to a specific dimension, which is used to express the feature representation of each data item; the data distance matrix is ​​a matrix calculated based on the target data feature matrix, which is used to represent the distance or similarity between data items.

[0082] The target data feature matrix can be understood as a matrix that integrates the target features of each data item in the data set to be classified, which can centrally present the feature representation of all data items. This matrix is ​​usually high-dimensional, and each row or column represents the feature information of a data item. By constructing the target data feature matrix, the system can quickly calculate the distance relationship between data items in subsequent steps to facilitate clustering or classification. For example, in a text classification task, the features of each row may contain feature information such as the subject word distribution and sentiment tendency of the text, while in an image classification task, the features may contain information such as color and edge detection.

[0083] The data distance matrix can be understood as a matrix generated by calculating the similarity or difference between each data item in the target data feature matrix, which is used to represent the distance relationship between each pair of data items. The data distance matrix can help the system identify similar data items and classify them into the same category. There are many ways to calculate distance, such as Euclidean distance, cosine similarity, etc. The structure of this matrix is ​​usually symmetrical, where each value reflects the similarity between the corresponding data items, and the smaller the value, the more similar the data items are. For example, in image classification, the values ​​in the distance matrix may be based on the difference between pixels; in text analysis, the distance matrix represents the similarity of different texts through the cosine distance of word vectors.

[0084] It should be noted that determining data distance information through a data distance matrix can be understood as extracting and quantifying the similarities or differences between data items from the matrix. Specifically, each value in the distance matrix can be used as data distance information, and cluster analysis can be performed based on the distance value between each pair of data items; data items can also be grouped according to values ​​less than a set threshold in the distance matrix, thereby screening out closer data pairs; the distance matrix can also be standardized to convert different distance values ​​into relative distance levels, so as to more accurately classify and identify similarities in complex data sets. This specification does not impose any restrictions on this.

[0085] Determining data distance information through the data distance matrix can effectively reveal the similarities and differences between data items, allowing the system to more accurately identify similar data sets during large-scale data classification and clustering. With this information, the system can significantly improve classification efficiency and accuracy, avoid misclassification problems caused by uneven data distribution, and lay a solid foundation for subsequent feature extraction and data summary generation. The determination of data distance information also supports flexible distance threshold settings, allowing the system to adjust similarity standards according to different application requirements, thereby achieving more adaptive classification and analysis in various complex data scenarios.

[0086] Step 106: Determine at least one data set of the same type and a data summary of the same type corresponding to each data set of the same type in the data set to be classified according to the data distance information.

[0087] In practical applications, a homogeneous data set is a set of data items that have similar characteristics and are classified into the same category, and a homogeneous data summary is a concise description that summarizes the core information of the homogeneous data set.

[0088] Exemplarily, a homogeneous data set can be understood as a set of data items with high similarity in feature dimensions, which show consistency in certain key features. Based on the data distance information, the system can identify similar data items and classify them into the same set, thus forming a homogeneous data set. For example, in text analysis, multiple documents with similar topics or keywords may be classified into the same category for further processing; in image classification, a homogeneous data set may include images with similar shapes, colors or contents, thus providing a basis for classification or clustering analysis.

[0089] A homogeneous data summary can be understood as a text description that extracts and summarizes the main features or themes of a homogeneous data set. It can remove redundant data while retaining the core information of the set, so that the overall characteristics of the set can be presented concisely and clearly. The process of generating a homogeneous data summary is usually based on the integration of representative data items or feature information in the set. For example, for a set of weather report documents, a homogeneous data summary can summarize the main weather conditions, temperature trends and other key information of this type of document; for a set of similar customer feedback, a homogeneous data summary may contain the main feedback topics, such as product quality or customer service.

[0090] It should be noted that determining at least one similar data set and similar data summaries corresponding to each similar data set based on data distance information can be understood as a process of identifying and aggregating similar data items and generating summaries through data distance information. The specific method is to obtain the sequence information of the data to be classified through the data distance information, and organize the target data feature information set based on the sequence information of the data to be classified, thereby forming data feature sequence information, so as to determine the similar data set and generate its summary; a clustering algorithm can also be used to group data with similar distances into similar sets, and generate corresponding summaries based on the core features of each group; hierarchical clustering can also be used to gradually build a hierarchical relationship tree based on the distance between data, so as to identify sets of similar data items, and then perform feature extraction on each set to generate a summary. This specification does not impose any restrictions on this.

[0091] By using distance information to determine each similar data set and the corresponding similar data summary of each similar data set, accurate data clustering and feature extraction can be effectively achieved, ensuring high classification accuracy in large-scale data processing, while greatly improving data summarization and processing efficiency. In this way, the system can automatically identify and summarize the core features of data items and generate concise summary information, which helps to quickly understand and use data and provide a reliable basis for subsequent analysis and decision-making.

[0092] Further, determining at least one data set of the same type and a data summary of the same type corresponding to each data set of the same type in the data set to be classified according to the data distance information includes:

[0093] Acquire sequence information of the data to be classified according to the data distance information;

[0094] Sorting and processing the target data feature information set based on the sequence information of the data to be classified to obtain data feature sequence information;

[0095] At least one similar data set and similar data summaries corresponding to each similar data set are determined based on the data feature sequence information.

[0096] In practical applications, the order information of the data to be classified is the order mark after the data to be classified is arranged according to the data distance information, and the data feature sequence information is the feature matrix or sequence formed by rearranging the target data feature information according to the order information of the data to be classified.

[0097] The order information of the data to be classified can be understood as an order mark or arrangement scheme obtained by analyzing the distance information between data items, which is used to guide the sorting and subsequent processing of data items. This order information can effectively reflect the similarity between data items, so that similar data can be arranged in similar positions to facilitate further analysis and feature extraction. For example, in the text classification of customer feedback, the order information of the data to be classified may be arranged according to the similarity of the feedback content, so that the relevant feedback is more concentrated, providing a clearer arrangement basis for subsequent clustering and analysis.

[0098] Data feature sequence information can be understood as a feature sequence or matrix generated by rearranging the target data feature information according to the order information of the data to be classified. This information organizes the feature information of the data items into an ordered structure according to the similarity or distance relationship, which is convenient for the model to extract features and generate summaries. For example, in image clustering, data feature sequence information may arrange images with similar color or shape features together, thereby forming an ordered feature matrix, which is convenient for the model to efficiently identify and process similar features.

