SMS monitoring and early warning methods, devices, equipment, media, and program products

Abstract

This invention provides a method, apparatus, device, medium, and program product for SMS monitoring and early warning. The method includes: periodically acquiring new complaint SMS samples; fine-tuning a large language model based on the new complaint SMS samples using an efficient parameter fine-tuning method to obtain a fine-tuned large model; acquiring SMS content service requests; and issuing SMS early warnings based on the SMS content service requests using the fine-tuned large model. The SMS monitoring and early warning method provided by this invention, by periodically acquiring new complaint SMS samples and fine-tuning a large language model using an efficient parameter fine-tuning method, enables the large language model to periodically upgrade its ability to respond to SMS monitoring and early warnings. Upon receiving an SMS content service request, it can issue real-time SMS early warnings based on the fine-tuned large model, efficiently identifying potential business compliance issues, realizing an intelligent real-time SMS monitoring and early warning process, reducing manual judgment time, and improving the overall efficiency of SMS monitoring and early warning.

Description

Technical Field

[0001] This invention relates to the field of information processing technology, and in particular to a method, apparatus, equipment, medium, and program product for SMS monitoring and early warning. Background Technology

[0002] In recent years, the issue of telecom operators sending commercial text messages through service numbers has become increasingly prominent, leading to higher requirements for the identification, monitoring, and early warning of commercial text message content types in the cybersecurity field. Against this backdrop, identifying text message types, configuring monitoring and early warning systems for text message content, and proactively reporting non-compliant text message content have gradually become primary issues to be addressed.

[0003] In existing technologies, SMS type identification and monitoring primarily employ keyword matching and blacklist / whitelist filtering techniques to identify SMS types and intercept SMS transmission. Keyword matching methods include precisely filtering SMS content or identifying SMS types through feature vector similarity comparison. Blacklist / whitelist filtering technology mainly achieves precise filtering by accurately matching the sending request user with a list.

[0004] However, relying on keyword matching and blacklist / whitelist filtering technologies for SMS type identification is somewhat unreliable and prone to misjudgments and omissions. It also requires significant human intervention, and the accuracy of the matching heavily depends on manually maintained word databases. Therefore, existing SMS monitoring and early warning technologies are not very efficient. Summary of the Invention

[0005] This invention provides a method, apparatus, device, medium, and program product for SMS monitoring and early warning, which solves the problem of low efficiency in existing SMS monitoring and early warning methods.

[0006] In a first aspect, the present invention provides a method for monitoring and issuing early warnings via SMS, comprising:

[0007] Regularly retrieve newly added complaint SMS samples;

[0008] Based on the newly added complaint SMS samples, the large language model is fine-tuned using an efficient parameter fine-tuning method to obtain the fine-tuned large model.

[0009] Request to retrieve SMS content service;

[0010] Based on the SMS content service request, SMS alerts are generated using the finely tuned large model.

[0011] In one embodiment, the step of fine-tuning the large language model based on the newly added complaint SMS samples using a parameter-efficient fine-tuning method to obtain the fine-tuned large model includes:

[0012] The newly added complaint SMS samples are preprocessed to obtain target complaint SMS samples;

[0013] The target complaint SMS sample is labeled to obtain the tags of the target complaint SMS sample; the tags include complaint SMS tags and SMS category tags;

[0014] Based on the target complaint SMS samples and their tags, the fine-tuning parameters of the large language model are adjusted;

[0015] Based on the fine-tuning parameters, the large language model is fine-tuned to obtain the fine-tuned large model.

[0016] In one embodiment, the step of providing SMS alerts based on the SMS content service request using the fine-tuned large model includes:

[0017] The call intent of the SMS content service request is parsed to obtain the SMS service specification monitoring and early warning intent;

[0018] Based on the SMS service specification monitoring and early warning intent, the first target SMS content is parsed from the SMS content service request;

[0019] The content of the first target SMS message is input into the fine-tuned large model to obtain the SMS service specification result output by the fine-tuned large model; the SMS service specification result includes complaint result and SMS category result;

[0020] If the SMS service specification result does not conform to the SMS service specification, an SMS warning will be issued based on the content of the first target SMS message.

[0021] In one embodiment, the step of providing SMS alerts based on the content of the first target SMS message includes:

[0022] Generate early warning information and optimization suggestions for the content of the first target SMS message;

[0023] The warning information and optimization suggestions are sent to the business platform that sent the SMS content service request.

[0024] In one embodiment, after obtaining the SMS content service request, the method further includes:

[0025] The call intent of the SMS content service request is parsed to obtain the SMS specification solution acquisition intent;

[0026] Based on the intent obtained from the SMS specification solution, the second target SMS content and the SMS service specification issues of the second target SMS content are parsed from the SMS content service request; the SMS service specification issues include SMS specification early warning rules and common SMS specification issues;

[0027] Vector extraction is performed on the SMS service specification issue to obtain the query vector;

[0028] Based on the second target SMS content and the query vector, obtain the SMS standardization solution for the second target SMS content.

[0029] In one embodiment, the SMS specification solution for obtaining the second target SMS content based on the second target SMS content and the query vector includes:

[0030] Based on the query vector, a query is performed in the vector database to obtain knowledge data for resolving the SMS service specification problem; the vector database is used to store response knowledge for different SMS service specification problems.

[0031] The knowledge data is encapsulated into prompt words for generating standardized SMS solutions;

[0032] The prompt word and the second target SMS content are input into the fine-tuned large model to obtain the SMS standard solution for the second target SMS content output by the fine-tuned large model.

[0033] In one embodiment, the large language model is pre-trained in the following manner:

[0034] Based on Flink CDC's non-intrusive data collection capabilities, we obtained the first historical SMS sample that was complained about, the second historical SMS samples of different categories, the third historical SMS sample with SMS standard warning rules, and the fourth historical SMS sample with common SMS standard problems.

