Interactive personalized process suggestion system and method based on large model

Abstract

The invention discloses an interactive personalized process suggestion system and method based on a large model, and relates to the technical field of workflow optimization, and the system comprises a data collection module which is responsible for monitoring and recording behavior data of a user in real time; the model processing module is responsible for preprocessing the behavior data of the user and analyzing the behavior data through a pre-trained large language model to generate personalized process suggestions conforming to user behaviors; the interactive interface module is responsible for providing a user interface, and the user interface supports various input modes, supports understanding of user input through natural language processing, supports real-time feedback and presents personalized process suggestions; and the feedback adjustment module is responsible for collecting user feedback, classifying and deeply analyzing the user feedback, identifying common problems and improvement points, and generating optimization suggestions. According to the method, the personalized process suggestions can be dynamically generated according to the specific requirements and operation behaviors of the user, the working efficiency is greatly improved, and the user experience is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of workflow optimization, and in particular to an interactive personalized process suggestion system and method based on a large model. Background Art

[0002] 1. Challenges of Workflow Management

[0003] Workflow management is a vital task in modern enterprises and personal work. Workflow involves the allocation, execution, monitoring and optimization of tasks, with the goal of making the work process as efficient and seamless as possible. However, with the increase in task complexity and the expansion of information volume, traditional workflow management systems often face the following challenges:

[0004] Efficiency bottleneck: Complex workflows often lead to efficiency bottlenecks, and users are prone to making mistakes or wasting time when switching between multiple tasks.

[0005] Lack of personalization: Traditional systems often adopt a unified workflow model and fail to fully consider users' personalized needs and preferences, thus affecting operational efficiency and user satisfaction.

[0006] Lack of feedback mechanism: Existing systems usually lack effective feedback mechanisms and cannot dynamically adjust workflows based on user operating habits and feedback.

[0007] 2. Development of personalized technology

[0008] With the development of data analysis and machine learning technology, personalization technology has made significant progress in many fields. Personalized recommendation systems, intelligent assistants, and automated optimization technologies play an important role in providing personalized experiences. These technologies are based on user behavior data, history, and preference settings to provide tailored services and suggestions.

[0009] However, the application of personalization technology in workflow management is still in its infancy. Although there have been some attempts to introduce personalization technology into workflow systems, they usually face the following problems:

[0010] Difficulty of data integration: Personalization systems need to integrate a large amount of user data, which involves data collection, processing and analysis.

[0011] Real-time requirements: In order to provide instant personalized recommendations, the system must have efficient data processing and real-time response capabilities.

[0012] User privacy: User privacy and data security must be ensured during the collection and use of user data.

[0013] 3. Current Personalization Technologies

[0014] Most existing personalized process suggestion systems use static models and cannot adapt to users' changing needs in real time. They usually make personalized process suggestions based on preset rules or fixed models, and cannot make full use of real-time data and dynamic feedback. Such systems often lack in-depth analysis of user behavior, resulting in insufficient accuracy and practicality of suggestions. Summary of the invention

[0015] In view of the needs and shortcomings of current technological development, the present invention provides an interactive personalized process suggestion system and method based on a large model, which automatically generates personalized process suggestions that meet user needs and operating habits by dynamically analyzing user behavior data.

[0016] In a first aspect, the present invention provides an interactive personalized process suggestion system based on a large model, and the technical solution adopted to solve the above technical problems is as follows:

[0017] An interactive personalized process suggestion system based on a large model, which is embedded in a workflow management system, includes:

[0018] The data collection module is responsible for real-time monitoring and recording of user behavior data, including process data generated by user-initialized process templates, iterative process templates, and user requests and queries for process templates, and records the frequency and number of times users use different process templates;

[0019] The model processing module is responsible for preprocessing user behavior data, analyzing user behavior data through a pre-trained large language model, and generating personalized process suggestions that meet user needs and click operations;

[0020] The interactive interface module is responsible for providing a friendly user interface that supports multiple input methods, understands the user's natural language input through natural language processing (NLU), and supports real-time feedback and the presentation of personalized process suggestions.