[0099] By using data distance information to arrange the target data feature information to obtain data feature sequence information, at least one similar data set and the similar data summary corresponding to each similar data set can be determined, which can effectively optimize the data classification process and improve the system's recognition ability and clustering accuracy for similar data items. This method ensures that data items with similar features are centrally arranged in the feature matrix, making it easier for the model to quickly detect and classify similar data, thereby speeding up data processing. By generating concise similar data summaries, the system can effectively reduce data redundancy while maintaining overall information, provide users with clear and concise results, and improve the efficiency of subsequent decision-making and analysis.

[0100] Further, determining at least one similar data set and similar data summaries corresponding to each similar data set in the data set to be classified based on the data feature sequence information includes:

[0101] Determine at least one similar data feature set based on the data feature sequence information, wherein the similar data feature set includes at least one target data feature information;

[0102] Generate similar data summaries corresponding to similar data feature sets according to the characteristic information of each target data;

[0103] Based on the data feature sequence information, the first similar data set corresponding to the first similar data feature set is determined, and the similar data summary corresponding to the first similar data feature set is determined to be the similar data summary corresponding to the first similar data set, wherein the first similar data feature set is any one of the similar data feature sets, and the first similar data set is any one of the similar data sets.

[0104] In practical applications, a similar data feature set is a set of data items with similar features, which is used to summarize target data feature information with common characteristics. A similar data feature set can be understood as a set formed by combining data with similar features through the feature similarity of each data item in the data feature sequence. This set can effectively reflect the association and commonality between data items, thereby simplifying subsequent analysis and processing steps. For example, in customer feedback analysis, a similar data feature set may include all feedback data involving keywords related to "product quality", allowing the system to quickly identify and process issues related to product quality. In image classification, a similar data feature set can group images with similar textures or color distributions into the same set, facilitating subsequent clustering or feature extraction operations.

[0105] Exemplarily, determining the first similar data set corresponding to the first similar data feature set based on the data feature sequence information, and determining the similar data summary corresponding to the first similar data feature set as the similar data summary corresponding to the first similar data set, can be understood as determining the similar data summary corresponding to the similar data set as the similar data summary corresponding to the corresponding similar data feature set. For example, similar data set A corresponds to similar data feature set A', and similar data set B corresponds to similar data feature set B'; and similar data feature set A' corresponds to similar data summary a, and similar data feature set B' corresponds to similar data summary b, then it is determined that the similar data summary corresponding to similar data set A is a, and the similar data summary corresponding to similar data set B is b.

[0106] By building a collection of similar data features, the system can effectively organize data items with similar features in large-scale data into more concentrated units. This process not only improves the classification and recognition efficiency of the system, but also provides a basis for generating concise and representative summaries of similar data, thereby better supporting the actual needs of automated data analysis and application scenarios.

[0107] Furthermore, a summary of similar data corresponding to each similar data feature set is generated according to each target data feature information, including:

[0108] Obtain data category feature information corresponding to each similar data feature set according to each target data feature information;

[0109] Based on the data category feature information corresponding to the second similar data feature set, a similar data summary corresponding to the second similar data feature set is generated, wherein the second similar data feature set is any one of the similar data feature sets.

[0110] In practical applications, data category feature information is feature information that extracts and summarizes the main characteristics of a set of similar data features and is used to represent common attributes in the set.

[0111] Data category feature information can be understood as a representative information generated by analyzing the commonalities of each feature in a set of similar data features, which can effectively describe the overall characteristics of the set. These feature information usually contain key category features, which are used to provide identification basis for specific categories in subsequent data classification or analysis. For example, in the classification of customer feedback, data category feature information may include topic keywords such as "product quality" and "customer service" to reflect the main focus of this type of feedback; in image analysis, data category feature information may include characteristics such as color distribution and edge features to describe the visual characteristics of this type of image.

[0112] The method of generating a summary of similar data corresponding to a certain similar data feature set based on the data category feature information corresponding to the similar data feature set is to input the data category feature information corresponding to the similar data feature set into the summary generation model, and obtain the summary of similar data generated by the summary generation model. The summary generation model is a model that generates a brief summary text by analyzing the feature information of the input data, and is used to refine the core content of the data set into a concise summary. In addition, the summary generation model can be understood as a large model that can automatically generate text summaries, which is widely used in document summary extraction, news summaries, report generation and other scenarios. The summary generation model can receive feature information in vector form to generate a text summary corresponding to the feature information.

[0113] It should be noted that obtaining the data category feature information corresponding to each similar data feature set based on each target data feature information can be understood as generating high-level information reflecting the overall characteristics of similar data by integrating and processing the target data feature information. Specifically, each target data feature information can be spliced ​​into a matrix according to similarity, and the matrix can be processed so that each row or column of the matrix carries the characteristics of the current target data feature information in the overall data, thereby generating the corresponding data category feature information; each feature information in the similar data feature set can also be statistically analyzed, such as calculating the mean or weighted average to generate category features representing the characteristics of the set; and the category feature information can also be refined by aggregating representative features (such as cluster centers) in each similar data feature set to ensure that the core characteristics of the set are efficiently reflected. This specification does not impose any restrictions on this.

[0114] The generation of data category feature information enables the system to efficiently summarize the core features of similar data, avoiding processing each data item one by one. This summary information not only helps improve the efficiency of classification and clustering tasks, but also provides strong support for the system in generating summaries and identifying patterns, allowing users to quickly understand the main characteristics of each data category.

[0115] Furthermore, data category feature information corresponding to each similar data feature set is obtained according to each target data feature information, including:

[0116] Generate data feature sequence information according to each target data feature information;

[0117] Inputting the data feature sequence information into a preset global feature extraction network to obtain a target data global feature matrix generated by the global feature extraction network;

[0118] Based on the target data global feature matrix, data category feature information corresponding to each similar data feature set is obtained.

[0119] In practical applications, the global feature matrix of target data is a matrix generated after processing all target data feature information through a global feature extraction network, which is used to centrally express the global characteristics of the entire data set; the global feature extraction network is a deep learning network used to extract global information from data features.

[0120] The target data global feature matrix can be understood as a matrix that can comprehensively characterize the feature distribution of all data items in the overall data space after the feature information of each target data is processed by the global feature extraction network. Each row or column of the matrix not only reflects the characteristics of the current data item, but also contains its relative position and pattern in the data set, so that the system can accurately grasp the global properties of each data item. For example, in image analysis tasks, each row of the target data global feature matrix may contain the main texture, color and other features of the image, and reflect the characteristic distribution of the image in the overall image set; in text analysis, each row or column may contain the theme or emotional features of each text fragment, and show the relationship and similarity of the text in the overall semantic space.