[0035] The first historical SMS sample, the second historical SMS sample, the third historical SMS sample, and the fourth historical SMS sample are labeled to obtain the complaint label of the first historical SMS sample, the category label of the second historical SMS sample, the warning rule solution label of the third historical SMS sample, and the common problem solution label of the fourth historical SMS sample.

[0036] The large language model is obtained by training the model based on the first historical SMS sample and its complaint label, the second historical SMS sample and its category label, the third historical SMS sample and its early warning rule solution label, and the fourth historical SMS sample and its common problem solution label.

[0037] Secondly, the present invention also provides a text message monitoring and early warning device, comprising:

[0038] The first acquisition module is used to periodically acquire newly added complaint SMS samples;

[0039] The model fine-tuning module is used to fine-tune the large language model based on the newly added complaint SMS samples using an efficient parameter fine-tuning method, so as to obtain the fine-tuned large model.

[0040] The second acquisition module is used to acquire SMS content service requests;

[0041] The SMS alert module is used to issue SMS alerts based on the SMS content service request and through the fine-tuned large model.

[0042] Thirdly, the present invention provides a device comprising an electronic device, the electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of any of the above-described SMS monitoring and early warning methods.

[0043] Fourthly, the present invention also provides a medium comprising a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the above-described SMS monitoring and early warning methods.

[0044] Fifthly, the present invention also provides a product comprising a computer program product, the computer program product comprising a computer program, the computer program being able to be stored on a non-transitory computer-readable storage medium, and the computer program, when executed by the processor, implementing the steps of any of the above-described SMS monitoring and early warning methods.

[0045] The SMS monitoring and early warning method, device, equipment, medium, and program products provided by this invention acquire new complaint SMS samples at regular intervals and fine-tune a large language model using an efficient parameter fine-tuning method. This allows the large language model to be upgraded regularly to cope with SMS monitoring and early warning, and the fine-tuning efficiency is high. Therefore, upon receiving an SMS content service request, it can perform real-time SMS early warning based on the fine-tuned large model, efficiently identify potential business compliance issues, realize an intelligent real-time SMS monitoring and early warning process, reduce the time for manual judgment, and improve the overall efficiency of SMS monitoring and early warning. At the same time, it comprehensively enhances the intelligent configuration and early warning capabilities for business SMS and the source governance capabilities of SMS content. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0047] Figure 1This is a flowchart illustrating the SMS monitoring and early warning method provided by the present invention.

[0048] Figure 2 This is a technical framework diagram of the AI ​​large model technology provided by the present invention.

[0049] Figure 3 This is a schematic diagram of the non-intrusive data acquisition method of Flink CDC provided by the present invention.

[0050] Figure 4 This is a flowchart of the model fine-tuning process provided by the present invention.

[0051] Figure 5 This is a flowchart of the real-time inference process for large models provided by this invention.

[0052] Figure 6 This is a flowchart of the overall service process for SMS monitoring and early warning provided by the present invention.

[0053] Figure 7 This is a flowchart of the overall service process for the solution to obtain SMS specifications provided by the present invention.

[0054] Figure 8 This is a flowchart illustrating the vector database knowledge plugin provided by this invention.

[0055] Figure 9 This is a schematic diagram of the SMS monitoring and early warning device provided by the present invention.

[0056] Figure 10 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0057] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0058] The terms "first," "second," etc., used in this invention are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that embodiments of the invention can be implemented in orders other than those illustrated or described herein.

[0059] The following is combined Figures 1-10 This invention describes the SMS monitoring and early warning method, apparatus, equipment, medium, and program products provided by the present invention.

[0060] Combination Figure 1 , Figure 1 This is a flowchart illustrating the SMS monitoring and early warning method provided by the present invention.

[0061] like Figure 1 As shown, the method includes the following:

[0062] Step 101: Periodically obtain samples of newly added complaint text messages;

[0063] Step 102: Based on the newly added complaint SMS samples, the large language model is fine-tuned using an efficient parameter fine-tuning method to obtain the fine-tuned large model;

[0064] Step 103: Obtain SMS content service request;

[0065] Step 104: Based on the SMS content service request, perform SMS alerts using the fine-tuned large model.

[0066] It should be noted that the SMS monitoring and early warning method provided in this embodiment of the invention is implemented based on an SMS monitoring and early warning device. This embodiment of the invention uses an SMS monitoring and early warning device as the executing entity to describe the SMS monitoring and early warning method.

[0067] Typically, after business personnel configure SMS templates or specific SMS content in the business system or platform, when a normal SMS delivery request is triggered, the request message is directly sent to the SMS gateway, which then forwards it to the target user. However, the SMS monitoring and early warning method provided by this invention primarily utilizes large-scale AI models and real-time data stream processing technology. It adds the automatic identification capability of the large-scale AI model before the SMS content service request reaches the SMS gateway, and establishes a real-time interaction mechanism between the business platform and the underlying capability platform. This adds an SMS monitoring and early warning step before the request reaches the SMS gateway, thereby improving the business standardization of SMS services and the source governance capability of SMS content at the source.

[0068] This invention, based on an underlying AI framework and model fine-tuning inference technology, empowers the entire SMS sending process, enabling automated SMS classification, SMS monitoring and early warning, SMS compliance early warning rules, and responses to common SMS compliance questions. First, it performs supervised fine-tuning (SFT) on a large model using existing user historical complaints, known SMS types, and new complaint data to achieve SMS type judgment and monitoring and early warning model construction. Second, it performs knowledge vectorization and semantic storage for SMS compliance early warning rules and common SMS compliance questions, and uses RAG (Retrieval Augmented Generation) for invocation. During the judgment and early warning process, even when management is unfamiliar with the SMS compliance early warning rules, it enables them to quickly obtain response knowledge and resolve common issues, empowering SMS management stakeholders in customer service scenarios. The overall AI technology framework includes four layers, such as... Figure 2 As shown, Figure 2 This is a technical framework diagram of the AI ​​large model technology provided by the present invention. The AI ​​large model technology includes an algorithm base, a storage and computing base, training and push technologies, and external services.