[0021] The feedback adjustment module is responsible for collecting user feedback, classifying user feedback through a combination of natural language processing technology and manual review, conducting in-depth analysis of the classified user feedback, identifying common problems and improvement points, and generating optimization suggestions for personalized processes.

[0022] Optionally, the model processing module involved preprocesses the user's behavior data, and the preprocessing process includes:

[0023] Data cleaning: remove noise, duplicate data and irrelevant information;

[0024] Data conversion: converting user behavior data into a format that can be processed by large language models;

[0025] Data standardization: standardize the format of user behavior data.

[0026] Further optionally, the model processing module involved analyzes the user's behavior data through a pre-trained large language model to generate personalized process suggestions that meet the user's needs, wherein the pre-training process of the large language model is as follows:

[0027] First, select the large language model GPT as the pre-training model;

[0028] Secondly, based on the user’s historical behavior data and corresponding suggestions, a fine-tuning dataset is constructed;

[0029] Use the fine-tuning dataset to train the large language model GPT. During the training process, set the model hyperparameters.

[0030] Verify and evaluate the performance indicators of the trained large language model GPT, and output the large language model GPT when it meets the preset requirements.

[0031] Further optionally, the model processing module involved analyzes the user's behavior data through a pre-trained large language model to generate personalized process suggestions that meet the user's needs. This process specifically includes:

[0032] Input the preprocessed user behavior data into the large language model GPT;

[0033] The large language model GPT performs reasoning on user data, including: extracting useful features from user behavior data, understanding the user's specific needs and background information, generating and outputting personalized process suggestions that meet user needs and click operations;

[0034] Post-process the personalized process suggestions output by the large language model GPT, including: converting the personalized process suggestions into user-friendly suggestion texts, formatting the suggestion texts, generating and outputting the final suggestions.

[0035] Optionally, the user interface provided by the interactive interface module supports three input modes: text, voice and touch, and supports presenting the personalized process suggestions generated by the model processing module in any mode of text box, chart and graph, and voice output;

[0036] The interactive interface module supports understanding the user's natural language input through natural language processing NLU, identifying the user's natural language input as three types of intentions: request, query and feedback, and extracting key information therefrom. Then, based on the recognition results, the key information extracted when it is identified as feedback intention is sent to the feedback adjustment module, and the key information extracted when it is identified as request or query intention is sent to the data collection module.

[0037] Further optionally, the user provides feedback on the presented personalized process suggestions through the user interface provided by the interactive interface module;

[0038] The feedback adjustment module collects user feedback through the interactive interface module, first classifies and deeply analyzes the user feedback through natural language processing technology, identifies common problems and improvement points, then manually reviews the identified common problems and improvement points, and generates optimization suggestions for user feedback after adjustment;

[0039] The model processing module retrains the large language model GPT using the optimization suggestions to optimize the generation strategy of the large language model GPT.

[0040] In a second aspect, the present invention further provides an interactive method for providing personalized process suggestions to users based on a large model based on the interactive personalized process suggestion system provided in the first aspect. The technical solution adopted to solve the above technical problems is as follows:

[0041] An interactive personalized process suggestion method based on a large model, which is based on the interactive personalized process suggestion system described in the first aspect, includes the following implementation steps:

[0042] S1. The user inputs natural language through the user interface provided by the interactive interface module. The interactive interface module understands the user's natural language input through natural language processing NLU, identifies the user's natural language input as three types of intents: request, query, and feedback, and extracts key information from it. The key information extracted when it is identified as a request or query intent is sent to the data collection module, and the key information extracted when it is identified as a feedback intent is sent to the feedback adjustment module;

[0043] S2, the data collection module monitors and records the user's behavior data in real time, including the process data generated by the user's initialization process template, iterative process template, and the user's request and query process template, records the frequency and number of times the user uses different process templates, and then sends the recorded behavior data to the model processing module;

[0044] S3, the model processing module first pre-processes the user's behavior data, then analyzes the user's behavior data through a pre-trained large language model, generates personalized process suggestions that meet user needs and click operations, and presents them on the user interface of the interactive interface module;

[0045] S4, the user views the personalized process suggestions presented in the user interface, and inputs feedback information through the user interface provided by the interactive interface module. The interactive interface module understands the feedback information input by the user through natural language processing NLU, extracts key information and sends it to the feedback adjustment module;

[0046] S5. The feedback adjustment module collects user feedback, classifies the user feedback by combining natural language processing technology with manual review, conducts in-depth analysis of the classified user feedback, identifies common problems and improvement points, and generates optimization suggestions.