[0121] The global feature extraction network can be understood as a network that extracts high-level and global features from input data through a deep learning architecture. Global feature extraction networks usually include models such as CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), and LSTM (Long Short-Term Memory), which can capture the overall pattern and complex features of the data. For example, in text analysis tasks, the global feature extraction network can extract the topic distribution or emotional tendency of the text, while in video processing tasks, the network can capture the continuity and change characteristics in the frame sequence. Through the global feature extraction network, the system can obtain the overall characteristics of the data and provide support for the generation of the global feature matrix of the target data, thereby achieving higher accuracy and adaptability in classification and clustering.

[0122] refer to Figure 2 , Figure 2 A schematic diagram of a data classification process provided for an embodiment of the present specification, wherein each data set to be classified in the data set to be classified first passes through an initial data feature extraction module to extract respective initial feature information. Each initial feature information after extraction is input into a text prediction model to generate target data feature information including more knowledge. Subsequently, each target data feature information is spliced, and the spliced ​​data is calculated by a data distance matrix module to obtain distance information between each data item, and data feature sequence information is generated based on the distance information between each data item obtained above. The data feature sequence information is processed by a global feature extraction network to generate a global feature matrix corresponding to the data set to be classified. Subsequently, the target data global feature information corresponding to each data to be classified is extracted from the global feature matrix. On this basis, the system integrates similar target data global feature information into similar data feature sets, so that each set contains multiple target data global feature information, and generates category feature information corresponding to the set based on the target data global feature information in each set. Finally, the category feature information corresponding to each similar data feature set is respectively input into the summary generation model to generate the similar data summary of each similar data feature set, and the data to be classified corresponding to the global feature information of the target data in each similar data feature set is divided into the same similar data set, and the similar data summary corresponding to each similar data feature set is determined to be the similar data summary corresponding to each corresponding similar data set.

[0123] The construction of the global feature matrix enables the system to accurately locate the relative characteristics of each data item during the classification or clustering process, thereby improving the overall effect of data processing.

[0124] Furthermore, obtaining data category feature information corresponding to each similar data feature set based on the target data global feature matrix includes:

[0125] Determine at least one target data global feature information corresponding to each similar data feature set in the target data global feature matrix according to the data feature sequence information;

[0126] Based on the global feature information of each target data corresponding to the third similar data feature set, the data category feature information corresponding to the third similar data feature set is generated, wherein the third similar data feature set is any one of the similar data feature sets.

[0127] In practical applications, generating data category feature information corresponding to a set of similar data features based on the global feature information of each target data corresponding to the set can be understood as generating the main features representing the set by comprehensively processing the global features of the similar data. Specifically, the mean or weighted average of the global features of each target data can be calculated to obtain the feature representative of the set as a whole; a clustering algorithm can also be used to extract representative feature points or feature vectors in the set as data category feature information; and a statistical analysis can also be performed on the distribution of feature information in the set (such as calculating the median or mode) to generate a representative feature description to more accurately reflect the commonality of the set. This specification does not impose any restrictions on this.

[0128] Exemplarily, determining at least one target data global feature information corresponding to each similar data feature set in the target data global feature matrix according to the data feature sequence information can be understood as using the data feature sequence information to determine the sequence number of the feature corresponding to each similar data feature set, and then using the sequence number corresponding to each similar data feature set to determine the global feature information corresponding to each sequence number from the target data global feature matrix as the global feature information corresponding to each similar data feature set. For example, according to the data feature sequence information, it is determined that the sequence numbers of the features corresponding to the similar data set A are 2, 4, and 5, and the sequence numbers of the similar data features corresponding to the similar data set B are 0, 1, 3, and 6, and then it is confirmed that the three vectors corresponding to the 2nd, 4th, and 5th rows in the target data global feature matrix are the target data global feature information corresponding to the set A, and it is confirmed that the four vectors corresponding to the 0th, 1st, 3rd, and 6th rows in the target data global feature matrix are the target data global feature information corresponding to the set B.

[0129] By integrating the global feature information of all target data in each similar data feature set, more representative category feature information can be extracted. In this way, the system can more accurately reflect the overall attributes and uniqueness of this type of data, which helps to clearly divide the classification and clustering results. The global feature information of each target data contains the high-level features of each data item. By integrating these features to generate data category feature information, not only can redundant data be removed, but also the commonalities of similar data can be highlighted, which facilitates subsequent summary generation and rapid understanding, thereby ensuring that the system can efficiently and accurately process large-scale data, thereby improving the applicability and accuracy of the analysis results.

[0130] By applying the solution of the embodiments of this specification, the method breaks through the bottleneck of traditional technology in processing complex and heterogeneous data by introducing the generation and processing of target data feature information sets in the data classification process. Specifically, by processing the initial feature information based on the text prediction model, obtaining the target data feature information, and then generating data distance information and determining the same type of data set, accurate classification and feature extraction can be achieved in large-scale data processing. This technology effectively captures the correlation between different data by constructing a feature matrix and calculating the data distance matrix, ensuring the efficiency of data processing.

[0131] In addition, by introducing the combination of data feature sequence information and the global feature extraction network, in the process of determining similar data sets and generating summaries, not only the accuracy of data clustering is improved, but also the summarization and extraction of key information is accelerated. In this way, it is possible to better handle massive and heterogeneous data and improve the intelligence level of data classification. In general, this technical solution significantly enhances the efficiency and accuracy of big data classification and summary generation, is suitable for a variety of complex data processing tasks, and provides strong support for data mining.

[0132] Corresponding to the above method embodiment, this specification also provides a work order classification method embodiment, see Figure 3 , Figure 3 A flowchart of a work order classification method provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0133] Step 302: Acquire a target work order feature information set corresponding to the work order data set, wherein the target work order feature information in the target work order feature information set has the same dimension.

[0134] Step 304: Generate work order distance information corresponding to the work order data set based on the target work order feature information set, wherein the work order distance information represents the degree of similarity between each work order data in the work order data set.

[0135] Step 306: Determine at least one similar work order set and similar work order summaries corresponding to each similar work order set in the work order data set according to the work order distance information.

[0136] In practical applications, the work order data set is a group of work order data that needs to be classified and analyzed, the target work order feature information set is the key feature information extracted from the work order data set, the work order distance information is the data to measure the similarity between different work orders, forming the basis for classifying work orders, the same work order set is a group of work orders with similar features classified according to the distance information, and the same work order summary is a concise description of the core information of the same work order set.

