Short message classification method and system

Abstract

The invention discloses a short message classification method and system, and relates to the technical field of artificial intelligence, and the method comprises the steps: carrying out the feature extraction of a short message text, and obtaining a multi-dimensional feature; and inputting the multi-dimensional features into a pre-trained extreme gradient boosting tree model to obtain a probability that the to-be-predicted short message belongs to a specified type, if the importance probability is greater than a preset threshold value, determining that the to-be-predicted short message is an important short message, and if the importance probability is less than the preset threshold value, determining that the to-be-predicted short message is a non-important short message. The problem that the importance of the short messages cannot be efficiently, accurately and automatically classified in the prior art is solved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method and system for classifying text messages. Background Technology

[0002] With the development of mobile communication technology, the number of text messages received by mobile phone users every day continues to grow. These messages include various types of information such as notifications, verification codes, and advertisements. Quickly identifying important text messages has become a core need for users.

[0003] Currently, the distinction of SMS importance mainly relies on three methods: manual screening, rule engine filtering, and traditional machine learning models. However, all of these methods have significant limitations: manual screening is time-consuming and labor-intensive, and has poor real-time performance, making it difficult to cope with scenarios with a surge in SMS volume; rule engines are based on keyword matching and classification, which has weak generalization ability, and require manual maintenance of rules when facing new SMS content, resulting in serious lag in updates; traditional machine learning models such as logistic regression and support vector machines rely on manual feature engineering, have weak feature representation capabilities, cannot capture the contextual semantics and implicit information of SMS messages, are sensitive to data imbalance, tend to favor the majority class, have poor adaptability to sentence structure changes and synonymous expressions, and show a significant decline in prediction performance across scenarios.

[0004] The aforementioned problems result in low efficiency, insufficient accuracy, and poor adaptability of existing technologies when processing massive amounts of text messages, making it difficult to meet users' needs for rapid identification of important information.

[0005] Therefore, there is an urgent need for a text message classification method that can efficiently and accurately classify the importance of text messages automatically. Summary of the Invention

[0006] In view of this, the present invention proposes a method and system for classifying text messages, which can achieve efficient and accurate automatic classification of the importance of text messages.

[0007] To achieve the above objectives, the present invention provides the following technical solution: A method for classifying text messages includes: Feature extraction is performed on the SMS text to obtain multi-dimensional features; The multi-dimensional features are input into a pre-trained extreme gradient boosting tree model to obtain the probability that the SMS message to be predicted belongs to a specified type. If the importance probability is greater than a preset threshold, it is determined to be an important SMS message; if the importance probability is less than the preset threshold, it is determined to be an unimportant SMS message.

[0008] Based on the above technical solution, the present invention can be further improved as follows: Optionally, before performing feature extraction on the SMS text, the method further includes: The SMS text is cleaned and its characters are standardized.

[0009] Optionally, the feature extraction of the SMS text to obtain multi-dimensional features includes: Character-level features, regular expression features, and statistical features of the SMS text are extracted to obtain multi-dimensional features.

[0010] Optionally, before the step of inputting the multi-dimensional features into the pre-trained extreme gradient boosting tree model, the following steps are included: The optimal hyperparameters of the extreme gradient boosting tree model are searched within a preset parameter range using the Bayesian optimization algorithm. The optimal hyperparameters are used as the final parameters for training the extreme gradient boosting tree model.

[0011] Optionally, before the step of inputting the multi-dimensional features into the pre-trained extreme gradient boosting tree model, the method further includes: The trained extreme gradient boosting tree model, its corresponding vocabulary, and hyperparameter information are compressed into a self-verifying binary file to enable the storage of the extreme gradient boosting tree model and its rapid loading and use in the prediction end.

[0012] Optionally, before the step of inputting the multi-dimensional features into the pre-trained extreme gradient boosting tree model, the method further includes: Misclassified text messages are marked to create corrective samples; When the extreme gradient boosting tree model is updated to the preset time, incremental training is triggered: the server reads the correction sample and uses the correction sample to perform incremental training on the trained extreme gradient boosting tree model to generate a new extreme gradient boosting tree model. The original extreme gradient boosting tree model is replaced with the new extreme gradient boosting tree model to complete the incremental update of the extreme gradient boosting tree model.

[0013] Optionally, inputting the multi-dimensional features into a pre-trained extreme gradient boosting tree model to obtain the probability that the SMS message to be predicted belongs to a specified type includes: The extreme gradient boosting tree model scores the multi-dimensional features through multiple decision units, resulting in several scores. The probability that the SMS message to be predicted belongs to a specified type is obtained by summing several scores and then using a non-linear activation function.

[0014] A system for classifying text messages, comprising: The feature extraction module is used to extract features from SMS text to obtain multi-dimensional features; The importance prediction module is used to input the multi-dimensional features into a pre-trained extreme gradient boosting tree model to obtain the probability that the SMS message to be predicted belongs to a specified type. If the importance probability is greater than a preset threshold, it is determined to be an important SMS message; if the importance probability is less than the preset threshold, it is determined to be an unimportant SMS message.