[0047] Optionally, the model processing module involved first performs preprocessing operations of data cleaning, data conversion and data standardization on the user's behavior data, and then analyzes the user's behavior data through a pre-trained large language model to generate personalized process suggestions that meet user needs and click operations, wherein the pre-training process of the large language model is as follows:

[0048] First, select the large language model GPT as the pre-training model;

[0049] Secondly, based on the user’s historical behavior data and corresponding suggestions, a fine-tuning dataset is constructed;

[0050] Use the fine-tuning dataset to train the large language model GPT. During the training process, set the model hyperparameters.

[0051] Verify and evaluate the performance indicators of the trained large language model GPT, and output the large language model GPT when it meets the preset requirements.

[0052] Optionally, the model processing module analyzes the user's behavior data through a pre-trained large language model to generate personalized process suggestions that meet the user's needs. This process specifically includes:

[0053] Input the preprocessed user behavior data into the large language model GPT;

[0054] The large language model GPT performs reasoning on user data, including: extracting useful features from user behavior data, understanding the user's specific needs and background information, generating and outputting personalized process suggestions that meet user needs and click operations;

[0055] Post-process the personalized process suggestions output by the large language model GPT, including: converting the personalized process suggestions into user-friendly suggestion texts, formatting the suggestion texts, generating and outputting the final suggestions.

[0056] Optionally, the user interface provided by the interactive interface module supports three input modes: text, voice and touch, and supports presenting the personalized process suggestions generated by the model processing module in any mode of text box, chart and graph, and voice output;

[0057] The feedback adjustment module collects user feedback through the interactive interface module, first classifies and deeply analyzes the user feedback through natural language processing technology, identifies common problems and improvement points, then manually reviews the identified common problems and improvement points, and generates optimization suggestions for user feedback after adjustment;

[0058] The model processing module retrains the large language model GPT using the optimization suggestions to optimize the generation strategy of the large language model GPT.

[0059] Compared with the prior art, the interactive personalized process suggestion system and method based on a large model of the present invention has the following beneficial effects:

[0060] 1. The present invention combines the existing workflow management system and adopts a large language model and a real-time user interaction mechanism to dynamically generate personalized suggestions for prefabricated processes or drawn processes according to the specific needs and operation behaviors of users, thereby greatly improving work efficiency and user experience;

[0061] 2. The data collection module of the present invention monitors the user's operating behavior in real time and obtains key information to support subsequent analysis. The model processing module deeply analyzes user data based on pre-trained large-scale language models, such as GPT, and generates personalized process suggestions that meet user needs. The interactive interface module provides a friendly user interface and supports natural language interaction, allowing users to have real-time conversations with the system through text or voice, obtain suggestions and provide feedback. The feedback adjustment module continuously optimizes the model generation strategy according to user feedback information to ensure the relevance and practicality of the suggestions. The present invention can not only optimize pre-made or newly drawn workflows, but also provide personalized suggestions for the workflows selected or drawn by users through pre-trained large-scale language models, and has broad application prospects for process management and decision support in various industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0062] Attached Figure 1 It is a module connection block diagram of the first embodiment of the present invention. DETAILED DESCRIPTION

[0063] In order to make the technical solution, the technical problem solved and the technical effect of the present invention more clearly understood, the technical solution of the present invention is clearly and completely described below in conjunction with specific embodiments.