[0137] The work order data set can be understood as a set of work order data to be processed, usually from customer requests, technical support or internal task reports. Each work order contains multiple fields such as problem description, request type, customer information, etc. When classifying and processing, the system analyzes these work orders as a whole data set to identify the commonalities or differences between the data. For example, a work order data set may contain maintenance requests from different departments, involving different fault types or priority levels. The target work order feature information set can be understood as the key feature information extracted from the work order data set for analysis and classification. This feature information set can effectively represent the content or nature of each work order, usually including information such as extracted keywords, request type or problem complexity. By extracting features from the work order content, the system can better identify and understand the core content of each work order. For example, in an IT support system, the target work order feature information set may include information such as error code, device type or request priority.

[0138] Work order distance information can be understood as a set of numerical values ​​used to quantify the similarity between work orders. Work order distance information is calculated based on the target work order feature information set and reflects the degree of association between different work order items. Through distance information, the system can identify work orders that are similar or related to each other. For example, in a customer service system, work orders that describe similar faults will show a shorter distance in the distance information and thus be considered similar work orders.

[0139] A similar work order set can be understood as a set of work orders that are similar in feature dimensions and are classified into the same category. The system aggregates similar work order items based on the work order distance information to form a work order group with similar themes or features. For example, a similar work order set may include all work orders related to a certain type of equipment failure, which makes it easier for the technical team to quickly locate the problem and provide batch solutions. A similar work order summary can be understood as a concise summary of the key information in a similar work order set. The summary can effectively summarize the core content of the work order set, so that the essential characteristics of the data are clearly presented. For example, for a network failure work order set, a similar work order summary may include the main manifestations of the failure, the scope of impact, and common processing steps, to facilitate subsequent analysis and rapid response.

[0140] The above is a schematic scheme of a work order classification method of this embodiment. It should be noted that the technical scheme of the work order classification method and the technical scheme of the above data classification method belong to the same concept, and the details not described in detail in the technical scheme of the work order classification method can be referred to the description of the technical scheme of the above data classification method.

[0141] By applying the solution of the embodiments of this specification, accurate classification and information extraction can be achieved in large-scale work order data by acquiring and generating work order distance information based on the target work order feature information set. This method effectively improves the recognition of relationships between work orders and the efficiency of data processing by constructing a feature matrix and calculating the distance matrix of work order data. In addition, the division and summary generation of similar work order sets based on work order distance information not only enhances the accuracy of work order clustering, but also accelerates the abstraction and refinement of information. In this way, different categories of work orders can be quickly identified during the work order data processing process, further improving the processing efficiency and accuracy. In general, this method provides a more accurate and efficient solution for work order classification, can cope with complex work order data mining tasks, improve the level of intelligence in the overall classification process, and optimize the overall efficiency of work order processing.

[0142] Corresponding to the above method embodiment, this specification also provides a data classification method embodiment applied to a cloud device, see Figure 4 , Figure 4 A flow chart of a data classification method applied to a cloud device provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0143] Step 402: A data classification request sent by a receiving end-side device, wherein the data classification request carries a set of data to be classified.

[0144] Step 404: Obtain a target data feature information set corresponding to the data set to be classified, wherein the target data feature information in the target data feature information set has the same dimension.

[0145] Step 406: Generate data distance information corresponding to the to-be-classified data set based on the target data feature information set, wherein the data distance information represents the degree of similarity between each to-be-classified data in the to-be-classified data set.

[0146] Step 408: Determine at least one data set of the same type and a data summary of the same type corresponding to each data set of the same type in the data set to be classified according to the data distance information.

[0147] Step 410: Generate a data classification result based on each similar data set and a similar data summary corresponding to each similar data set, and send the data classification result to the terminal side device.

[0148] In actual applications, cloud devices are remote servers or systems responsible for data storage and computing processing, and end-side devices are devices that directly interact with users, such as mobile phones or sensors. Data classification requests are information packets containing data classification requirements sent by end-side devices to cloud devices. Data classification results are classification outputs returned to the end-side devices by the cloud devices after completing the classification operation.

[0149] Cloud devices can be understood as servers or data centers located remotely, mainly used to perform data processing tasks with high computing power. Cloud devices usually have powerful storage and computing capabilities, can handle requests from multiple end-side devices, and have advantages in large-scale data processing, model training or analytical computing. For example, when end-side devices (such as smartphones or sensors) collect a large amount of data and it is difficult to complete the analysis locally, the data can be transmitted to cloud devices for centralized processing and the results are returned.

[0150] The end-side device can be understood as a device that directly interacts with the user and is responsible for data collection or request sending, usually including smartphones, smart sensors and other smart terminal devices. Such devices upload data or requests and receive processing results through communication with cloud devices. The end-side device assumes the role of data collection and transmission in the data classification process. Typical scenarios include mobile applications collecting data locally and sending it to the cloud for analysis, or IoT devices uploading collected environmental data to the cloud to obtain classification results.

[0151] A data classification request can be understood as an instruction information packet sent by the end device to the cloud device, which contains a set of data to be classified and its classification requirements. A data classification request usually includes data items and feature information so that the cloud device can identify the classification task and perform analysis. For example, the sensor device in the smart home system will send a classification request to the cloud, which includes environmental data or user behavior data, so that the cloud can classify specific scenario modes or device control instructions.

[0152] Data classification results can be understood as the output content returned to the end-side device after the cloud device completes the analysis of the data in the classification request, which usually contains classification labels, category summaries, or other structured information. Data classification results enable the end-side device to take corresponding actions based on the classification situation. For example, the intelligent voice assistant generates a specific voice response after receiving the cloud classification results, or the autonomous driving vehicle adjusts its driving strategy based on the classification results returned by the cloud.

[0153] The above is a schematic scheme of a data classification method applied to a cloud device in this embodiment. It should be noted that the technical scheme of the data classification method applied to the cloud device belongs to the same concept as the technical scheme of the above data classification method. For details not described in detail in the technical scheme of the data classification method applied to the cloud device, please refer to the description of the technical scheme of the above data classification method.

[0154] By applying the scheme of the embodiments of this specification, the generation and processing of the target data feature information set is introduced in the data classification process of the cloud device. This method breaks through the bottleneck of the traditional cloud data classification scheme when facing large-scale and diversified data. Specifically, by calculating the data distance information and dividing the same type of data sets based on the generation of the target data feature information set, the classification accuracy and information extraction efficiency can be significantly improved. This process effectively captures the similarities and correlations between different data by constructing a feature matrix and a data distance matrix, ensures the efficient generation of classification results, and improves the overall data processing efficiency while ensuring the classification accuracy.