[0015] An electronic device includes a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor executes the computer program to implement the steps of the method described herein.

[0016] A non-transitory computer-readable storage medium having a computer program stored thereon, the computer program implementing the steps of the method when executed by a processor.

[0017] The present invention has the following advantages: The SMS classification method in this invention replaces manual screening through automated data preprocessing, feature extraction, and model prediction processes, significantly reducing the cost of manual intervention. Even with a surge in SMS volume, it can quickly classify SMS messages and meet real-time identification requirements. Attached Figure Description

[0018] For illustrative and not limiting purposes, the present invention will now be described in conjunction with embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart illustrating the SMS classification method in an embodiment of the present invention. Figure 2 A schematic diagram showing the main components of the SMS classification system in an embodiment of the present invention; Figure 3 This is a schematic diagram of the physical structure of the electronic device provided by the present invention. Detailed Implementation

[0019] To enable those skilled in the art to better understand the present invention, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0020] It should be noted that the terms "first," "second," etc., in the specification and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate for the embodiments of the invention described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0021] It should be noted that, where there is no conflict, the embodiments and features of the present invention can be combined with each other. The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0022] Figure 1 This is a flowchart illustrating the SMS classification method in an embodiment of the present invention, as shown below. Figure 1 As shown, the SMS classification method provided in this embodiment of the invention includes the following steps S101 to S103.

[0023] S101, extract features from the SMS text to obtain multi-dimensional features.

[0024] The text messages to be predicted are subjected to text cleaning and character standardization.

[0025] Text cleaning process: Remove URLs, redundant punctuation, and special characters; Character standardization: unify uppercase and lowercase letters, filter meaningless characters; Original SMS: [Best Neighborhood] Please pick up your package before 8 PM at the second floor of Yimeijia Supermarket next to Gate 4, Area A of Sino-Germany, using pickup code 42539 and item number box-389. For delivery inquiries, please call 18888888888t.cn / RNiw5RK.

[0026] After the URL is removed: [Best Neighborhood] Please pick up your package before 8 PM at the second floor of Yimeijia Supermarket next to Gate 4 in Zone A of Sino-Germany with pick-up code 42539 and item number box-389. For delivery inquiries, please contact 3533.

[0027] Character-level features, regular expression features, and statistical features of the SMS text are extracted to obtain multi-dimensional features.

[0028] Character-level n-gram (which breaks down text into a sequence of n consecutive words) features: extracting 1-gram to 2-gram character combinations.

[0029] For example: [China Construction Bank] repayment → ["CCB", "Bank", "Repay", "Loan", "CCB", "Bank Repay", "Repayment"]; Regular features: matching key semantics based on predefined patterns.

[0030] Example: Key feature patterns: Balance reminder | Reply.* Confirm (?: Accepted | Processed | Ordered) | Processing in progress | Expires at.*; Non-essential feature patterns: Advertisement | Promotion | Offer | Discount; Statistical characteristics: basic statistics such as text length and the proportion of numbers / letters.

[0031] By organically combining character-level semantic features, rule features, and statistical features, the model's ability to understand complex SMS messages is improved.

[0032] Automatically adjust feature weights based on domain keywords to enhance important semantic signals.

[0033] S102, input the multi-dimensional features into the pre-trained extreme gradient boosting tree model to obtain the probability that the SMS to be predicted belongs to the specified type. If the importance probability is greater than the preset threshold, it is determined to be an important SMS. If the importance probability is less than the preset threshold, it is determined to be an unimportant SMS.

[0034] Before the step of inputting the multi-dimensional features into the pre-trained extreme gradient boosting tree model, the following steps are included: The optimal hyperparameters of the extreme gradient boosting tree model are searched within a preset parameter range using the Bayesian optimization algorithm; the optimal classification threshold is searched using the validation set to balance precision and recall.

[0035] The optimal hyperparameters are used as the final parameters for training the extreme gradient boosting tree model.

[0036] The trained extreme gradient boosting tree model (including the vocabulary and hyperparameter information corresponding to the extreme gradient boosting tree model) is compressed into a self-verifying binary file for the training end to save and the prediction end to load in seconds, while ensuring that file corruption can be detected.

[0037] Misclassified text messages are marked to create corrective samples; Specifically, when a mobile phone user marks a misclassified text message as "important" or "unimportant" within the SMS app, this action is considered a "correction sample".

[0038] The SMS-tag pairs are sent to the server database, awaiting further batch processing.

[0039] When the extreme gradient boosting tree model is updated to the preset time, incremental training is triggered: the server reads the correction sample, uses the correction sample to perform incremental training on the trained extreme gradient boosting tree model, and generates a new extreme gradient boosting tree model; the user-corrected data is used for retraining to further improve the accuracy.

[0040] The original extreme gradient boosting tree model is replaced with the new extreme gradient boosting tree model to complete the incremental update of the extreme gradient boosting tree model.

[0041] The extreme gradient boosting tree model scores the multi-dimensional features through multiple decision units, resulting in several scores. The importance probability of the SMS message to be predicted is obtained by summing several scores and then using a nonlinear activation function.

