Intelligent file classification method and device based on multi-model fusion and computer equipment

Through multi-model fusion and ensemble learning methods, the problem of low efficiency and accuracy of existing file classification methods is solved, efficient and accurate file classification is achieved, labor costs are reduced, and the adaptability and performance of the model is improved.

CN119068494APending Publication Date: 2024-12-03SHANGHAI BESTON LEGAL INFORMATION TECH CO LTD
View PDF 0 Cites 1 Cited by

Patent Information

Application Number
CN202411094094.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-09
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing file classification method has low efficiency and accuracy, and relies on manual operations and a single machine learning model. The data preprocessing is inaccurate, the manual annotation dependency is strong, and the model generalization ability is limited. It is difficult to adapt to new types of files, which increases the system maintenance cost. and complexity.

Method used

Using the intelligent file classification method of multi-model fusion, the data is preprocessed, multiple classification models are trained, and the optimal model is selected through cross-validation, combined with the integrated learning method for classification identification, the model performance is tested using the verification data set, and the model performance is tested in Update the model if necessary to accommodate new data and requirements.

Benefits of technology

It improves the efficiency and accuracy of file classification, reduces manual operation requirements, reduces labor costs, enhances the generalization ability and adaptability of the model, and ensures high reliability and continuous performance improvement of classification results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119068494A_ABST
    Figure CN119068494A_ABST
Patent Text Reader

Abstract

The invention relates to an intelligent file classification method and device based on multi-model fusion and computer equipment, and the method comprises the steps: carrying out the preprocessing of collected data, and taking the preprocessed data as model training data; training a plurality of classification models by using the model training data to obtain a plurality of trained classification models; selecting an optimal classification model from the plurality of trained classification models; and performing classification identification on a to-be-classified file by utilizing the optimal classification model to obtain a classification result of the to-be-classified file. Through the file classification method and device, the problem that file classification methods in related technologies are low in efficiency and accuracy is solved, and the effect of improving the efficiency and accuracy of the file classification methods is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of image recognition and intelligent file classification, and in particular to an intelligent file classification method, device and computer equipment based on multi-model fusion. Background Art

[0002] In the digital age, processing and understanding large amounts of image data has become a critical task. Existing technologies typically perform optical character recognition (OCR) on paper and digital documents, then perform vectorization to infer the document type. Manually preparing data for model training relies heavily on subjective judgment. Misjudgments can lead to inaccurate classifications, impacting the overall accuracy of the model. High-quality data annotation is key to improving model performance, but acquiring large amounts of accurately labeled, high-quality data is costly and time-consuming.

[0003] Traditional image classification methods rely on manual operations or single machine learning models. These methods are not only inefficient when processing large numbers of heterogeneous images, but also have limitations in accuracy and consistency, as follows:

[0004] 1. Data preprocessing challenges: OCR recognition and vectorization processing, as key steps in data preprocessing, may be affected by factors such as document quality and format diversity, resulting in inaccurate feature extraction and, in turn, affecting classification results.

[0005] 2. Dependence on manual labeling: Although manual labeling can improve the efficiency of data cleaning, this process still requires a lot of human resources and is susceptible to subjective judgment, which may limit the degree of automation of the entire system.

[0006] 3. Limitations in model generalization: Model accuracy is highly dependent on the quality and diversity of training data. If the training data is biased or insufficient, the model's generalization ability in real-world applications may be limited, affecting classification accuracy.

[0007] 4. Adaptability to new file types: As the business evolves, new file types or formats may emerge. Existing models may require additional adjustments and training to adapt to these changes, which increases the maintenance cost and complexity of the system.

[0008] Currently, the file classification methods in related technologies have problems of low efficiency and accuracy, and no effective solution has been proposed. Summary of the Invention

[0009] The purpose of this application is to address the deficiencies in the prior art and to provide an intelligent file classification method, apparatus, computer device, and computer-readable storage medium based on multi-model fusion, so as to at least solve the problem of low efficiency and accuracy of file classification methods in related technologies.

[0010] To achieve the above objectives, the technical solutions adopted in this application are:

[0011] In a first aspect, an embodiment of the present application provides an intelligent file classification method based on multi-model fusion, the method comprising:

[0012] Preprocess the collected data and use the preprocessed data as model training data;

[0013] Using the model training data to train multiple classification models to obtain multiple trained classification models;

[0014] Selecting an optimal classification model from the multiple trained classification models;

[0015] The optimal classification model is used to classify and identify the files to be classified, and a classification result of the files to be classified is obtained.

[0016] In some embodiments, the preprocessing of the collected data includes:

[0017] Determining the type of the collected data, wherein the type includes: text, image;

[0018] Extracting text content from text data and converting the text content into a numerical vector;

[0019] Extracting image features from image data and converting the image features into numerical vectors;

[0020] Compute the cosine similarity between numeric vectors;

[0021] The collected data is deduplicated and filtered according to the calculation result of the cosine similarity.

[0022] In some embodiments, the pre-processing of the collected data further comprises:

[0023] Find the intersection between numeric vectors;

[0024] The data in the intersection are labeled, and the labeled data are used as the model training data.