[0069] (1) Algorithm foundation: including technical components related to large models. First, in terms of AI models, including large language models, intent recognition models, and vectorization models. The basic large language model can perform intelligent interaction of general knowledge, but there are erroneous reasoning phenomena in the reasoning decision-making of scenarios such as SMS classification and detection and early warning. The model needs to be fine-tuned to improve the recognition accuracy. The intent recognition model is used to identify the calling intent of external callers. For example, when an external system calls the interface in real time, it needs to be routed to different processing logic according to different intents. The vectorization model assists the search engine, which not only realizes the vectorization calculation and query of knowledge corpus, but also supports the mixed retrieval of semantic query and keyword query, expanding the query coverage and accuracy. Secondly, the framework includes PyTorch, CUDA, and vLLM. PyTorch is the programming framework used in this solution for model training, fine-tuning, and real-time inference. It performs computations based on a dynamic computation graph, enabling customization of AI model architecture, optimizers, loss functions, and other aspects. CUDA is the underlying parallel computing tool and GPU programming model, enabling efficient computation of the AI ​​model on GPUs. vLLM is a large-model computation acceleration framework that effectively manages the attention memory of large models based on the PagedAttention algorithm, improving the inference throughput and memory usage efficiency of large models.

[0070] (2) Storage and computing platform: including offline and real-time storage and computing services. First, for offline data, offline SMS rules and log files are collected through FTP, SFTP, and JDBC, business specification issues are aggregated, and delivered to the AI ​​model for regular model optimization and knowledge storage; second, for real-time data, a Kafka+Flink data collection and computing architecture is adopted to realize real-time aggregation and computing of complaint data, SMS content, SMS type, warning signs, etc. under the dual-center deployment mode of central + edge, and data preprocessing and feature extraction are achieved through data filtering, cleaning, completion, and association, user large model is fine-tuned regularly and real-time inference is performed to shorten data latency; in addition, persistent data storage is achieved through HBase.

[0071] (3) Training and Inference Technology: This includes three modules: corpus management, model fine-tuning, and model inference. The corpus management module not only targets training and fine-tuning data but also stores data that needs to be queried in real time during the inference and prediction process, such as SMS types, business specification warning rules, and common business specification questions, in a vector database using embedding vectorization technology. HNSW indexing is used to improve the retrieval speed of the vector database. The fine-tuning training module is based on components such as Jupyter and WebSSH, encapsulates algorithms such as QLoRA and P-Tuning v2, and trains data on corpora such as complaints and SMS classification to generate SMS judgment and prediction models. The model inference module is implemented based on prompt word engineering + RAG technology. It uses prompt word technologies such as Few Shot and Chain of Thought to enhance the semantic understanding of the large model of business specification warning rules and common business specification questions. It uses RAG technology attached to the vector database to enhance the understanding of SMS content of the large model, enabling the large model to generate inference responses related to SMS rules and common questions using generative technology.

[0072] (4) External services: Using technologies and frameworks such as Java, Python, and JavaScript, the model provides external services via Application Programming Interface (API) or Web interface.

[0073] Therefore, within the technical framework of AI large model technology, after the large language model has completed pre-training, it will periodically acquire new complaint SMS samples to fine-tune the large language model, enabling the model to quickly adjust based on continuously generated sample data and improve prediction accuracy.

[0074] Specifically, based on Flink CDC's non-intrusive data collection technology, newly added complaint SMS samples are acquired periodically.

[0075] It should be noted that, based on Flink CDC's non-intrusive data acquisition technology and employing a highly available streaming computing framework, this system leverages lock-free, non-intrusive concurrent reading, and breakpoint resume capabilities to monitor and capture real-time changes in complaint logs and SMS production databases from the business platform. This includes data insertion (INSERT), update (UPDATE), and deletion (DELETE) operations, which are then written to the message middleware in chronological order. This enables visualization and millisecond-level real-time transmission of multi-source data, such as complaint logs and SMS messages from the business platform, ensuring continuous and real-time updates to the AI ​​large-scale model training data source. (See reference.) Figure 3 , Figure 3 This is a schematic diagram of the non-intrusive data acquisition method of Flink CDC provided by the present invention.

[0076] Furthermore, based on the newly added complaint SMS samples, the large language model is fine-tuned using an efficient parameter fine-tuning method to obtain the fine-tuned large model. This allows for performance improvement through a small number of parameter adjustments without changing all parameters of the original model, significantly reducing computational resources and memory overhead.

[0077] Therefore, after configuring the specific SMS content on the business platform, the business personnel generate an SMS content service request and send it to the SMS monitoring and early warning system. The SMS monitoring and early warning system can then use the latest fine-tuned big model to determine whether the specific SMS content in the SMS content service request complies with the SMS business specifications.

[0078] If the message complies with SMS service specifications, the request to send SMS content service to the gateway is allowed. The gateway will then parse the SMS content and send it to the target user. If the message does not comply with SMS service specifications, a warning message will be generated and sent to the business platform to inform the business personnel that the SMS content poses a risk to compliance with service specifications.

[0079] The SMS monitoring and early warning method provided by this invention acquires new complaint SMS samples at regular intervals and fine-tunes a large language model using an efficient parameter fine-tuning method. This allows the large language model to be upgraded regularly to cope with SMS monitoring and early warning, and the fine-tuning efficiency is high. Therefore, upon receiving an SMS content service request, it can perform real-time SMS early warning based on the fine-tuned large model, efficiently identify potential business compliance issues, realize an intelligent real-time SMS monitoring and early warning process, reduce the time for manual judgment, and improve the overall efficiency of SMS monitoring and early warning. At the same time, it comprehensively enhances the intelligent configuration and early warning capabilities for business SMS and the source governance capabilities of SMS content.