[0155] In addition, the combination of target data feature information and summary generation of similar data sets further improves the clustering accuracy and information extraction speed in the data classification process. By combining data distance information with global feature extraction strategies, the system can quickly identify and classify different categories in massive data, which not only accelerates the generation of data classification results, but also significantly improves the intelligence level of the system in large-scale data processing. In short, this technical solution provides an efficient and accurate solution for data classification in cloud devices, which can adapt to complex and changeable data types, effectively improve the ability and intelligence level of data processing, and has strong adaptability and advantages in big data application scenarios.

[0156] Corresponding to the above method embodiment, this specification also provides a work order classification method embodiment applied to a cloud device, see Figure 5 , Figure 5 A flowchart of a work order classification method applied to a cloud device provided according to an embodiment of the present specification is shown, which specifically includes the following steps.

[0157] Step 502: receiving a work order classification request sent by a terminal-side device, wherein the data classification request includes a work order data set.

[0158] Step 504: Acquire a target work order feature information set corresponding to the work order data set, wherein the target work order feature information in the target work order feature information set has the same dimension.

[0159] Step 506: Generate work order distance information corresponding to the work order data set based on the target work order feature information set, wherein the work order distance information represents the degree of similarity between each work order data in the work order data set.

[0160] Step 508: Determine at least one similar work order set and similar work order summaries corresponding to each similar work order set in the work order data set according to the work order distance information.

[0161] Step 510: Generate a work order classification result based on the similar work order summaries corresponding to each similar work order set, and send the work order classification result to the terminal side device.

[0162] In actual applications, a work order classification request is a request sent by a terminal device to a server containing work order data to be classified, and a work order classification result is classification information returned by the server after classifying and processing the work order data.

[0163] A work order classification request can be understood as an instruction issued by the end-side device to request classified work order data, which includes a set of work order data to be classified. This request is generally issued when work orders need to be automatically classified or analyzed so that the server can process them efficiently. For example, in an enterprise management system, the customer service center can send customer service requests to the server through work order classification requests, so that the system can automatically classify different service requests according to the content, thereby improving response speed and processing efficiency.

[0164] The work order classification result can be understood as the feedback information generated by the server after analyzing and classifying the received work order data set. The result contains the classification information or cluster analysis results of each work order, as well as the corresponding brief description or summary, so that users can quickly understand the type and urgency of the work order. For example, the work order classification result can classify a batch of work orders about product failures into the "technical support" category, or classify a batch of common problems into the "general inquiry" category, helping managers to quickly judge and allocate resources.

[0165] The above is a schematic scheme of a work order classification method applied to a cloud device in this embodiment. It should be noted that the technical scheme of the work order classification method applied to the cloud device belongs to the same concept as the technical scheme of the above work order classification method. For details not described in detail in the technical scheme of the work order classification method applied to the cloud device, please refer to the description of the technical scheme of the above work order classification method.

[0166] By applying the solution of the embodiments of this specification, by generating work order distance information based on a work order feature information set, it is possible to accurately classify work orders of different categories, and by calculating the distance matrix to capture the similarities between different work orders, the accuracy of work order classification is improved. In addition, determining a set of similar work orders based on work order distance information and generating corresponding work order summaries not only enhances the accuracy of work order clustering, but also accelerates the extraction and abstraction of information. In this way, different categories of work orders can be quickly identified when processing large-scale work order data, thereby improving overall processing efficiency and accuracy.

[0167] In particular, by applying this method to cloud devices, efficient data transmission and processing is achieved. After receiving a work order classification request, the cloud device can quickly generate work order distance information and determine a set of similar work orders based on the target work order feature information, thereby achieving rapid response and real-time classification. The application of this method in a cloud environment not only effectively reduces the computing burden of local devices, but also speeds up the generation of data processing and classification results, further improving the intelligence level of work order processing. In general, this technical solution improves data processing efficiency, accuracy, and intelligence by optimizing the work order classification process in a cloud environment, providing cloud devices with a more accurate and efficient work order classification solution.

[0168] See also Figure 6 , Figure 6 An architecture diagram of a data classification system provided by an embodiment of the present specification is shown, and the data classification system may include a client 100 and a server 200;

[0169] The client 100 is used to send a data classification request carrying a data set to be classified to the server 200;

[0170] The server 200 is used to receive a data classification request; obtain a target data feature information set corresponding to a data set to be classified, wherein the target data feature information in the target data feature information set has the same dimension; generate data distance information corresponding to the data set to be classified based on the target data feature information set, wherein the data distance information represents the degree of similarity between each data to be classified in the data set to be classified; determine at least one similar data set and a similar data summary corresponding to each similar data set in the data set to be classified according to the data distance information; generate a data classification result based on each similar data set and the similar data summary corresponding to each similar data set; and send the data classification result to the client 100;

[0171] The client 100 is also used to receive the data classification result sent by the server 200.

[0172] By applying the solution of the embodiments of this specification, the server can accurately identify and classify different data categories by receiving the data classification request sent by the client and generating data distance information based on the target data feature information, thereby improving the accuracy of data classification. By calculating the data distance matrix and constructing the feature matrix, this technology can effectively capture the correlation between different data, allowing the system to efficiently and accurately process a large amount of data to be classified.

[0173] In addition, by determining the same data set and generating the same data summary on the server side, the system not only improves the accuracy of data clustering, but also accelerates the extraction of key information. The close collaboration between the client and the server enables the timely feedback and transmission of data classification results to the client, achieving rapid response and optimized processing. Combining the processing of target data feature information with global data analysis, the system effectively improves the ability to process massive data, optimizes data classification efficiency, and ensures the accuracy and intelligence level of classification results. In general, the data classification system provides a reliable solution for the intelligent classification of large-scale complex data through efficient feature information generation and data processing mechanisms, and significantly improves the overall efficiency and accuracy of data processing.

[0174] The data classification system may include multiple clients 100 and a server 200, wherein the client 100 may be referred to as a terminal device and the server 200 may be referred to as a cloud device. Multiple clients 100 may establish a communication connection through the server 200. In the data classification scenario, the server 200 is used to provide data classification services between multiple clients 100. Multiple clients 100 may serve as a sender or a receiver, respectively, and realize communication through the server 200.