[0025] In some embodiments, selecting the optimal classification model from the multiple trained classification models includes:

[0026] Divide the model training data into K subsets, each subset is used as a test set in turn, and the remaining K-1 subsets constitute the training set;

[0027] Repeat the following process K times: use the training set to train multiple classification models respectively, and use the test set to evaluate the performance parameters of the multiple trained classification models;

[0028] Get the average value of the performance parameters of each trained classification model;

[0029] The optimal classification model is selected according to the average value.

[0030] In some embodiments, after selecting the optimal classification model from the multiple trained classification models, the method includes:

[0031] If multiple optimal classification models are selected, the multiple optimal classification models are combined into an integrated model;

[0032] Combining the classification and recognition results of multiple optimal classification models using an ensemble learning method;

[0033] Performing parameter adjustment processing on the integrated model to obtain a processed integrated model;

[0034] The performance of the processed integrated model is tested using the validation dataset.

[0035] In some embodiments, after obtaining the classification result of the file to be classified, the method further includes:

[0036] The trained classification model is updated using the classification result of the file to be classified.

[0037] In a second aspect, an embodiment of the present application provides an intelligent file classification device based on multi-model fusion, comprising:

[0038] A preprocessing unit is used to preprocess the collected data and use the preprocessed data as model training data;

[0039] A training unit, configured to train a plurality of classification models using the model training data to obtain a plurality of trained classification models;

[0040] A selection unit, configured to select an optimal classification model from the plurality of trained classification models;

[0041] The classification unit is used to classify and identify the files to be classified using the optimal classification model to obtain a classification result of the files to be classified.

[0042] In some embodiments, the pre-processing unit includes:

[0043] A determination module, configured to determine the type of the collected data, wherein the type includes: text, image;

[0044] The first extraction module is used to extract text content from data of type text and convert the text content into a numerical vector;

[0045] The second extraction module is used to extract image features from image-type data and convert the image features into numerical vectors.

[0046] A calculation module, used to calculate the cosine similarity between numerical vectors;

[0047] The deduplication module is used to perform deduplication and filtering on the collected data based on the calculation results of cosine similarity.

[0048] Thirdly, embodiments of this application provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the intelligent file classification method based on multi-model fusion as described in the first aspect above.

[0049] Fourthly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the intelligent file classification method based on multi-model fusion as described in the first aspect above.

[0050] This application adopts the above technical solution. Compared with the prior art, the embodiments of this application provide an intelligent file classification method based on multi-model fusion. The method includes: preprocessing the collected data and using the preprocessed data as model training data; training multiple classification models using the model training data to obtain multiple trained classification models; selecting the optimal classification model from the multiple trained classification models; and using the optimal classification model to classify and identify the file to be classified to obtain the classification result of the file to be classified. This method solves the problem of low efficiency and accuracy in related file classification methods, and achieves the effect of improving the efficiency and accuracy of file classification methods.

[0051] The details of one or more embodiments of the present application are set forth in the following drawings and description to make other features, objects, and advantages of the present application more readily apparent. Attached Figure Description

[0052] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0053] Figure 1 is a structural block diagram of a mobile terminal according to an embodiment of the present application;

[0054] Figure 2 is a flow chart of an intelligent file classification method based on multi-model fusion according to an embodiment of the present application;

[0055] Figure 3 1 is a flow chart of an intelligent file classification method based on multi-model fusion according to a preferred embodiment of the present application;

[0056] Figure 4 is a schematic diagram of a specific sub-process of the intelligent file classification method according to a preferred embodiment of the present application;

[0057] Figure 5 This is a structural block diagram of an intelligent file classification device based on multi-model fusion according to an embodiment of the present application;

[0058] Figure 6 Schematic diagram of the hardware structure of a computer device according to an embodiment of the present application. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solutions and advantages of this application more clearly understood, the present application is described and illustrated below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are merely used to explain this application and are not intended to limit this application. Based on the embodiments provided in this application, all other embodiments obtained by those of ordinary skill in the art without making any creative efforts are within the scope of protection of this application.

[0060] Obviously, the drawings described below are merely examples or embodiments of the present application. Those skilled in the art can, without inventive effort, apply the present application to other similar scenarios based on these drawings. Furthermore, it is also understood that, although the effort involved in such a development process may be complex and lengthy, for those skilled in the art related to the content disclosed in this application, changes in design, manufacturing, or production based on the technical content disclosed in this application are merely conventional technical means and should not be construed as an insufficiency of the content disclosed in this application.

[0061] References to "embodiments" in this application mean that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it refer to independent or alternative embodiments that are mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described in this application may be combined with other embodiments unless there is a conflict.

[0062] Unless otherwise defined, the technical or scientific terms used in this application should have the ordinary meaning understood by a person of ordinary skill in the technical field to which this application belongs. The words "one", "a", "the" and the like used in this application do not indicate a limit on quantity and may indicate the singular or plural. The terms "include", "comprise", "have" and any variations thereof used in this application are intended to cover non-exclusive inclusions; for example, a process, method, system, product or device that includes a series of steps or modules (units) is not limited to the listed steps or units, but may also include steps or units that are not listed, or may also include other steps or units that are inherent to these processes, methods, products or devices. The words "connect", "connected", "coupled" and the like used in this application are not limited to physical or mechanical connections, but may include electrical connections, whether direct or indirect. The word "multiple" used in this application refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can mean: A exists alone, A and B exist at the same time, and B exists alone. The character " / " generally indicates that the objects before and after are in an "or" relationship. The terms "first", "second", "third", etc. involved in this application are only used to distinguish similar objects and do not represent a specific order for the objects.