[0080] In one embodiment, based on step 102, the step of fine-tuning the large language model using a parameter-efficient fine-tuning method based on the newly added complaint SMS samples to obtain the fine-tuned large model includes:

[0081] The newly added complaint SMS samples are preprocessed to obtain target complaint SMS samples;

[0082] The target complaint SMS sample is labeled to obtain the tags of the target complaint SMS sample; the tags include complaint SMS tags and SMS category tags;

[0083] Based on the target complaint SMS samples and their tags, the fine-tuning parameters of the large language model are adjusted;

[0084] Based on the fine-tuning parameters, the large language model is fine-tuned to obtain the fine-tuned large model.

[0085] Combination Figure 4 , Figure 4 This is a flowchart of the model fine-tuning process provided by the present invention, and the following content describes the process.

[0086] Specifically, in order to better meet the input needs of the large language model, the newly added complaint samples are preprocessed to obtain the target complaint SMS samples after data preprocessing.

[0087] During data preprocessing, newly added complaint SMS samples need to be cleaned, filtered, and categorized. Simultaneously, missing and outlier values ​​need to be analyzed, and real-time data completion based on historical data averages is performed to ensure data integrity. While common sample processing methods achieve the desired result by deleting missing values, this approach has significant limitations. Therefore, a Lagrange interpolation method is used to complete the data in real-time.

[0088] Given n+1 discrete data pairs (xi, yi), where i = 0, 1, ..., n, the goal of Lagrange interpolation is to construct an nth-order polynomial L(x) such that L(xi) = yi. This polynomial can be expressed as:

[0089]

[0090] Each of them It is a Lagrange basis function, defined as follows:

[0091]

[0092] Calculate each Lagrange basis function based on the given data points. Then substitute The final interpolation polynomial is obtained from the expression.

[0093] Meanwhile, leveraging the cloud-based Flink streaming computing capabilities, a real-time data processing approach combining the Flink Java API and SQL is employed, enabling real-time application development through a drag-and-drop visual interface. A dual-center deployment (central + edge) is adopted to filter, clean, complete, and correlate data in real time, extracting features such as SMS messages, to avoid data errors, reduce latency, and ensure data security. This results in target complaint SMS samples that serve as the input data source for the AI ​​model. The newly added complaint SMS samples include SMS template identifiers, SMS service content, and the SMS sending port, which are used as input for training the model.

[0094] Furthermore, the target complaint SMS samples are labeled to obtain the tags of the target complaint SMS samples. The tags specifically include complaint SMS tags and SMS category tags. The complaint SMS tags are used to reflect that the target complaint SMS samples have been complained about by users, and the SMS category tags are used to reflect the category to which the target complaint SMS samples belong.

[0095] Furthermore, based on the target complaint SMS samples and their labels, the fine-tuning parameters of the large language model can be adjusted, including batch size, gradient accumulation steps, learning rate, and variation strategies.

[0096] Furthermore, the QLoRA method can be initiated to fine-tune the large language model based on the fine-tuning parameters. After fine-tuning, the adapted weights generated by QLoRA are merged with the weights of the base model to obtain the fine-tuned large model. The fine-tuned large model can be saved using an appropriate framework to generate the corresponding model file for easy use and deployment later.

[0097] For model fine-tuning scenarios, efficient fine-tuning (partial parameter fine-tuning) is adopted. First, the model loads the pre-trained parameters. Initialize the parameters, then adjust them by maximizing the probability of the conditional language model. :

[0098]

[0099] Where x represents the input sequence (features) of the large model; y represents the output sequence of the large model; Z represents the input and output space; and t represents the token sequence. It is a multi-task learner based on full fine-tuning, representing the fully fine-tuned parameters. The conditional probability of predicting the output from the input.

[0100] Furthermore, fewer parameters are used to represent the parameter increments to be learned. ,in, Originally searching The optimization objective becomes finding :

[0101]

[0102] Where x represents the input sequence (features) of the large model; y represents the output sequence of the large model; Z represents the input and output space; and t represents the token sequence. It is a multi-task learner based on efficient fine-tuning of mini-batch parameters, representing the initialization parameters. Add small batch fine-tuning parameters The conditional probability of predicting the output from the input.

[0103] This invention, through periodic acquisition of new complaint SMS samples and the use of efficient parameter fine-tuning to fine-tune the large language model, significantly improves the model's understanding and response accuracy to complaint SMS. This allows the large language model to be upgraded periodically to address SMS monitoring and early warning capabilities, and the fine-tuning efficiency is high. Therefore, upon receiving an SMS content service request, it can perform real-time SMS early warning based on the fine-tuned large model, efficiently identify potential business compliance issues, and realize an intelligent real-time SMS monitoring and early warning process.

[0104] In some embodiments, based on step 104, the step of performing SMS alerts based on the SMS content service request using the fine-tuned large model includes:

[0105] The call intent of the SMS content service request is parsed to obtain the SMS service specification monitoring and early warning intent;

[0106] Based on the SMS service specification monitoring and early warning intent, the first target SMS content is parsed from the SMS content service request;

[0107] The content of the first target SMS message is input into the fine-tuned large model to obtain the SMS service specification result output by the fine-tuned large model; the SMS service specification result includes complaint result and SMS category result;

[0108] If the SMS service specification result does not conform to the SMS service specification, an SMS warning will be issued based on the content of the first target SMS message.

[0109] It should be noted that, in combination Figure 5 , Figure 5 This is the flowchart of real-time inference for large models provided by the present invention, consisting of... Figure 5 It can be seen that the finely tuned large language model can be used for real-time inference. After obtaining the SMS content service request in real time, the intent of the SMS content service request is first determined. When it is confirmed that the intent of the request is to monitor SMS service specifications and implement SMS monitoring and early warning, the specific SMS content is obtained, and the large model is called to complete real-time inference and generate results.

[0110] Combination Figure 6 , Figure 6 This is a flowchart of the overall service process for SMS monitoring and early warning provided by the present invention, and the process is described below.