[0042] The SMS classification method in this invention achieves a prediction accuracy of over 90%, significantly outperforming traditional methods. It boasts millisecond-level response time for single SMS messages, meeting real-time requirements. It reduces manual intervention by over 80%, lowering the cost of user information filtering. Furthermore, it exhibits excellent cross-scenario adaptability, maintaining stable performance across various SMS types, including notifications, financial messages, and social messages.

[0043] Figure 2 This is a schematic diagram illustrating the main components of the SMS classification system according to an embodiment of the present invention. (See diagram below.) Figure 2 As shown, the SMS classification system 1 provided in this embodiment of the invention includes a feature extraction module 10 and an importance prediction module 20.

[0044] Feature extraction module 10 is used to extract features from SMS text to obtain multi-dimensional features; The importance prediction module 20 is used to input the multi-dimensional features into a pre-trained extreme gradient boosting tree model to obtain the probability that the SMS to be predicted belongs to a specified type. If the importance probability is greater than a preset threshold, it is determined to be an important SMS; if the importance probability is less than the preset threshold, it is determined to be an unimportant SMS.

[0045] Figure 3 This is a schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention, such as... Figure 3 As shown, the electronic device 30 includes: a processor 301, a memory 302, and a bus 303; The processor 301 and the memory 302 communicate with each other via the bus 303. The processor 301 is used to call program instructions in the memory 302 to execute the methods provided in the above-described method embodiments, and to execute the methods provided in the embodiments of the present invention.

[0046] This embodiment provides a non-transitory computer-readable storage medium that stores computer instructions, which cause a computer to execute the method provided in this embodiment of the invention.

[0047] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the steps of the above method embodiments. The aforementioned storage medium includes various storage media capable of storing program code, such as ROM, RAM, magnetic disk, or optical disk.

[0048] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method of classifying short messages, characterized by, The method comprises the following steps: characteristic extraction is performed on the short message text to obtain multi-dimensional characteristics; the multi-dimensional characteristics are input into a pre-trained extreme gradient boosting tree model to obtain a probability that the short message to be predicted belongs to a specified type, if the importance probability is greater than a preset threshold, the short message is determined as an important short message, and if the importance probability is less than the preset threshold, the short message is determined as a non-important short message.

2. The method of short message classification according to claim 1, characterized in that, Before the characteristic extraction is performed on the short message text, the method further comprises the following steps: the short message text is cleaned and processed and character standardization is performed.

3. The method of short message classification according to claim 1, characterized in that, The characteristic extraction is performed on the short message text to obtain multi-dimensional characteristics, which comprises the following steps: characteristic extraction is performed on the short message text to obtain multi-dimensional characteristics.

4. The method of short message classification according to claim 1, characterized in that, Before the multi-dimensional characteristics are input into the pre-trained extreme gradient boosting tree model, the following steps are further included: a Bayesian optimization algorithm is used to search for optimal hyperparameters of the extreme gradient boosting tree model in a preset parameter range; the optimal hyperparameters are used as final parameters of the extreme gradient boosting tree model to train the extreme gradient boosting tree model.

5. The method of short message classification according to claim 4, characterized in that, Before the multi-dimensional characteristics are input into the pre-trained extreme gradient boosting tree model, the following steps are further included: the trained extreme gradient boosting tree model, a vocabulary corresponding to the extreme gradient boosting tree model, and hyperparameter information are compressed into a self-checking binary file to realize storage of the extreme gradient boosting tree model and rapid loading and use of a prediction end of the extreme gradient boosting tree model.

6. The method of short message classification according to claim 5, characterized in that, Before the multi-dimensional characteristics are input into the pre-trained extreme gradient boosting tree model, the following steps are further included: misclassified short messages are marked to form correction samples; when the extreme gradient boosting tree model is updated for a preset time, incremental training is triggered: the server reads the correction samples, and uses the correction samples to perform incremental training on the trained extreme gradient boosting tree model to generate a new extreme gradient boosting tree model; the new extreme gradient boosting tree model is used to replace the original extreme gradient boosting tree model to complete incremental updating of the extreme gradient boosting tree model.

7. The method of short message classification according to claim 1, characterized in that, The multi-dimensional characteristics are input into the pre-trained extreme gradient boosting tree model to obtain a probability that the short message to be predicted belongs to a specified type, which comprises the following steps: the extreme gradient boosting tree model scores the multi-dimensional characteristics through multiple decision units to obtain a plurality of scores; the plurality of scores are added and then calculated through a nonlinear activation function to obtain the probability that the short message to be predicted belongs to the specified type.

8. A system for classifying short messages, characterized by The method comprises the following steps: a feature extraction module is configured to perform characteristic extraction on short message text to obtain multi-dimensional characteristics; an importance prediction module is configured to input the multi-dimensional characteristics into a pre-trained extreme gradient boosting tree model to obtain a probability that the short message to be predicted belongs to a specified type, if the importance probability is greater than a preset threshold, the short message is determined as an important short message, and if the importance probability is less than the preset threshold, the short message is determined as a non-important short message.

9. An electronic device comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

10. A non-transitory computer readable medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method according to any one of claims 1 to 7.

Images