[0063] This embodiment provides a mobile terminal. Figure 1 : is a structural block diagram of a mobile terminal according to an embodiment of the present application. Figure 1 As shown, the mobile terminal includes components such as a radio frequency (RF) circuit 110, a memory 120, an input unit 130, a display unit 140, a sensor 150, an audio circuit 160, a wireless fidelity (WiFi) module 170, a processor 180, and a power supply 190. Those skilled in the art will understand that Figure 1 The structure of the mobile terminal shown in the figure does not constitute a limitation to the mobile terminal, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0064] The following combination Figure 1A detailed introduction to the various components of the mobile terminal is given below:

[0065] The RF circuit 110 can be used to receive and send signals during information transmission or calls. In particular, after receiving downlink information from the base station, it is sent to the processor 180 for processing; in addition, the designed uplink data is sent to the base station. Generally, the RF circuit includes but is not limited to an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (LNA), a duplexer, etc. In addition, the RF circuit 110 can also communicate with the network and other devices through wireless communication. The above-mentioned wireless communication can use any communication standard or protocol, including but not limited to Global System of Mobile Communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Messaging Service (SMS), etc.

[0066] The memory 120 can be used to store software programs and modules. The processor 180 executes various functional applications and data processing of the mobile terminal by running the software programs and modules stored in the memory 120. The memory 120 may mainly include a program storage area and a data storage area. The program storage area may store an operating system, at least one application required for a function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created based on the use of the mobile terminal (such as audio data, a phone book, etc.). In addition, the memory 120 may include a high-speed random access memory and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage device.

[0067] The input unit 130 can be used to receive input digital or character information, and to generate key signal input related to the user settings and function control of the mobile terminal. Specifically, the input unit 130 may include a touch panel 131 and other input devices 132. The touch panel 131, also known as a touch screen, can collect user touch operations on or near it (such as operations performed by the user using any suitable object or accessory such as a finger, stylus, etc. on or near the touch panel 131) and drive the corresponding connection device according to a pre-set program. Optionally, the touch panel 131 may include two parts: a touch detection device and a touch controller. Among them, the touch detection device detects the user's touch direction, detects the signal caused by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device and converts it into touch point coordinates, which are then sent to the processor 180, and can receive commands sent by the processor 180 and execute them. In addition, the touch panel 131 can be implemented using various types such as resistive, capacitive, infrared, and surface acoustic wave. In addition to the touch panel 131, the input unit 130 may also include other input devices 132. Specifically, other input devices 132 may include, but are not limited to, one or more of the following: physical keyboard, function keys (such as volume control buttons, power buttons, etc.), trackball, mouse, joystick, etc.

[0068] The display unit 140 can be used to display information input by the user or information provided to the user and various menus of the mobile terminal. The display unit 140 may include a display panel 141. Optionally, the display panel 141 may be configured in the form of a liquid crystal display (LCD), an organic light-emitting diode (OLED), etc. Further, the touch panel 131 may cover the display panel 141. When the touch panel 131 detects a touch operation on or near it, it is transmitted to the processor 180 to determine the type of touch event. Subsequently, the processor 180 provides corresponding visual output on the display panel 141 according to the type of touch event. Although in Figure 1 In this embodiment, the touch panel 131 and the display panel 141 are two separate components to realize the input and output functions of the mobile terminal. However, in some embodiments, the touch panel 131 and the display panel 141 can be integrated to realize the input and output functions of the mobile terminal.

[0069] The mobile terminal may also include at least one sensor 150, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display panel 141 according to the brightness of the ambient light, and the proximity sensor may turn off the display panel 141 and / or the backlight when the mobile terminal is moved to the ear. As a type of motion sensor, the accelerometer sensor can detect the magnitude of acceleration in all directions (generally three axes), and can detect the magnitude and direction of gravity when stationary. It can be used for applications that identify the posture of the mobile terminal (such as horizontal and vertical screen switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer, tapping), etc.; as for other sensors that can be configured in the mobile terminal, such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc., they will not be described here.

[0070] The speaker 161 and microphone 162 in the audio circuit 160 provide an audio interface between the user and the mobile terminal. The audio circuit 160 converts received audio data into electrical signals and transmits them to the speaker 161, which then converts the signals into sound signals for output. Furthermore, the microphone 162 converts the collected sound signals into electrical signals, which are then received by the audio circuit 160 and converted into audio data. The audio data is then processed by the processor 180 and transmitted to, for example, another mobile terminal via the RF circuit 110, or the audio data is output to the memory 120 for further processing.

[0071] WiFi is a short-range wireless transmission technology. Mobile terminals using a WiFi module 170 can help users send and receive emails, browse web pages, and access streaming media, providing users with wireless broadband internet access. Although Figure 1 WiFi module 170 is shown, but it is understood that it is not a necessary component of a mobile terminal and can be omitted or replaced with other short-range wireless transmission modules, such as Zigbee modules or WAPI modules, as needed without changing the nature of the invention.