[0111] Specifically, the intent of the SMS content service request is parsed. This can be achieved by using an input intent recognition model and AI intelligent analysis to obtain the intent in the SMS content service request.

[0112] If the intent in the SMS content service request is an intent for monitoring and warning according to the SMS business specifications, then the first target SMS content is parsed from the SMS content service request according to the intent for monitoring and warning according to the SMS business specifications.

[0113] Furthermore, the content of the first target SMS message is input into the fine-tuned large model to obtain the SMS service specification result output by the fine-tuned large model, which includes the complaint result and the SMS category result.

[0114] If the SMS service specification result does not comply with the SMS service specification, it means that the content of the first target SMS in this SMS category is not in compliance with the SMS service specification, which is very likely to be complained by users. In this case, an SMS warning will be issued based on the content of the first target SMS.

[0115] This invention selects the SMS service specification monitoring and early warning processing logic based on intent analysis results, further parses the first target SMS content in the SMS content service request, and uses a fine-tuned large language model for SMS early warning. This effectively realizes the automatic monitoring and early warning of potential SMS problems, improves the intelligence level of SMS processing, and enhances response efficiency.

[0116] Based on the above description, the step of providing SMS alerts based on the content of the first target SMS message includes:

[0117] Generate early warning information and optimization suggestions for the content of the first target SMS message;

[0118] The warning information and optimization suggestions are sent to the business platform that sent the SMS content service request.

[0119] Specifically, after confirming that the SMS service specification results do not conform to the SMS service specification, an early warning message for the first target SMS content is generated. At the same time, optimization suggestions for the first target SMS content are also generated. These suggestions can be input into a finely tuned large model and intelligently generated by AI.

[0120] Furthermore, the service platform that sent the SMS content service request sends an early warning message and optimization suggestions for the first target SMS content. Therefore, after receiving the early warning message on the service platform, the business personnel first cancel the SMS sending request, and then modify the SMS content according to the obtained optimization suggestions.

[0121] In this embodiment of the invention, when a text message is found to be non-compliant with SMS service specifications, a warning message and corresponding optimization suggestions are generated for the message content. The warning message and optimization suggestions are then sent to the business platform to ensure that business personnel can receive timely feedback and take necessary adjustments. This not only improves the monitoring efficiency of potentially problematic text messages but also provides data support for optimizing business processes, promoting smooth overall operations and improving customer satisfaction.

[0122] In some embodiments, after obtaining the SMS content service request, the method further includes:

[0123] The call intent of the SMS content service request is parsed to obtain the SMS specification solution acquisition intent;

[0124] Based on the intent obtained from the SMS specification solution, the second target SMS content and the SMS service specification issues of the second target SMS content are parsed from the SMS content service request; the SMS service specification issues include SMS specification early warning rules and common SMS specification issues;

[0125] Vector extraction is performed on the SMS service specification issue to obtain the query vector;

[0126] Based on the second target SMS content and the query vector, obtain the SMS standardization solution for the second target SMS content.

[0127] It should be noted that, by Figure 5 It can be seen that after obtaining the SMS content service request in real time, the intent of the SMS content service request is first determined. When it is confirmed that the intent of the request is to obtain the SMS business specification warning or the solution to common problems, the specific SMS content and the SMS business specification problem are obtained. Based on the response knowledge of semantic retrieval, keyword retrieval and matching vector database, the knowledge is encapsulated into prompt words. Real-time reasoning is completed by calling the large model to generate the result.

[0128] Combination Figure 7 , Figure 7 This is a flowchart of the overall service process for the SMS specification acquisition solution provided by the present invention, and the process is described below.

[0129] Specifically, the intent of the SMS content service request is parsed. This can be achieved by using an input intent recognition model and AI intelligent analysis to obtain the intent in the SMS content service request.

[0130] If the intent in the SMS content service request is the intent to obtain SMS specification solutions, then based on the intent of SMS business specification monitoring and early warning, the second target SMS content and the SMS business specification issues of the second target SMS content are parsed from the SMS content service request. The SMS business specification issues include SMS specification early warning rules and common SMS specification issues.

[0131] Vector extraction is performed on SMS service specification issues to obtain the corresponding query vector.

[0132] Furthermore, based on the second target SMS content and the query vector, a standardized SMS solution for obtaining the second target SMS content is developed.

[0133] This invention employs intent analysis to select a solution for SMS compliance, further analyzing the second target SMS content and SMS service compliance issues within the SMS content service request. Vector extraction is performed on the data, enabling data to be queried in a vector database. This allows for the retrieval of knowledge data that can resolve SMS service compliance issues. Combined with the second target SMS content, a corresponding SMS compliance solution is generated, providing business personnel with a timely and effective solution and improving SMS service compliance.

[0134] Based on the above description, the SMS specification solution for obtaining the second target SMS content based on the second target SMS content and the query vector includes:

[0135] Based on the query vector, a query is performed in the vector database to obtain knowledge data for resolving the SMS service specification problem; the vector database is used to store response knowledge for different SMS service specification problems.

[0136] The knowledge data is encapsulated into prompt words for generating standardized SMS solutions;

[0137] The prompt word and the second target SMS content are input into the fine-tuned large model to obtain the SMS standard solution for the second target SMS content output by the fine-tuned large model.

[0138] Specifically, based on the query vector, a query is performed in the vector database to obtain knowledge data for resolving SMS service standardization issues.

[0139] It should be noted that the vector database stores response knowledge for different SMS service specification issues and connects to the large model in a knowledge-plug-in mode. When the model performs inference, it triggers a RAG call to retrieve database knowledge, which is then provided to the large model as supplementary knowledge to assist in generating SMS specification solutions. Figure 8 , Figure 8This is a flowchart illustrating the vector database knowledge plugin provided by the present invention. Specifically, it involves: acquiring SMS standard warning rule samples and SMS standard common question samples; splitting these samples into Block data segments; generating corresponding responses for each Block data segment based on the inductive summarization capability of the large model, thereby forming multiple QA question-answer pairs; furthermore, using a vectorization model to vectorize the data of these QA question-answer pairs, and simultaneously constructing an HNSW index to input the data into the vector database.