[0064] Embodiment 1:

[0065] Combined with Figure 1 , this embodiment proposes an interactive personalized process suggestion system based on a large model, which is embedded in a workflow management system, including:

[0066] The data collection module is responsible for real-time monitoring and recording of user behavior data, including process data generated by user-initialized process templates, iterative process templates, and user requests and queries for process templates, and records the frequency and number of times users use different process templates;

[0067] The model processing module is responsible for preprocessing user behavior data, analyzing user behavior data through a pre-trained large language model, and generating personalized process suggestions that meet user needs and click operations;

[0068] The interactive interface module is responsible for providing a user-friendly interface that supports multiple input methods, understands the user's natural language input through natural language processing (NLU), and supports real-time feedback and presentation of personalized process suggestions.

[0069] The feedback adjustment module is responsible for collecting user feedback, classifying user feedback through a combination of natural language processing technology and manual review, conducting in-depth analysis of the classified user feedback, identifying common problems and improvement points, and generating optimization suggestions for personalized processes.

[0070] In this embodiment, the interactive interface module can meet the following functional requirements: ① Support three input methods of text, voice and touch through the user interface, among which text input is performed through a keyboard, a virtual keyboard or handwriting input, and the user's voice input is converted into text using voice recognition technology, and gesture operations and click input are supported on touch screen devices to meet the needs of different users; ② Support the presentation of personalized process suggestions generated by the model processing module through any of text boxes, charts and graphics, and voice output through the user interface to facilitate user understanding; ③ Support understanding the user's natural language input through natural language processing NLU, identify the user's natural language input as three types of intentions: request, query and feedback, and extract key information therefrom, and then send the key information extracted when it is identified as feedback intention to the feedback adjustment module based on the recognition result, and send the key information extracted when it is identified as request or query intention to the data collection module.

[0071] It should be added that (1) after the user completes the input, the interactive interface module needs to preprocess the natural language input by the user to ensure data quality. The preprocessing includes: a) text cleaning: removing irrelevant characters, correcting spelling errors, processing uppercase and lowercase, etc. b) speech recognition: converting speech input into text and using the speech recognition engine to process audio data c) formatting: converting input data into a format that the system can understand. After completing the preprocessing, the interactive interface module uses natural language processing (NLU) to understand the user's natural language input. (2) Use the natural language data set with labeled query intent, request intent, and feedback intent to train the intent recognition model. After the interactive interface module understands the user's natural language input through natural language processing (NLU), it uses the trained intent recognition model to identify the user's natural language input as three types of intent: request, query, and feedback, and extract key information from it. The key information extracted when it is identified as feedback intent is sent to the feedback adjustment module, and the key information extracted when it is identified as request or query intent is sent to the data collection module. (3) The interactive interface module supports real-time feedback, which is specifically manifested in: providing timely feedback mechanisms such as likes and dislikes for the presented personalized process suggestions. Of course, users can also input their own feedback on the presented personalized process suggestions through the user interface.

[0072] In this embodiment, the data collection module monitors and records the user's behavior data in real time through the interactive interface module, including the user's input when searching and requesting prefabricated process templates in the workflow management system (including two types of queries and requests), and the click operation performed by the user when drawing or selecting a workflow diagram. Specifically, JavaScript or other front-end scripts can be embedded in the user interface provided by the interactive interface module to record the user's behavior data in real time, and the data collection module collects the user's behavior data and sends it to the model processing module.

[0073] In this embodiment, the model processing module first preprocesses the user's behavior data. The preprocessing process includes:

[0074] Data cleaning: remove noise, duplicate data and irrelevant information;

[0075] Data conversion: converting user behavior data into a format that can be processed by large language models;

[0076] Data standardization: standardize the format of user behavior data.

[0077] The model processing module then analyzes the user's behavior data through a pre-trained large language model to generate personalized process recommendations that meet the user's needs. The pre-training process of the large language model is as follows:

[0078] First, select the large language model GPT as the pre-training model;

[0079] Secondly, based on the user’s historical behavior data and corresponding suggestions, a fine-tuning dataset is constructed;

[0080] Use the fine-tuning dataset to train the large language model GPT. During the training process, set appropriate model hyperparameters such as learning rate, batch size, and training rounds;

[0081] Verify and evaluate the performance indicators of the trained large language model GPT, use indicators such as accuracy, recall, and F1 value to measure the effectiveness of the model, and output the large language model GPT when it meets the preset requirements.