[0072] The processor 180 is the control center of the mobile terminal, connecting all components of the mobile terminal using various interfaces and circuits. By running or executing software programs and / or modules stored in the memory 120 and accessing data stored in the memory 120, it executes various functions of the mobile terminal and processes data, thereby providing overall monitoring of the mobile terminal. Optionally, the processor 180 may include one or more processing units; preferably, the processor 180 may integrate an application processor and a modem processor. The application processor primarily handles the operating system, user interface, and application programs, while the modem processor primarily handles wireless communications. It is understood that the modem processor may not be integrated into the processor 180.

[0073] The mobile terminal also includes a power supply 190 (such as a battery) for supplying power to various components. Preferably, the power supply can be logically connected to the processor 180 through a power management system, thereby managing charging, discharging, and power consumption through the power management system.

[0074] Although not shown, mobile terminals may also include cameras, Bluetooth modules, etc., which will not be described in detail here.

[0075] In this embodiment, the processor 180 is configured to:

[0076] The collected data is preprocessed and used as training data for the model.

[0077] Using the model training data to train multiple classification models to obtain multiple trained classification models;

[0078] Selecting an optimal classification model from the multiple trained classification models;

[0079] The optimal classification model is used to classify and identify the file to be classified, and the classification result of the file to be classified is obtained.

[0080] In some embodiments, the processor 180 is further configured to:

[0081] The type of the collected data is determined, wherein the type includes: text and images;

[0082] Extracting text content from text data and converting the text content into a numerical vector;

[0083] Extract image features from image-type data and convert the image features into numerical vectors;

[0084] Calculate the cosine similarity between numerical vectors;

[0085] The collected data is deduplicated and filtered according to the calculation result of the cosine similarity.

[0086] In some embodiments, the processor 180 is further configured to:

[0087] Find the intersection of numerical vectors;

[0088] The data in the intersection are labeled, and the labeled data are used as the model training data.

[0089] In some embodiments, the processor 180 is further configured to:

[0090] The model training data is divided into K subsets, each subset is used as the test set in turn, and the remaining K-1 subsets form the training set;

[0091] Repeat the following process K times: train multiple classification models using the training set, and evaluate the performance parameters of the multiple trained classification models using the test set;

[0092] Obtain the average performance parameters of each trained classification model;

[0093] The optimal classification model is selected based on the average value.

[0094] In some embodiments, the processor 180 is further configured to:

[0095] If multiple optimal classification models are selected, then these multiple optimal classification models are combined into an ensemble model.

[0096] The classification and recognition results of multiple optimal classification models are combined using an ensemble learning method;

[0097] The integrated model is subjected to parameter adjustment to obtain the processed integrated model;

[0098] The performance of the processed ensemble model was tested using a validation dataset.

[0099] In some embodiments, the processor 180 is further configured to:

[0100] The trained classification model is updated using the classification results of the files to be classified.

[0101] This embodiment provides an intelligent file classification method based on multi-model fusion. Figure 2 This is a flowchart of an intelligent file classification method based on multi-model fusion according to an embodiment of this application, such as... Figure 2 As shown, the process includes the following steps:

[0102] Step S201: Preprocess the collected data and use the preprocessed data as model training data;

[0103] Step S202: Train multiple classification models using the model training data to obtain multiple trained classification models;

[0104] Step S203: Select the optimal classification model from the plurality of trained classification models;

[0105] Step S204: Use the optimal classification model to classify and identify the file to be classified, and obtain the classification result of the file to be classified.

[0106] It should be noted that the aforementioned multi-model training employs various types of models, such as text classification models and image recognition models, and uses cross-validation to select the optimal model combination. For text data, feature extraction methods may include, but are not limited to, bag-of-words models, TF-IDF, and Word2Vec; for image data, methods may include edge detection, color histograms, and deep learning features. The multiple classification models mentioned above may include, but are not limited to, Support Vector Machines (SVM), decision trees, random forests, and neural networks. Each model has its unique algorithm and advantages, and each model is trained independently using the same training dataset to learn document features and classification rules. This application's embodiment improves classification accuracy through multi-model training and significantly broadens the application scope of the document classification system through rich document recognition capabilities.

[0107] In some embodiments, the step S201 of pre-processing the collected data includes:

[0108] The type of the collected data is determined, wherein the type includes: text and images;

[0109] Extracting text content from text data and converting the text content into a numerical vector;

[0110] Extract image features from image-type data and convert the image features into numerical vectors;

[0111] Calculate the cosine similarity between numerical vectors;

[0112] The collected data is deduplicated and filtered according to the calculation result of the cosine similarity.

[0113] The data preprocessing described in this application improves the accuracy of document content recognition and ensures data quality by employing advanced OCR technology and image processing algorithms. This step involves processing the original document, extracting key information, and converting it into a format suitable for further analysis.

[0114] In some embodiments, the step S201 of pre-processing the collected data further includes:

[0115] Find the intersection of numerical vectors;

[0116] The data in the intersection are labeled, and the labeled data are used as the model training data.

[0117] This application's embodiments, through manual review of the initial automatic classification results, can correct errors that may occur during the automatic classification process, thereby improving the overall classification accuracy. Manual annotation not only categorizes documents but also allows for a deeper understanding and analysis of the document content, which is crucial for the training and optimization of the automatic classification algorithm.