[0140] Furthermore, the acquired knowledge data is encapsulated into prompt words for generating standardized SMS solutions.

[0141] Furthermore, the prompt words and the second target SMS content are input into the fine-tuned large model. The prompt words guide the fine-tuned large model to output the SMS specification solution for the second target SMS content, and the SMS specification solution is sent to the business platform so that business personnel can modify the SMS content according to the obtained SMS specification solution.

[0142] This invention utilizes a generated query vector to perform precise queries in a vector database, retrieving relevant knowledge data on SMS service standardization issues. The acquired knowledge data is then encapsulated into prompt words. Based on these prompt words and the content of a second target SMS message, a finely tuned large model is used for intelligent analysis, outputting corresponding SMS standardization solutions. This provides business personnel with timely and effective solutions, improving the standardization of SMS services.

[0143] In some embodiments, the large language model is pre-trained in the following manner:

[0144] Based on Flink CDC's non-intrusive data collection capabilities, we obtained the first historical SMS sample that was complained about, the second historical SMS samples of different categories, the third historical SMS sample with SMS standard warning rules, and the fourth historical SMS sample with common SMS standard problems.

[0145] The first historical SMS sample, the second historical SMS sample, the third historical SMS sample, and the fourth historical SMS sample are labeled to obtain the complaint label of the first historical SMS sample, the category label of the second historical SMS sample, the warning rule solution label of the third historical SMS sample, and the common problem solution label of the fourth historical SMS sample.

[0146] The large language model is obtained by training the model based on the first historical SMS sample and its complaint label, the second historical SMS sample and its category label, the third historical SMS sample and its early warning rule solution label, and the fourth historical SMS sample and its common problem solution label.

[0147] Specifically, based on Flink CDC's non-intrusive data collection capabilities, the first historical SMS sample that was complained about, the second historical SMS samples of different categories, the third historical SMS samples with SMS standard warning rules, and the fourth historical SMS samples with common SMS standard problems are obtained. The obtained sample data is preprocessed in the same way as above, and will not be described in detail here.

[0148] For example, being complained about means that the content of a historical email was found to be non-compliant and was subsequently reported by the user. Different categories can include marketing, service, business reminders, payment reminders, public service notices, internal task reminders, etc. SMS compliance warning rules can include service SMS messages containing marketing content, marketing content with embedded links, marketing content without real-name signatures, marketing content without clearly stating the SMS unsubscribe method, SMS content that is too long, etc. Common SMS compliance issues include "Please confirm the SMS type of this SMS template", "Please note that service SMS messages should not contain marketing content", "Please confirm whether there are any embedded links", "Do you want to add a real-name signature?", "Do you want to add the SMS unsubscribe method normally?", "Please note the character limit of the SMS content", etc.

[0149] The first and second historical SMS samples were manually labeled to obtain complaint tags for the first historical SMS sample and category tags for the second historical SMS sample.

[0150] Meanwhile, the third and fourth historical SMS samples were labeled with large model prompt words to obtain the warning rule solution tags for the third historical SMS sample and the common problem solution tags for the fourth historical SMS sample.

[0151] Furthermore, the model is trained based on the first historical SMS sample and its complaint label, the second historical SMS sample and its category label, the third historical SMS sample and its early warning rule solution label, and the fourth historical SMS sample and its common problem solution label to obtain a pre-trained large language model.

[0152] This invention utilizes Flink CDC technology to acquire diverse SMS samples, including first historical SMS samples that have been complained about, second historical SMS samples of different categories, third historical SMS samples containing SMS standard warning rules, and fourth historical SMS samples involving common problems. After accurately labeling these samples, complaint tags, category tags, warning rule solution tags, and common problem solution tags are generated, providing rich and structured input data for model training. This ensures the accuracy and comprehensiveness of the model in understanding and processing SMS standard-related issues. Therefore, the pre-trained large language model has the ability to effectively identify and standardize SMS content, significantly improving the intelligence level and response quality of SMS services.

[0153] In summary, by adding SMS monitoring and early warning mechanisms and Q&A sessions on SMS service specifications to the standard SMS delivery process, we can comprehensively improve the intelligent configuration and early warning capabilities for SMS services and the source governance capabilities for SMS content. The advantages are as follows:

[0154] 1. Intelligent Business Configuration: Based on the intent recognition model, the system determines the user's calling intent, performs inference based on the finely tuned large model, and generates the final SMS service specification recognition result and the inference response to SMS service specification questions.

[0155] 2. Convenient staffing: Through AI big data model for intelligent identification and classification feedback, the system can intelligently identify SMS types and complaint situations based on a continuously updated big data language model, and provide early warnings and optimization suggestions based on the results to improve the standardization of SMS services.

[0156] 3. Improve service efficiency: Through real-time data stream processing technology, real-time data collection and service efficiency can be improved throughout the entire process from data acquisition to data service.

[0157] 4. Reduce labor costs: Real-time intelligent SMS monitoring and early warning reduces the time for manual judgment and knowledge base queries, improving business configuration efficiency while reducing manual maintenance costs compared to existing technologies such as keywords and blacklists.

[0158] The SMS monitoring and early warning device provided by the present invention is described below. The SMS monitoring and early warning device described below can be referred to in correspondence with the SMS monitoring and early warning method described above.

[0159] Reference Figure 9 , Figure 9 This is a schematic diagram of the SMS monitoring and early warning device provided by the present invention.

[0160] The SMS monitoring and early warning device includes:

[0161] The first acquisition module 910 is used to periodically acquire newly added complaint SMS samples.

[0162] The model fine-tuning module 920 is used to fine-tune the large language model based on the newly added complaint SMS samples using an efficient parameter fine-tuning method, so as to obtain the fine-tuned large model.