[0175] The user can interact with the server 200 through the client 100 to receive data sent by other clients 100, or send data to other clients 100, etc. In the data classification scenario, the user can publish a data stream to the server 200 through the client 100, and the server 200 generates a data classification result based on the data stream and pushes the data classification result to other clients that have established communication.

[0176] The client 100 and the server 200 are connected via a network. The network provides a medium for a communication link between the client 100 and the server 200. The network may include various connection types, such as wired or wireless communication links or optical fiber cables, etc. The data transmitted by the client 100 may need to be encoded, transcoded, compressed, etc. before being released to the server 200.

[0177] The client 100 can be a browser, an APP (Application), or a web application such as an H5 (HyperText Markup Language 5, Hypertext Markup Language Version 5) application, or a light application (also known as a mini-program, a lightweight application) or a cloud application, etc. The client 100 can be based on the software development kit (SDK, Software Development Kit) of the corresponding service provided by the server 200, such as based on the real-time communication (RTC, Real Time Communication) SDK development and acquisition. The client 100 can be deployed in an electronic device and needs to rely on the device to run or some APPs in the device to run. For example, the electronic device can have a display screen and support information browsing, such as a personal mobile terminal such as a mobile phone, a tablet computer, a personal computer, etc. Various other types of applications can also be configured in the electronic device, such as human-computer dialogue applications, model training applications, text processing applications, web browser applications, shopping applications, search applications, instant messaging tools, email clients, social platform software, etc.

[0178] The server 200 may include servers that provide various services, such as servers that provide communication services to multiple clients, servers for background training that support models used on clients, and servers that process data sent by clients. It should be noted that the server 200 can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. The server can also be a server of a distributed system, or a server combined with a blockchain. The server can also be a cloud server for basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks (CDN, Content Delivery Network), and big data and artificial intelligence platforms, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology.

[0179] It is worth noting that the data classification method provided in the embodiments of this specification is generally executed by the server, but in other embodiments of this specification, the client may also have similar functions to the server, thereby executing the data classification method provided in the embodiments of this specification. In other embodiments, the data classification method provided in the embodiments of this specification may also be jointly executed by the client and the server.

[0180] The following combination Figure 7 , taking the application of the data classification method provided in this specification in image classification as an example, the data classification method is further described. Figure 7 A processing flow chart of an image classification method provided by an embodiment of the present specification is shown, which specifically includes the following steps.

[0181] Step 702: For each image to be classified, a unified image feature extraction method is used to obtain a 1024-dimensional initial image feature information corresponding to each image to be classified.

[0182] In practical applications, the dimension of the initial image feature information can be adjusted as needed.

[0183] Step 704: The extracted initial image feature information is composed of a data feature set and input into the generative large model, and the 10 subsequent features of each initial image feature information are obtained through prediction by the generative large model, and each initial image feature information is fused with its corresponding subsequent features to obtain the target image feature information.

[0184] Step 706: Based on the target image feature information corresponding to each image, a feature matrix of the image set dimension is constructed, and a distance matrix between images in the image set is calculated and obtained.

[0185] Step 708: Obtain the coordinates corresponding to the characteristic distance values ​​less than the set threshold in the distance matrix, and generate data sequence information using the position coordinates corresponding to each characteristic distance.

[0186] Step 710: Use the data sequence information to construct a feature matrix of the image set dimension for the target image feature information, rearrange it to obtain a data sorting feature matrix, and process the data sorting feature matrix through a convolution operation to generate a data global feature matrix.

[0187] Step 712: Determine the global feature set corresponding to each data of the same type from the global feature matrix according to the above data sequence information.

[0188] Step 714: A plurality of global features corresponding to each of the same type of data sets are fused to obtain category feature information corresponding to each of the same type of data sets.

[0189] Step 716: Input the category feature information corresponding to each similar data set into the text generation model to generate summary text corresponding to each similar data set.

[0190] Step 718: Based on the above data sequence information, multiple images belonging to the same data set are determined in the image set, and summary texts corresponding to the same type of images are determined.

[0191] By applying the scheme of the embodiments of this specification, the generation and processing of the target image feature information set is introduced into the image classification process, and this method breaks through the limitations of traditional image classification technology in processing large-scale and complex image data. Specifically, the initial image feature information is predicted using a generative large model and fused with subsequent features to obtain the target image feature information. This process not only significantly improves the expressive power of image features, but also deepens the intrinsic relationship between images. By constructing a feature matrix and calculating the distance matrix between images, similar images can be accurately identified, thereby achieving efficient classification and information extraction.

[0192] In addition, by combining data sequence information with the construction of a global feature matrix, the system effectively improves the accuracy and efficiency of image classification. In the process of clustering and summarizing similar images, it not only accelerates the extraction of key information, but also ensures that the image classification results are highly intelligent. By inputting the category feature information of similar data sets into the text generation model to generate summary text, the system can automatically summarize the core features of each type of image and output efficient and accurate classification results. Overall, this method provides an efficient and accurate solution in image classification, which can effectively improve processing efficiency when facing large-scale image data, and provide strong support for deep mining and intelligent analysis of image data.

[0193] Corresponding to the above method embodiment, this specification also provides a data classification device embodiment, Figure 8 FIG. 1 shows a schematic diagram of the structure of a data classification device provided by an embodiment of the present specification. Figure 8 As shown, the device comprises:

[0194] The acquisition module 802 is configured to acquire a target data feature information set corresponding to the data set to be classified, wherein the target data feature information in the target data feature information set has the same dimension;

[0195] A generating module 804 is configured to generate data distance information corresponding to the to-be-classified data set based on the target data feature information set, wherein the data distance information represents the degree of similarity between each to-be-classified data in the to-be-classified data set;

[0196] The classification module 806 is configured to determine at least one data set of the same type and a data summary of the same type corresponding to each data set of the same type in the data set to be classified according to the data distance information.

[0197] Optionally, the acquisition module 802 is further configured to:

[0198] Determine target data to be classified, wherein the target data to be classified is any one of the data to be classified in the data set to be classified;

[0199] Acquire initial feature information corresponding to the target data to be classified, wherein the initial feature information represents the features of the target data to be classified itself in the text dimension;

[0200] Processing the initial feature information based on a text prediction model to obtain target data feature information, wherein the target data feature information represents extended features of the target data to be classified in a text dimension;

[0201] A target data feature information set is obtained according to the target data feature information corresponding to each to-be-classified data in the to-be-classified data set.

[0202] Optionally, the acquisition module 802 is further configured to:

[0203] Inputting the initial feature information into the text prediction model, and obtaining at least one reference feature information output by the text prediction model;

[0204] The target data feature information is generated based on each reference feature information.