[0118] In some embodiments, step S203 of selecting the optimal classification model from the plurality of trained classification models includes:

[0119] The model training data is divided into K subsets, each subset is used as the test set in turn, and the remaining K-1 subsets form the training set;

[0120] Repeat the following process K times: train multiple classification models using the training set, and evaluate the performance parameters of the multiple trained classification models using the test set;

[0121] Obtain the average performance parameters of each trained classification model;

[0122] The optimal classification model is selected based on the average value.

[0123] The embodiment of the present application selects the optimal classification model based on the results of cross-validation, which can achieve the purpose of improving the accuracy of file classification and recognition by using the optimal classification model. The embodiment of the present application also introduces a manual review mechanism assisted by machine learning. After the initial automatic classification, the system will mark files that may be inaccurately classified or require further confirmation and hand them over to manual review. This not only improves the accuracy of classification, but also makes manual review work more efficient. At the same time, the system will learn from the results of manual review and continuously optimize its own classification algorithm to achieve a virtuous cycle of human-computer collaboration.

[0124] In some embodiments, step S203, after selecting the optimal classification model from the plurality of trained classification models, includes:

[0125] If multiple optimal classification models are selected, then these multiple optimal classification models are combined into an ensemble model.

[0126] The classification and recognition results of multiple optimal classification models are combined using an ensemble learning method;

[0127] The integrated model is subjected to parameter adjustment to obtain the processed integrated model;

[0128] The performance of the processed ensemble model was tested using a validation dataset.

[0129] The embodiments of this application, through cross-validation and model optimization algorithms, can select the best model combination from multiple models, ensuring high-accuracy classification results under different file types and scenarios.

[0130] In some embodiments, after obtaining the classification result of the file to be classified, step S204 further includes:

[0131] The trained classification model is updated using the classification results of the files to be classified.

[0132] The embodiments of this application update the model based on business feedback to adapt to new data and requirements. This ensures that the model can adapt to new data and business needs, and achieve continuous performance improvement.

[0133] Through the above steps, this application solves the problem of low efficiency and accuracy in document classification methods in related technologies, and achieves the effect of improving the efficiency and accuracy of document classification methods.

[0134] The embodiments of the present application are described and illustrated below through preferred embodiments.

[0135] Figure 3 This is a flow chart of an intelligent file classification method based on multi-model fusion according to a preferred embodiment of the present application. Figure 3 As shown, the execution process specifically includes:

[0136] 1. Data preparation: Collect various documents and materials to lay the foundation for subsequent processing.

[0137] 2. Manual annotation: Manually classify and label documents to establish an initial dataset.

[0138] 3. Model training: Use labeled data to train multiple models, including text, image, and other classification models.

[0139] 4. Model reasoning: Apply the trained model to new documents for automatic classification and reasoning.

[0140] 5. Business Deployment: Integrate the inference model into the business process and classify and process actual documents.

[0141] 6. Collect feedback: Collect feedback data on model performance from business applications.

[0142] 7. Model Optimization: Adjust and optimize the model based on feedback data to improve classification accuracy and efficiency.

[0143] 8. Iterative Loop: Return to the model training step and perform continuous iteration and optimization.

[0144] The specific sub-processes of the intelligent file classification method in the preferred embodiment of this application are as follows: Figure 4 Shown, including:

[0145] 1. Start: Identifies the starting point of a sub-process.

[0146] 2. Data collection: Collect various types of document materials, including text, images and other files in different formats.

[0147] 3. Document type determination: Determine whether the document is text type or image type based on the file content and format.

[0148] 4.OCR processing: Perform OCR recognition on text documents and extract text content.

[0149] 5. Image processing: process image documents and extract image features.

[0150] 6. Text vectorization: Convert the extracted text content into numerical vectors for easy calculation and comparison.

[0151] 7. Image vectorization: Convert the extracted image features into numerical vectors.

[0152] 8. Calculate cosine similarity: Calculate the cosine similarity between document vectors to obtain an approximate result of the similarity between documents.

[0153] 9. Data cleaning: Based on the vectorization results and cosine similarity, data cleaning and preprocessing are performed. Data cleaning usually involves identifying and removing noisy, erroneous, or inconsistent data points in the dataset.

[0154] The data cleaning process may include the following steps:

[0155] Vectorization: First, the document content is converted into numerical vectors. This is typically achieved by extracting features from the document (e.g., keywords, term frequencies, etc.).

[0156] Cosine similarity calculation: Next, the cosine similarity between these vectors is calculated to determine the similarity between documents.

[0157] Similarity analysis: By analyzing the results of cosine similarity, document pairs with abnormally high similarity can be identified. These may be duplicates or very similar documents.

[0158] Deduplication and filtering: Based on the results of similarity analysis, duplicate documents are removed or documents that are irrelevant to the main classification target are filtered out.

[0159] Cleaned datasets can reduce noise interference during model training, thereby improving the model's generalization ability and prediction accuracy.

[0160] 10. Manual classification and labeling: Combining manual operations to classify and label documents improves the accuracy of data cleaning.

[0161] The purpose of manual classification and labeling is to improve the quality of dataset annotation, especially when there may be uncertainty in the automatic classification algorithm. Through the intervention of human experts, the results of automatic classification can be verified and corrected to ensure the accuracy and reliability of the dataset.

[0162] 11. Create intersection: Find the intersection between document vectors for further manual annotation.

[0163] Manual annotation refers to the process of classifying and labeling documents. Specifically, human annotators assign one or more category labels to each document based on its content, format, semantics, and other characteristics. These labels are then used to train and optimize classification models.