[0082] The model processing module analyzes user behavior data through a pre-trained large language model and generates personalized process recommendations that meet user needs. This process specifically includes:

[0083] The pre-processed user behavior data is input into the large language model GPT, which stands for Generative Pre-trained Transformer and is suitable for tasks that generate natural language text;

[0084] The large language model GPT performs reasoning on user data, including: extracting useful features from user behavior data, understanding the user's specific needs and background information, generating and outputting personalized process suggestions that meet user needs and click operations;

[0085] Post-process the personalized process suggestions output by the large language model GPT, including: converting the personalized process suggestions into user-friendly suggestion texts, formatting the suggestion texts, generating and outputting the final suggestions.

[0086] In this embodiment, the feedback adjustment module collects user feedback (such as user likes and dislikes on personalized process suggestions, or user input feedback) through the interactive interface module, first classifies and deeply analyzes user feedback through natural language processing technology, identifies common problems and improvement points, and then manually reviews the identified common problems and improvement points, and generates optimization suggestions for user feedback after adjustment. The subsequent model processing module uses the optimization suggestions to retrain the large language model GPT to optimize the generation strategy of the large language model GPT.

[0087] It should be added that (1) user feedback information is not limited to the three types of satisfaction rating (user satisfaction rating of suggestions, such as 1 to 5 stars), comments and suggestions (user detailed comments on suggestions, including specific improvement opinions and suggestions), and error reports (system errors or problems reported by users, such as inaccurate suggestions, interface problems, etc.). (2) The feedback adjustment module first classifies user feedback through natural language processing technology, and the classification results are: satisfaction, suggestions, and problems; then the classified user feedback is deeply analyzed through natural language processing technology, including a) statistical analysis: statistical analysis of feedback data to identify common problems and user needs. For example, calculate the frequency of common keywords or questions in feedback, b) Sentiment analysis: use sentiment analysis technology to evaluate the user's overall emotional tendency towards the system, such as positive, negative or neutral, c) Trend analysis: analyze the time trend of feedback data and identify the changing trends of user needs and problems; then summarize the common problems in user feedback, such as inaccurate suggestions, unfriendly interface, etc., propose corresponding improvement points, and then generate optimization suggestions. The optimization suggestions can be: a) Strategy adjustment: adjust the system's suggestion generation strategy, such as improving the model's training data, optimizing the model algorithm, etc., b) Functional improvement: improve the system function, such as optimizing the user interface, adding new functions, etc., c) Performance improvement: improve the performance of the system, such as increasing response speed, reducing error rate, etc. (3) The optimization suggestions can be subsequently implemented into the system, such as i) system update: updating the system's suggestion generation algorithm, functional modules, and user interface; ii) data retraining: retraining the model based on the new optimization suggestions to improve its performance; iii) function release: releasing system updates to ensure that the optimization function takes effect in the production environment, and then monitoring the system's performance indicators in real time, such as response time, suggestion accuracy, etc., continuing to collect user feedback, and evaluating the effects of the optimization adjustments from three aspects: performance indicator comparison, user feedback, and effect verification, to determine whether the improvement has achieved the expected goals.

[0088] Embodiment 2:

[0089] Reference Figure 1 Based on the interactive personalized process suggestion system described in Example 1, this embodiment further proposes an interactive method for providing personalized process suggestions to users based on a large model, including the following implementation steps:

[0090] S1. The user inputs natural language through the user interface provided by the interactive interface module. The interactive interface module understands the user's natural language input through natural language processing NLU, identifies the user's natural language input as three types of intents: request, query, and feedback, and extracts key information from it. The key information extracted when it is identified as a request or query intent is sent to the data collection module, and the key information extracted when it is identified as a feedback intent is sent to the feedback adjustment module;

[0091] S2, the data collection module monitors and records the user's behavior data in real time, including the process data generated by the user's initialization process template, iterative process template, and the user's request and query process template, records the frequency and number of times the user uses different process templates, and then sends the recorded behavior data to the model processing module;

[0092] S3, the model processing module first pre-processes the user's behavior data, and then analyzes the user's behavior data through the pre-trained large language model GPT, generates personalized process suggestions that meet user needs and click operations, and presents them on the user interface of the interactive interface module;

[0093] S4, the user views the personalized process suggestions presented in the user interface, and inputs feedback information through the user interface provided by the interactive interface module. The interactive interface module understands the feedback information input by the user through natural language processing NLU, extracts key information and sends it to the feedback adjustment module;

[0094] S5. The feedback adjustment module collects user feedback, classifies the user feedback by combining natural language processing technology with manual review, conducts in-depth analysis of the classified user feedback, identifies common problems and improvement points, and generates optimization suggestions.