[0205] Optionally, the generating module 804 is further configured to:

[0206] Based on each target data feature information in the target data feature information set, generating a target data feature matrix, wherein the target data feature matrix is ​​used to present the corresponding feature representation of each to-be-classified data in the to-be-classified data combination;

[0207] A data distance matrix is ​​calculated according to the target data feature matrix, and data distance information is obtained based on the data distance matrix, wherein the data distance matrix is ​​used to represent the distance relationship between each data to be classified.

[0208] Optionally, the classification module 806 is further configured to:

[0209] Acquire sequence information of the data to be classified according to the data distance information;

[0210] Sorting and processing the target data feature information set based on the sequence information of the data to be classified to obtain data feature sequence information;

[0211] At least one similar data set and similar data summaries corresponding to each similar data set are determined based on the data feature sequence information.

[0212] Optionally, the classification module 806 is further configured to:

[0213] Determine at least one similar data feature set based on the data feature sequence information, wherein the similar data feature set includes at least one target data feature information;

[0214] Generate similar data summaries corresponding to similar data feature sets according to the characteristic information of each target data;

[0215] Based on the data feature sequence information, the first similar data set corresponding to the first similar data feature set is determined, and the similar data summary corresponding to the first similar data feature set is determined to be the similar data summary corresponding to the first similar data set, wherein the first similar data feature set is any one of the similar data feature sets, and the first similar data set is any one of the similar data sets.

[0216] Optionally, the classification module 806 is further configured to:

[0217] Obtain data category feature information corresponding to each similar data feature set according to each target data feature information;

[0218] Based on the data category feature information corresponding to the second similar data feature set, a similar data summary corresponding to the second similar data feature set is generated, wherein the second similar data feature set is any one of the similar data feature sets.

[0219] Optionally, the classification module 806 is further configured to:

[0220] Generate data feature sequence information according to each target data feature information;

[0221] Inputting the data feature sequence information into a preset global feature extraction network to obtain a target data global feature matrix generated by the global feature extraction network;

[0222] Based on the target data global feature matrix, data category feature information corresponding to each similar data feature set is obtained.

[0223] Optionally, the classification module 806 is further configured to:

[0224] Determine at least one target data global feature information corresponding to each similar data feature set in the target data global feature matrix according to the data feature sequence information;

[0225] Based on the global feature information of each target data corresponding to the third similar data feature set, the data category feature information corresponding to the third similar data feature set is generated, wherein the third similar data feature set is any one of the similar data feature sets.

[0226] By applying the solution of the embodiments of this specification, this technology breaks through the efficiency and accuracy bottleneck of traditional data classification methods when processing massive data by introducing the acquisition and processing of target data feature information sets in the data classification device. In particular, the acquisition module uses the text prediction model to generate target data feature information, thereby expanding the initial data features and generating data distance information. This process not only strengthens the inherent relationship between data, but also ensures accurate identification and efficient processing in large-scale data classification by constructing a data feature matrix and calculating the distance matrix between data.

[0227] The above is a schematic scheme of a data classification device of this embodiment. It should be noted that the technical scheme of the data classification device and the technical scheme of the above data classification method belong to the same concept, and the details not described in detail in the technical scheme of the data classification device can be referred to the description of the technical scheme of the above data classification method.

[0228] Fig. 9 The structure block diagram of a computing device 900 provided according to an embodiment of the present specification is shown. The components of the computing device 900 include but are not limited to a memory 910 and a processor 920. The processor 920 is connected to the memory 910 via a bus 930, and a database 950 is used to store data.

[0229] The computing device 900 also includes an access device 940 that enables the computing device 900 to communicate via one or more networks 960. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 940 may include one or more of any type of network interface (e.g., a network interface card (NIC)) that is wired or wireless, such as an IEEE 802.11 wireless local area network (WLAN) wireless interface, a world-wide interoperability for microwave access (Wi-MAX) interface, an Ethernet interface, a universal serial bus (USB) interface, a cellular network interface, a Bluetooth interface, and a near field communication (NFC).

[0230] In one embodiment of the present specification, the above components of the computing device 900 and Fig. 9 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Fig. 9 The computing device structure block diagram shown is only for the purpose of illustration, and is not intended to limit the scope of this specification. Those skilled in the art can add or replace other components as needed.

[0231] The computing device 900 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.), or other types of mobile devices, or a stationary computing device such as a desktop computer or a personal computer (PC). The computing device 900 may also be a mobile or stationary server.

[0232] Among them, the processor 920 is used to execute the following computer executable instructions, which, when executed by the processor, implement the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and steps of the work order classification method applied to cloud devices.

[0233] The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and technical scheme of work order classification method applied to cloud devices belong to the same concept. For details not described in detail in the technical scheme of the computing device, please refer to the description of the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and technical scheme of work order classification method applied to cloud devices.

[0234] An embodiment of the present specification also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and steps of the work order classification method applied to cloud devices.

[0235] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and technical scheme of the work order classification method applied to cloud devices belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and technical scheme of the work order classification method applied to cloud devices.

[0236] An embodiment of the present specification also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and steps of the work order classification method applied to cloud devices.

[0237] The above is a schematic scheme of a computer program of this embodiment. It should be noted that the technical scheme of the computer program and the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and technical scheme of work order classification method applied to cloud devices belong to the same concept, and the details not described in detail in the technical scheme of the computer program can be referred to the description of the above-mentioned data classification method, work order classification method, data classification method applied to cloud devices, and technical scheme of work order classification method applied to cloud devices.

[0238] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0239] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0240] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0241] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0242] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to only the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can well understand and use this specification. This specification is only limited by the claims and their full scope and equivalents.

Claims

1. A data classification method, comprising: Acquire a target data feature information set corresponding to the data set to be classified, wherein the target data feature information in the target data feature information set has the same dimension, and the data to be classified includes at least one of an image, a text, and a work order; Generating data distance information corresponding to the to-be-classified data set based on the target data feature information set, wherein the data distance information represents the degree of similarity between each to-be-classified data in the to-be-classified data set; Generate data feature sequence information according to each target data feature information in the target data feature information set; input the data feature sequence information into a preset global feature extraction network to obtain a target data global feature matrix generated by the global feature extraction network; determine at least one similar data set and similar data summaries corresponding to each similar data set in the data set to be classified according to the target data global feature matrix and the data distance information, wherein the target data global feature matrix represents the feature distribution of each data to be classified in the data set to be classified in the entire data set to be classified.