[0164] The intersection of these vectors is used to identify documents that may have classification uncertainty during the automatic classification process. These documents are close to each other in vector space and have high cosine similarity, making them difficult to accurately classify using automatic algorithms alone. By finding the intersection, we can filter out documents that require further human review and annotation to ensure that these key documents are correctly classified.

[0165] Finding the intersection is a supplementary step in the manual classification and annotation process. After the initial automatic classification, the system identifies document sets that may require manual review by calculating the cosine similarity between document vectors. Documents in the intersection are selected and then meticulously classified and labeled by human annotators. This process helps improve the quality of the training data, ensuring that the model can learn accurate classification boundaries. The high-quality manually annotated data can then be used to further train and optimize the classification model, forming a closed-loop optimization process. The system learns from the results of manual annotation, continuously adjusting and optimizing the automatic classification algorithm, achieving a virtuous cycle of human-machine collaboration.

[0166] In summary, the intersection is used to filter out documents requiring further manual processing from the initial automatic classification results, while manual annotation is the process of accurately classifying these documents. The combination of these two aims to improve the accuracy and efficiency of the entire document classification system, which embodies one of the innovations of this patent.

[0167] 12. Manual labeling: Manually label the documents in the intersection to form high-quality training data.

[0168] 13. Organize the data into Parquet files: Organize the processed data and labels into Parquet format files to facilitate model training.

[0169] 14. Model training: Use the organized data to train the model.

[0170] 15. Model validation: Validate the model using existing labels to evaluate the accuracy of the model.

[0171] 16. Multi-epoch training: A multi-epoch training process is introduced to ensure that the model fully learns the data characteristics. The optimal model is selected through model validation to ensure the model's effectiveness and accuracy in real-world applications. If the model accuracy is insufficient, multiple epochs of training are performed until it meets the requirements.

[0172] 17. Save model: Save the verified model.

[0173] 18. Select the optimal model: Select the optimal model from multiple models for actual business.

[0174] 19. Deploy the model to the business: Deploy the optimal model to the actual business for document classification.

[0175] 20. Collect business data feedback: Collect model classification results and feedback from the business.

[0176] 21. Inference document type: Use the model to perform classification inference on new documents.

[0177] 22. Model update: Update the model based on business feedback to adapt to new data and requirements.

[0178] 23. Return to model training: Return the updated model to the training step for a new round of iterative optimization.

[0179] The specific steps of the embodiment of the present application in an actual application scenario can be described as follows:

[0180] In the legal technology sector, efficient management and rapid classification of documents such as insurance policies, accident reports, and indictments are key to improving work efficiency. These documents often contain extensive technical terminology and formatting specifications, placing high demands on the accuracy of classification systems.

[0181] The implementation steps for this application scenario include:

[0182] 1. Data preparation:

[0183] A certain amount of electronic legal documents have been collected, including insurance policies, accident reports and indictments.

[0184] These documents were digitized to ensure they could be effectively read by subsequent algorithms and tools.

[0185] 2. Manual annotation:

[0186] Legal professionals meticulously categorize and annotate documents, assigning accurate category labels to each document.

[0187] User-friendly annotation tools have been developed to assist the annotation process and improve the consistency and accuracy of annotations.

[0188] 3. Model Training:

[0189] Using labeled datasets, a multi-model system that includes text analysis and natural language processing techniques was trained.

[0190] Through cross-validation and parameter tuning, the most suitable model combination for legal document classification was selected.

[0191] 4. Model Validation:

[0192] Use independent validation sets to test the model's performance, ensuring its accuracy and reliability in real-world applications.

[0193] After multiple rounds of testing, the model with the best performance was selected for practical application.

[0194] 5. Model optimization:

[0195] Based on feedback from legal experts, the model was further refined and optimized to better suit the characteristics of legal documents.

[0196] For example, the complaint now includes features to identify specific legal provisions and references.

[0197] 6. Deployment and Feedback:

[0198] The trained model is integrated into the document management system, and it begins to automatically classify newly received legal documents.

[0199] A feedback mechanism has been established to allow users to report documents that are not classified accurately, so that the model can continuously learn and improve.

[0200] 7. Continuous iteration:

[0201] As new document types emerge and the legal environment changes, the model should be updated regularly to adapt to new classification needs.

[0202] Regularly conduct comprehensive evaluations and updates to the model to ensure that the document classification system remains efficient and accurate.

[0203] It's important to note that during the data preparation phase, third-party data cleaning services can be considered to improve data quality. During model training, different deep learning frameworks can be experimented with to find the most suitable algorithm for legal document classification. During manual annotation, pre-trained language models can be introduced to assist in recognizing legal terminology and structure within documents, reducing the workload for manual review. During model optimization, expert systems can be considered, combining the knowledge and experience of legal experts to improve the model's ability to handle complex legal documents. Through these implementation steps and optional solutions, legal documents can be effectively classified and managed, improving work efficiency and ensuring the accuracy and professionalism of document management.

[0204] This application can achieve the following technical effects:

[0205] 1. Improved classification accuracy: By combining advanced text recognition and image processing technologies, this application significantly improves the recognition and classification accuracy of various types of documents, ensuring the high reliability of classification results.

[0206] 2. Improved processing efficiency: Automated document classification processes significantly reduce the need for manual operations, thereby speeding up document processing and improving overall work efficiency.