[0163] The second acquisition module 930 is used to acquire SMS content service requests.

[0164] The SMS alert module 940 is used to issue SMS alerts based on the SMS content service request and through the fine-tuned large model.

[0165] The SMS monitoring and early warning device provided by this invention acquires new complaint SMS samples at regular intervals and fine-tunes a large language model using an efficient parameter fine-tuning method. This allows the large language model to be upgraded periodically to cope with SMS monitoring and early warning, and the fine-tuning efficiency is high. Therefore, upon receiving an SMS content service request, it can perform real-time SMS early warning based on the fine-tuned large model, efficiently identify potential business compliance issues, realize an intelligent real-time SMS monitoring and early warning process, reduce the time for manual judgment, and improve the overall efficiency of SMS monitoring and early warning. At the same time, it comprehensively enhances the intelligent configuration and early warning capabilities for business SMS and the source governance capabilities of SMS content.

[0166] Furthermore, the model fine-tuning module 920 is also used for:

[0167] The newly added complaint SMS samples are preprocessed to obtain target complaint SMS samples;

[0168] The target complaint SMS sample is labeled to obtain the tags of the target complaint SMS sample; the tags include complaint SMS tags and SMS category tags;

[0169] Based on the target complaint SMS samples and their tags, the fine-tuning parameters of the large language model are adjusted;

[0170] Based on the fine-tuning parameters, the large language model is fine-tuned to obtain the fine-tuned large model.

[0171] Furthermore, the SMS alert module 940 is also used for:

[0172] The call intent of the SMS content service request is parsed to obtain the SMS service specification monitoring and early warning intent;

[0173] Based on the SMS service specification monitoring and early warning intent, the first target SMS content is parsed from the SMS content service request;

[0174] The content of the first target SMS message is input into the fine-tuned large model to obtain the SMS service specification result output by the fine-tuned large model; the SMS service specification result includes complaint result and SMS category result;

[0175] If the SMS service specification result does not conform to the SMS service specification, an SMS warning will be issued based on the content of the first target SMS message.

[0176] Furthermore, the SMS alert module 940 is also used for:

[0177] Generate early warning information and optimization suggestions for the content of the first target SMS message;

[0178] The warning information and optimization suggestions are sent to the business platform that sent the SMS content service request.

[0179] Furthermore, the SMS monitoring and early warning device is also used for:

[0180] The call intent of the SMS content service request is parsed to obtain the SMS specification solution acquisition intent;

[0181] Based on the intent obtained from the SMS specification solution, the second target SMS content and the SMS service specification issues of the second target SMS content are parsed from the SMS content service request; the SMS service specification issues include SMS specification early warning rules and common SMS specification issues;

[0182] Vector extraction is performed on the SMS service specification issue to obtain the query vector;

[0183] Based on the second target SMS content and the query vector, obtain the SMS standardization solution for the second target SMS content.

[0184] Furthermore, the SMS monitoring and early warning device is also used for:

[0185] Based on the query vector, a query is performed in the vector database to obtain knowledge data for resolving the SMS service specification problem; the vector database is used to store response knowledge for different SMS service specification problems.

[0186] The knowledge data is encapsulated into prompt words for generating standardized SMS solutions;

[0187] The prompt word and the second target SMS content are input into the fine-tuned large model to obtain the SMS standard solution for the second target SMS content output by the fine-tuned large model.

[0188] Furthermore, the SMS monitoring and early warning device is also used for:

[0189] Based on Flink CDC's non-intrusive data collection capabilities, we obtained the first historical SMS sample that was complained about, the second historical SMS samples of different categories, the third historical SMS sample with SMS standard warning rules, and the fourth historical SMS sample with common SMS standard problems.

[0190] The first historical SMS sample, the second historical SMS sample, the third historical SMS sample, and the fourth historical SMS sample are labeled to obtain the complaint label of the first historical SMS sample, the category label of the second historical SMS sample, the warning rule solution label of the third historical SMS sample, and the common problem solution label of the fourth historical SMS sample.

[0191] The large language model is obtained by training the model based on the first historical SMS sample and its complaint label, the second historical SMS sample and its category label, the third historical SMS sample and its early warning rule solution label, and the fourth historical SMS sample and its common problem solution label.

[0192] It should be noted that the SMS monitoring and early warning device provided by the present invention can execute the SMS monitoring and early warning method described in any of the above embodiments during specific operation, which will not be elaborated in this embodiment.

[0193] Figure 10 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 10 As shown, the electronic device may include a processor 1010, a communications interface 1020, a memory 1030, and a communication bus 1040. The processor 1010, communications interface 1020, and memory 1030 communicate with each other via the communication bus 1040. The processor 1010 can call logical instructions in the memory 1030 to execute a text message monitoring and early warning method. This method includes: periodically acquiring new complaint text message samples; fine-tuning a large language model based on the new complaint text message samples using a parameter-efficient fine-tuning method to obtain a fine-tuned large model; acquiring text message content service requests; and issuing text message early warnings based on the text message content service requests using the fine-tuned large model.

[0194] Furthermore, the logical instructions in the aforementioned memory 1030 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0195] On the other hand, the present invention also provides a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, and when the program instructions are executed by a computer, the computer can execute the SMS monitoring and early warning method provided in the above embodiments, the method including: periodically acquiring new complaint SMS samples; based on the new complaint SMS samples, fine-tuning a large language model using a parameter efficient fine-tuning method to obtain a fine-tuned large model; acquiring SMS content service requests; and based on the SMS content service requests, performing SMS early warning through the fine-tuned large model.

[0196] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the SMS monitoring and early warning method provided in the above embodiments. The method includes: periodically acquiring new complaint SMS samples; fine-tuning a large language model based on the new complaint SMS samples using a parameter-efficient fine-tuning method to obtain a fine-tuned large model; acquiring SMS content service requests; and performing SMS early warning based on the SMS content service requests using the fine-tuned large model.