[0095] In summary, the interactive personalized process suggestion system and method of the present invention that provides personalized process suggestions to users based on a large model can dynamically generate personalized process suggestions according to the user's specific needs and operating behaviors by combining a large language model with a real-time user interaction mechanism, thereby greatly improving work efficiency and enhancing user experience.

[0096] The above specific examples are used to explain the principles and implementation methods of the present invention in detail. These examples are only used to help understand the core technical content of the present invention. Based on the above specific embodiments of the present invention, any improvements and modifications made by technicians in this technical field without departing from the principles of the present invention should fall within the scope of patent protection of the present invention.

Claims

1. An interactive personalized process suggestion system based on a large model, characterized in that: It is embedded in the workflow management system, including: The data collection module is responsible for real-time monitoring and recording of user behavior data, including process data generated by user-initialized process templates, iterative process templates, and user requests and queries for process templates, and records the frequency and number of times users use different process templates; The model processing module is responsible for preprocessing user behavior data, analyzing user behavior data through a pre-trained large language model, and generating personalized process suggestions that meet user needs and click operations; The interactive interface module is responsible for providing a friendly user interface that supports multiple input methods, understands the user's natural language input through natural language processing (NLU), and supports real-time feedback and the presentation of personalized process suggestions. The feedback adjustment module is responsible for collecting user feedback, classifying user feedback through a combination of natural language processing technology and manual review, conducting in-depth analysis of the classified user feedback, identifying common problems and improvement points, and generating optimization suggestions for personalized processes.

2. The interactive personalized process suggestion system based on a large model according to claim 1 is characterized in that: The model processing module preprocesses the user's behavior data, and the preprocessing process includes: Data cleaning: remove noise, duplicate data and irrelevant information; Data conversion: converting user behavior data into a format that can be processed by large language models; Data standardization: standardize the format of user behavior data.

3. The interactive personalized process suggestion system based on a large model according to claim 2 is characterized in that: The model processing module analyzes the user's behavior data through a pre-trained large language model to generate personalized process suggestions that meet the user's needs. The pre-training process of the large language model is as follows: First, select the large language model GPT as the pre-training model; Secondly, based on the user’s historical behavior data and corresponding suggestions, a fine-tuning dataset is constructed; Use the fine-tuning dataset to train the large language model GPT. During the training process, set the model hyperparameters. Verify and evaluate the performance indicators of the trained large language model GPT, and output the large language model GPT when it meets the preset requirements.

4. The interactive personalized process suggestion system based on a large model according to claim 3 is characterized in that: The model processing module analyzes the user's behavior data through a pre-trained large language model to generate personalized process suggestions that meet the user's needs. This process specifically includes: Input the preprocessed user behavior data into the large language model GPT; The large language model GPT performs reasoning on user data, including: extracting useful features from user behavior data, understanding the user's specific needs and background information, generating and outputting personalized process suggestions that meet user needs and click operations; Post-process the personalized process suggestions output by the large language model GPT, including: converting the personalized process suggestions into user-friendly suggestion texts, formatting the suggestion texts, generating and outputting the final suggestions.

5. The interactive personalized process suggestion system based on a large model according to claim 1 is characterized in that: The user interface provided by the interactive interface module supports three input methods: text, voice and touch, and supports presenting the personalized process suggestions generated by the model processing module in any of the following ways: text boxes, charts and graphics, and voice output; The interactive interface module supports understanding the user's natural language input through natural language processing NLU, identifying the user's natural language input as three types of intentions: request, query and feedback, and extracting key information therefrom. Then, based on the recognition results, the key information extracted when it is identified as feedback intention is sent to the feedback adjustment module, and the key information extracted when it is identified as request or query intention is sent to the data collection module.