[0207] 3. Reduced labor costs: This application reduces reliance on large amounts of human resources, especially when dealing with large volumes of documents, and helps to reduce long-term labor costs and training expenses.

[0208] 4. Improve user interaction experience: Through an intuitive user interface and simplified operation process, this application improves the convenience of users participating in document classification and enhances the user experience.

[0209] 5. Optimize data organization: Accurate document classification helps to better organize and store data, facilitating subsequent data retrieval and analysis.

[0210] 6. Support for rapid retrieval: The implementation of this application makes document retrieval faster and more accurate. Users can easily find information on similar cases, improving the efficiency of the workflow.

[0211] 7. Facilitating Knowledge Sharing: Through effective document classification, this application supports the sharing and dissemination of knowledge, providing a solid foundation for team collaboration and knowledge management.

[0212] 8. Enhanced data security: Automated processing reduces the possibility of human intervention, thereby reducing the risk of data leakage and misoperation, and enhancing the security of data management.

[0213] It should be noted that the steps shown in the above process or in the flowchart of the accompanying figures can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0214] This embodiment provides an intelligent file classification device based on multi-model fusion, which is used to implement the above-mentioned embodiments and preferred implementation methods. Details that have already been described will not be repeated. As used below, the terms "module," "unit," "subunit," etc. may refer to a combination of software and / or hardware that implements a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation using hardware, or a combination of software and hardware, is also possible and contemplated.

[0215] Figure 5 This is a structural block diagram of an intelligent file classification device based on multi-model fusion according to an embodiment of the present application. Figure 5 As shown, the device includes:

[0216] A preprocessing unit 51 is used to preprocess the collected data and use the preprocessed data as model training data;

[0217] A training unit 52 is configured to train a plurality of classification models using the model training data to obtain a plurality of trained classification models;

[0218] Selection unit 53 is used to select the optimal classification model from the plurality of trained classification models;

[0219] The classification unit 54 is configured to classify and identify the file to be classified using the optimal classification model to obtain a classification result of the file to be classified.

[0220] In some embodiments, the pre-processing unit 51 includes:

[0221] A determination module, configured to determine the type of the collected data, wherein the type includes: text, image;

[0222] The first extraction module is used to extract text content from data of type text and convert the text content into a numerical vector;

[0223] The second extraction module is used to extract image features from image-type data and convert the image features into numerical vectors.

[0224] A calculation module, used to calculate the cosine similarity between numerical vectors;

[0225] The deduplication module is used to perform deduplication and filtering on the collected data based on the calculation results of cosine similarity.

[0226] In some embodiments, the preprocessing unit 51 further includes:

[0227] The find module is used to find the intersection between numerical vectors;

[0228] The annotation module is used to annotate the data within the intersection and use the annotated data as the training data for the model.

[0229] In some embodiments, the selection unit 53 includes:

[0230] The classification module is used to divide the model training data into K subsets, each subset taking turns as the test set, and the remaining K-1 subsets forming the training set;

[0231] A training module is configured to repeat K times the following process: using the training set to train a plurality of classification models respectively, and using the test set to evaluate the performance parameters of the plurality of trained classification models;

[0232] The acquisition module is used to obtain the average performance parameters of each trained classification model;

[0233] The selection module is used to select the optimal classification model based on the average value.

[0234] In some embodiments, the device further includes:

[0235] an integration unit, configured to, after selecting an optimal classification model from the plurality of trained classification models, combine the plurality of optimal classification models into an integration model if multiple optimal classification models are selected;

[0236] The combining unit is used to combine the classification and recognition results of multiple optimal classification models using an ensemble learning method;

[0237] An adjustment unit is used to perform parameter adjustment processing on the integrated model to obtain the processed integrated model;

[0238] The testing unit is used to perform a performance test on the processed integrated model using a validation data set.

[0239] In some embodiments, the device further includes:

[0240] The update unit is used to update the trained classification model using the classification result of the file to be classified after obtaining the classification result of the file to be classified.

[0241] It should be noted that the above modules can be functional modules or program modules, and can be implemented through software or hardware. For modules implemented through hardware, the above modules can be located in the same processor; or the above modules can be located in different processors in any combination.

[0242] An embodiment provides a computer device. The intelligent document classification method based on multi-model fusion, as described in this application embodiment, can be implemented using a computer device. Figure 6 This is a schematic diagram of the hardware structure of a computer device according to an embodiment of this application.

[0243] The computer device may include a processor 61 and a memory 62 storing computer program instructions.

[0244] Specifically, the processor 61 may include a central processing unit (CPU), or an application-specific integrated circuit (ASIC), or may be configured to implement one or more integrated circuits of the embodiments of the present application.

[0245] The memory 62 may include a mass storage device for data or instructions. For example, and not limitingly, the memory 62 may include a hard disk drive (HDD), a floppy disk drive, a solid-state drive (SSD), flash memory, an optical disk drive, a magneto-optical disk drive, magnetic tape, or a Universal Serial Bus (USB) drive, or a combination of two or more of these. Where appropriate, the memory 62 may include removable or non-removable (or fixed) media. Where appropriate, the memory 62 may be internal or external to a data processing device. In a particular embodiment, the memory 62 is non-volatile memory. In a particular embodiment, the memory 62 includes read-only memory (ROM) and random access memory (RAM). Where appropriate, the ROM may be a mask-programmed ROM, a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), an electrically alterable read-only memory (EAROM), or flash memory, or a combination of two or more of these. Where appropriate, the RAM can be Static Random-Access Memory (SRAM) or Dynamic Random-Access Memory (DRAM). DRAM can be Fast Page Mode Dynamic Random-Access Memory (FPMDRAM), Extended Data Out Dynamic Random-Access Memory (EDODRAM), Synchronous Dynamic Random-Access Memory (SDRAM), etc.