[0197] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0198] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0199] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for monitoring and issuing early warnings via SMS, characterized in that, include: Regularly retrieve newly added complaint SMS samples; Based on the newly added complaint SMS samples, the large language model is fine-tuned using an efficient parameter fine-tuning method to obtain the fine-tuned large model. Request to retrieve SMS content service; Based on the SMS content service request, SMS alerts are generated using the finely tuned large model. The large language model is pre-trained in the following way: Based on Flink CDC's non-intrusive data collection capabilities, we obtained the first historical SMS sample that was complained about, the second historical SMS samples of different categories, the third historical SMS sample with SMS standard warning rules, and the fourth historical SMS sample with common SMS standard problems. The first historical SMS sample, the second historical SMS sample, the third historical SMS sample, and the fourth historical SMS sample are labeled to obtain the complaint label of the first historical SMS sample, the category label of the second historical SMS sample, the warning rule solution label of the third historical SMS sample, and the common problem solution label of the fourth historical SMS sample. The large language model is obtained by training the model based on the first historical SMS sample and its complaint tag, the second historical SMS sample and its category tag, the third historical SMS sample and its early warning rule solution tag, and the fourth historical SMS sample and its common problem solution tag. The step of providing SMS alerts based on the SMS content service request and using the fine-tuned large model includes: The call intent of the SMS content service request is parsed to obtain the SMS service specification monitoring and early warning intent; Based on the SMS service specification monitoring and early warning intent, the first target SMS content is parsed from the SMS content service request; The content of the first target SMS message is input into the fine-tuned large model to obtain the SMS service specification result output by the fine-tuned large model; the SMS service specification result includes complaint result and SMS category result; If the SMS service specification result does not conform to the SMS service specification, an SMS warning will be issued based on the content of the first target SMS message.

2. The SMS monitoring and early warning method according to claim 1, characterized in that, Based on the newly added complaint SMS samples, the large language model is fine-tuned using an efficient parameter fine-tuning method to obtain the fine-tuned large model, including: The newly added complaint SMS samples are preprocessed to obtain target complaint SMS samples; The target complaint SMS sample is labeled to obtain the tags of the target complaint SMS sample; the tags include complaint SMS tags and SMS category tags; Based on the target complaint SMS samples and their tags, the fine-tuning parameters of the large language model are adjusted; Based on the fine-tuning parameters, the large language model is fine-tuned to obtain the fine-tuned large model.

3. The SMS monitoring and early warning method according to claim 1, characterized in that, The method of issuing SMS alerts based on the content of the first target SMS message includes: Generate early warning information and optimization suggestions for the content of the first target SMS message; The warning information and optimization suggestions are sent to the business platform that sent the SMS content service request.

4. The SMS monitoring and early warning method according to claim 1, characterized in that, After obtaining the SMS content service request, it also includes: The call intent of the SMS content service request is parsed to obtain the SMS specification solution acquisition intent; Based on the intent obtained from the SMS specification solution, the second target SMS content and the SMS service specification issues of the second target SMS content are parsed from the SMS content service request; the SMS service specification issues include SMS specification early warning rules and common SMS specification issues; Vector extraction is performed on the SMS service specification issue to obtain the query vector; Based on the second target SMS content and the query vector, obtain the SMS standardization solution for the second target SMS content.

5. The SMS monitoring and early warning method according to claim 4, characterized in that, The SMS specification solution for obtaining the second target SMS content based on the second target SMS content and the query vector includes: Based on the query vector, a query is performed in the vector database to obtain knowledge data for resolving the SMS service specification problem; the vector database is used to store response knowledge for different SMS service specification problems. The knowledge data is encapsulated into prompt words for generating standardized SMS solutions; The prompt word and the second target SMS content are input into the fine-tuned large model to obtain the SMS standard solution for the second target SMS content output by the fine-tuned large model.

6. A text message monitoring and early warning device, characterized in that, include: The first acquisition module is used to periodically acquire newly added complaint SMS samples; The model fine-tuning module is used to fine-tune the large language model based on the newly added complaint SMS samples using an efficient parameter fine-tuning method, so as to obtain the fine-tuned large model. The second acquisition module is used to acquire SMS content service requests; The SMS alert module is used to issue SMS alerts based on the SMS content service request and through the fine-tuned large model. The large language model is pre-trained in the following way: Based on Flink CDC's non-intrusive data collection capabilities, we obtained the first historical SMS sample that was complained about, the second historical SMS samples of different categories, the third historical SMS sample with SMS standard warning rules, and the fourth historical SMS sample with common SMS standard problems. The first historical SMS sample, the second historical SMS sample, the third historical SMS sample, and the fourth historical SMS sample are labeled to obtain the complaint label of the first historical SMS sample, the category label of the second historical SMS sample, the warning rule solution label of the third historical SMS sample, and the common problem solution label of the fourth historical SMS sample. The large language model is obtained by training the model based on the first historical SMS sample and its complaint tag, the second historical SMS sample and its category tag, the third historical SMS sample and its early warning rule solution tag, and the fourth historical SMS sample and its common problem solution tag. The step of providing SMS alerts based on the SMS content service request and using the fine-tuned large model includes: The call intent of the SMS content service request is parsed to obtain the SMS service specification monitoring and early warning intent; Based on the SMS service specification monitoring and early warning intent, the first target SMS content is parsed from the SMS content service request; The content of the first target SMS message is input into the fine-tuned large model to obtain the SMS service specification result output by the fine-tuned large model; the SMS service specification result includes complaint result and SMS category result; If the SMS service specification result does not conform to the SMS service specification, an SMS warning will be issued based on the content of the first target SMS message.

7. An electronic 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, it implements the steps of the SMS monitoring and early warning method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, comprising a non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the SMS monitoring and early warning method as described in any one of claims 1 to 5.

9. A computer program product, the computer program product comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the steps of the SMS monitoring and early warning method as described in any one of claims 1 to 5.

Images