2. The method according to claim 1, obtaining a target data feature information set corresponding to the data set to be classified, comprising: Determine target data to be classified, wherein the target data to be classified is any one of the data to be classified in the data set to be classified; Acquire initial feature information corresponding to the target data to be classified, wherein the initial feature information represents features of the target data to be classified in a text dimension; Processing the initial feature information based on a text prediction model to obtain target data feature information, wherein the target data feature information represents extended features of the target data to be classified in a text dimension; A target data feature information set is obtained according to the target data feature information corresponding to each to-be-classified data in the to-be-classified data set.

3. The method according to claim 2, wherein the initial feature information is processed based on a text prediction model to obtain target data feature information, comprising: Inputting the initial feature information into the text prediction model, and obtaining at least one reference feature information output by the text prediction model; The target data feature information is generated based on each reference feature information.

4. The method according to claim 1, generating data distance information corresponding to the to-be-classified data set based on the target data feature information set, comprising: Based on each target data feature information in the target data feature information set, generating a target data feature matrix, wherein the target data feature matrix is ​​used to present the corresponding feature representation of each to-be-classified data in the to-be-classified data combination; A data distance matrix is ​​calculated according to the target data feature matrix, and data distance information is obtained based on the data distance matrix, wherein the data distance matrix is ​​used to represent the distance relationship between each data to be classified.

5. The method according to claim 1, determining at least one similar data set and similar data summaries corresponding to each similar data set in the data set to be classified according to the target data global feature matrix and the data distance information, comprising: Acquire sequence information of the data to be classified according to the data distance information; Sorting and processing the target data feature information set based on the sequence information of the data to be classified to obtain data feature sequence information; At least one similar data set and similar data summaries corresponding to each similar data set are determined based on the data feature sequence information.

6. The method according to claim 5, determining at least one similar data set and similar data summaries corresponding to each similar data set based on the data feature sequence information, comprising: Determine at least one similar data feature set based on the data feature sequence information, wherein the similar data feature set includes at least one target data feature information; Generate similar data summaries corresponding to similar data feature sets according to the characteristic information of each target data; Based on the data feature sequence information, the first similar data set corresponding to the first similar data feature set is determined, and the similar data summary corresponding to the first similar data feature set is determined to be the similar data summary corresponding to the first similar data set, wherein the first similar data feature set is any one of the similar data feature sets, and the first similar data set is any one of the similar data sets.

7. The method according to claim 6, generating a summary of similar data corresponding to each similar data feature set according to each target data feature information, comprising: Acquire data category feature information corresponding to each similar data feature set based on the target data global feature matrix; Based on the data category feature information corresponding to the second similar data feature set, a similar data summary corresponding to the second similar data feature set is generated, wherein the second similar data feature set is any one of the similar data feature sets.

8. The method according to claim 7, obtaining data category feature information corresponding to each similar data feature set based on the target data global feature matrix, comprises: Determine at least one target data global feature information corresponding to each similar data feature set in the target data global feature matrix according to the data feature sequence information; Based on the global feature information of each target data corresponding to the third similar data feature set, the data category feature information corresponding to the third similar data feature set is generated, wherein the third similar data feature set is any one of the similar data feature sets.

9. A work order classification method, comprising: Acquire a target work order feature information set corresponding to the work order data set, wherein the target work order feature information in the target work order feature information set has the same dimension; Generating work order distance information corresponding to the work order data set based on the target work order feature information set, wherein the work order distance information represents the similarity between each work order data in the work order data set; Generate data feature sequence information based on the feature information of each target work order in the target work order feature information set; input the data feature sequence information into a preset global feature extraction network to obtain a target data global feature matrix generated by the global feature extraction network; determine at least one similar work order set and similar work order summaries corresponding to each similar work order set in the work order data set based on the target data global feature matrix and the work order distance information, wherein the target data global feature matrix represents the feature distribution of each work order data in the work order data set in the entire work order data set.

10. A data classification method, applied to a cloud device, comprising: A data classification request sent by a receiving end-side device, wherein the data classification request carries a set of data to be classified, wherein the data to be classified includes at least one of an image, text, and a work order; Acquire a target data feature information set corresponding to the data set to be classified, wherein the target data feature information in the target data feature information set has the same dimension; Generating data distance information corresponding to the to-be-classified data set based on the target data feature information set, wherein the data distance information represents the degree of similarity between each to-be-classified data in the to-be-classified data set; Generate data feature sequence information according to each target data feature information in the target data feature information set; input the data feature sequence information into a preset global feature extraction network to obtain a target data global feature matrix generated by the global feature extraction network; determine at least one similar data set and similar data summaries corresponding to each similar data set in the data set to be classified according to the target data global feature matrix and the data distance information, wherein the target data global feature matrix represents the feature distribution of each data to be classified in the data set to be classified in the entire data set to be classified; A data classification result is generated based on each similar data set and a similar data summary corresponding to each similar data set, and the data classification result is sent to the terminal side device.

11. A work order classification method, applied to a cloud device, comprising: A work order classification request sent by a receiving end-side device, wherein the work order classification request includes a work order data set; Acquire a target work order feature information set corresponding to the work order data set, wherein the target work order feature information in the target work order feature information set has the same dimension; Generating work order distance information corresponding to the work order data set based on the target work order feature information set, wherein the work order distance information represents the similarity between each work order data in the work order data set; Generate data feature sequence information according to the feature information of each target work order in the target work order feature information set; input the data feature sequence information into a preset global feature extraction network to obtain a target data global feature matrix generated by the global feature extraction network; determine at least one similar work order set and similar work order summaries corresponding to each similar work order set in the work order data set according to the target data global feature matrix and the work order distance information, wherein the target data global feature matrix represents the feature distribution of each work order data in the work order data set in the entire work order data set; A work order classification result is generated based on the similar work order summaries corresponding to each similar work order set, and the work order classification result is sent to the terminal side device.

12. A computing device comprising: Memory and processor; The memory is used to store computer programs / instructions, and the processor is used to execute the computer programs / instructions. When the computer program / instructions are executed by the processor, the steps of the method described in any one of claims 1 to 11 are implemented.

13. A computer-readable storage medium storing a computer program / instruction, wherein the computer program / instruction, when executed by a processor, implements the steps of the method according to any one of claims 1 to 11.

14. A computer program product, comprising a computer program / instruction, which, when executed by a processor, implements the steps of the method according to any one of claims 1 to 11.

Images