[0246] The memory 62 may be used to store or cache various data files required for processing and / or communication, as well as possible computer program instructions executed by the processor 61 .

[0247] The processor 61 reads and executes computer program instructions stored in the memory 62 to implement any one of the intelligent file classification methods based on multi-model fusion in the above embodiments.

[0248] In some embodiments, the computer device may further include a communication interface 63 and a bus 60. Figure 6 As shown, the processor 61, the memory 62, and the communication interface 63 are connected via a bus 60 and communicate with each other.

[0249] The communication interface 63 is used to implement communication between the various modules, devices, units, and / or devices in the embodiments of the present application. The communication interface 63 can also implement data communication with other components such as: external devices, image / data acquisition equipment, databases, external storage, and image / data processing workstations.

[0250] The bus 60 includes hardware, software, or both, and couples the components of the computer device to each other. The bus 60 includes, but is not limited to, at least one of the following: a data bus, an address bus, a control bus, an expansion bus, and a local bus. For example, and not as a limitation, bus 60 may include an Accelerated Graphics Port (AGP) or other graphics bus, an Extended Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a Hyper Transport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an InfiniBand interconnect, a Low Pin Count (LPC) bus, a memory bus, a Micro Channel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local Bus (VLB) bus, or other suitable buses, or a combination of two or more of these. Where appropriate, bus 60 may include one or more buses. Although embodiments herein describe and illustrate a particular bus, this application contemplates any suitable bus or interconnect.

[0251] In addition, in conjunction with the intelligent file classification method based on multi-model fusion in the above embodiments, embodiments of the present application may provide a computer-readable storage medium for implementation. The computer-readable storage medium stores computer program instructions; when the computer program instructions are executed by a processor, any of the intelligent file classification methods based on multi-model fusion in the above embodiments is implemented.

[0252] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0253] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.

Claims

1. An intelligent file classification method based on multi-model fusion, characterized in that: include: Preprocess the collected data and use the preprocessed data as model training data; Using the model training data to train multiple classification models to obtain multiple trained classification models; Selecting an optimal classification model from the multiple trained classification models; The optimal classification model is used to classify and identify the files to be classified to obtain the classification results of the files to be classified.

2. The method according to claim 1, characterized in that The preprocessing of the collected data includes: Determine the type of the collected data, wherein the type includes: text, image; Extracting text content from data of text type and converting the text content into a numerical vector; Extracting image features from image data and converting the image features into numerical vectors; Compute cosine similarity between numeric vectors; The collected data is deduplicated and filtered according to the calculation result of the cosine similarity.

3. The method according to claim 2, characterized in that The preprocessing of the collected data also includes: Find the intersection between numeric vectors; The data in the intersection are labeled, and the labeled data are used as the model training data.

4. The method according to claim 3, characterized in that The selecting the optimal classification model from the multiple trained classification models comprises: Divide the model training data into K subsets, each subset is used as a test set in turn, and the remaining K-1 subsets constitute the training set; Repeat the following process K times: use the training set to train multiple classification models respectively, and use the test set to evaluate the performance parameters of the multiple trained classification models; Get the average value of the performance parameters of each trained classification model; The optimal classification model is selected according to the average value.

5. The method according to claim 4, characterized in that After selecting the optimal classification model from the multiple trained classification models, the method further comprises: If multiple optimal classification models are selected, the multiple optimal classification models are combined into an integrated model; Combining the classification and recognition results of the multiple optimal classification models by using an ensemble learning method; Performing parameter adjustment processing on the integrated model to obtain a processed integrated model; The performance of the processed integrated model is tested using the validation data set.

6. The method according to any one of claims 1 to 5, characterized in that After obtaining the classification result of the to-be-classified file, the method further includes: The trained classification model is updated using the classification result of the file to be classified.

7. An intelligent file classification device based on multi-model fusion, characterized in that: include: A preprocessing unit, used to preprocess the collected data and use the preprocessed data as model training data; A training unit, used to train multiple classification models using the model training data to obtain multiple trained classification models; A selection unit, configured to select an optimal classification model from the plurality of trained classification models; The classification unit is used to classify and identify the files to be classified using the optimal classification model to obtain the classification results of the files to be classified.

8. The device according to claim 7, characterized in that The pre-processing unit comprises: A determination module, used to determine the type of the collected data, wherein the type includes: text, image; A first extraction module, used for extracting text content from data of text type and converting the text content into a numerical vector; A second extraction module is used to extract image features from the image type data and convert the image features into a numerical vector; A calculation module, used for calculating the cosine similarity between numerical vectors; The deduplication module is used to perform deduplication and filtering processing on the collected data according to the calculation result of the cosine similarity.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Citation Information

Cited By

  • An intelligent classification and retrieval system for government files

    CN120407796B