6. The interactive personalized process suggestion system based on a large model according to claim 5, characterized in that: The user provides feedback on the presented personalized process suggestions through the user interface provided by the interactive interface module; The feedback adjustment module collects user feedback through the interactive interface module, first classifies and deeply analyzes the user feedback through natural language processing technology, identifies common problems and improvement points, then manually reviews the identified common problems and improvement points, and generates optimization suggestions for user feedback after adjustment; The model processing module retrains the large language model GPT using the optimization suggestions to optimize the generation strategy of the large language model GPT.

7. An interactive personalized process suggestion method based on a large model, characterized in that: The interactive personalized process suggestion system according to claim 1 includes the following implementation steps: S1. The user inputs natural language through the user interface provided by the interactive interface module. The interactive interface module understands the user's natural language input through natural language processing NLU, identifies the user's natural language input as three types of intents: request, query, and feedback, and extracts key information from it. The key information extracted when it is identified as a request or query intent is sent to the data collection module, and the key information extracted when it is identified as a feedback intent is sent to the feedback adjustment module; S2, the data collection module monitors and records the user's behavior data in real time, including the process data generated by the user's initialization process template, iterative process template, and the user's request and query process template, records the frequency and number of times the user uses different process templates, and then sends the recorded behavior data to the model processing module; S3, the model processing module first pre-processes the user's behavior data, then analyzes the user's behavior data through a pre-trained large language model, generates personalized process suggestions that meet user needs and click operations, and presents them on the user interface of the interactive interface module; S4, the user views the personalized process suggestions presented in the user interface, and inputs feedback information through the user interface provided by the interactive interface module. The interactive interface module understands the feedback information input by the user through natural language processing NLU, extracts key information and sends it to the feedback adjustment module; S5. The feedback adjustment module collects user feedback, classifies the user feedback by combining natural language processing technology with manual review, conducts in-depth analysis of the classified user feedback, identifies common problems and improvement points, and generates optimization suggestions.

8. The interactive personalized process suggestion method based on a large model according to claim 7, characterized in that: The model processing module first performs preprocessing operations such as data cleaning, data conversion and data standardization on the user's behavior data, and then analyzes the user's behavior data through a pre-trained large language model to generate personalized process suggestions that meet user needs and click operations. The pre-training process of the large language model is as follows: First, select the large language model GPT as the pre-training model; Secondly, based on the user’s historical behavior data and corresponding suggestions, a fine-tuning dataset is constructed; Use the fine-tuning dataset to train the large language model GPT. During the training process, set the model hyperparameters. Verify and evaluate the performance indicators of the trained large language model GPT, and output the large language model GPT when it meets the preset requirements.

9. The interactive personalized process suggestion method based on a large model according to claim 8, characterized in that: The model processing module analyzes the user's behavior data through a pre-trained large language model to generate personalized process suggestions that meet the user's needs. This process specifically includes: Input the preprocessed user behavior data into the large language model GPT; The large language model GPT performs reasoning on user data, including: extracting useful features from user behavior data, understanding the user's specific needs and background information, generating and outputting personalized process suggestions that meet user needs and click operations; Post-process the personalized process suggestions output by the large language model GPT, including: converting the personalized process suggestions into user-friendly suggestion texts, formatting the suggestion texts, generating and outputting the final suggestions.

10. The interactive personalized process suggestion method based on a large model according to claim 7, characterized in that: The user interface provided by the interactive interface module supports three input methods: text, voice and touch, and supports presenting the personalized process suggestions generated by the model processing module in any of the following ways: text boxes, charts and graphics, and voice output; The feedback adjustment module collects user feedback through the interactive interface module, first classifies and deeply analyzes the user feedback through natural language processing technology, identifies common problems and improvement points, then manually reviews the identified common problems and improvement points, and generates optimization suggestions for user feedback after adjustment; The model processing module retrains the large language model GPT using the optimization suggestions to optimize the generation strategy of the large language model GPT.

Images