NLP large model analysis system based on artificial intelligence adaptation

Abstract

The present invention relates to the technical field of NLP big model analysis, and specifically discloses an NLP big model analysis system based on artificial intelligence adaptation, comprising an NLP big model text acquisition module, a first key text processing module, a second key text processing module, a third key text processing module, an adaptive adjustment module, a comprehensive index analysis and judgment module, a human-computer interaction module, a database module and a central processing unit; the present invention collects NLP big model data from a database, analyzes to obtain a first text coefficient, a second text coefficient and a third text coefficient, and then analyzes to obtain an NLP big model comprehensive evaluation index, introduces artificial intelligence adaptation technology, can comprehensively and accurately evaluate the performance of the NLP big model, helps to improve the accuracy, efficiency, stability and user experience of the model, and provides more comprehensive guidance for the optimization of the model.

Description

Technical Field

[0001] The present invention relates to the technical field of NLP large model analysis, and in particular to an NLP large model analysis system based on artificial intelligence adaptation. Background Art

[0002] The Natural Language Processing (NLP) big model is a high-parameter model that uses deep learning technology to understand, interpret, and generate human language. The NLP big model imitates the process of human language processing through encoding and decoding to achieve the ability to output natural language text.

[0003] With the rapid development of Internet technology, the e-commerce industry has become an important part of the global economy. The surge in the number of users and the diversification of shopping behaviors have led to the accumulation of massive amounts of data on e-commerce platforms. In order to improve user experience, optimize product recommendations, and enhance user loyalty, e-commerce platforms are in urgent need of efficient and intelligent user behavior analysis tools. In recent years, the rapid development of artificial intelligence (AI), especially natural language processing (NLP) technology, has brought new opportunities for innovation to e-commerce intelligent customer service systems.

[0004] Traditional NLP large-model analysis systems often only focus on a single performance indicator of the model, such as recall rate or response time, while ignoring other equally important factors, such as the model's text recognition error, text output efficiency, and result output accuracy. This single-dimensional evaluation method cannot fully reflect the pros and cons of the model, and can easily lead to a one-sided understanding of the model's performance; and it usually lacks sufficient dynamic adaptability and cannot adjust the analysis strategy and parameters in time according to new data and scenarios, resulting in the inability to continuously improve the model's performance, and the inability to correct errors in the text output, resulting in low efficiency and poor accuracy in intelligent customer service applications; in addition, traditional analysis systems usually rely on manually set rules and algorithms for analysis, lack intelligent analysis methods, resulting in the analysis process being interfered with by human factors and high labor costs. Summary of the invention

[0005] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides an NLP large model analysis system based on artificial intelligence adaptation to solve the problems raised in the above-mentioned background technology.

[0006] To achieve the above-mentioned objectives, the present invention provides the following technical solutions: an artificial intelligence adaptive NLP large model analysis system, comprising an NLP large model text acquisition module, a first key text processing module, a second key text processing module, a third key text processing module, an adaptive adjustment module, a comprehensive index analysis and judgment module, a human-computer interaction module, a database module and a central processing unit.

[0007] Database module: used to collect the target NLP large model text information to obtain the NLP large model text information database, and store the coefficient text threshold and the comprehensive evaluation index threshold in the database;

[0008] NLP large model text collection module: used to obtain the first key text, the second key text and the third key text from the database module;

[0009] The first key text processing module is used to import the first key text obtained by the NLP large model text acquisition module into the NLP large model text recognition error coefficient model to obtain the first text coefficient;

[0010] The second key text processing module is used to import the second key text obtained by the NLP large model text acquisition module into the NLP large model text output efficiency coefficient model to obtain the second text coefficient;

[0011] The third key text processing module is used to import the third key text obtained by the NLP large model text acquisition module into the NLP large model result output accuracy coefficient model to obtain the third text coefficient;

[0012] Adaptive adjustment module: used to compare the first text coefficient, the second text coefficient and the third text coefficient with the coefficient text threshold preset in the database to determine whether an abnormality occurs, and to start an adjustment and error correction link for the abnormal data;

[0013] Comprehensive index analysis and judgment module: used to import the first text coefficient, the second text coefficient and the third text coefficient adjusted by the adaptive adjustment module into the comprehensive evaluation index model, obtain the comprehensive evaluation index of the NLP large model, and compare and analyze it with the comprehensive evaluation index threshold preset in the database module to obtain the judgment result, and transmit the judgment result to the human-computer interaction module;

[0014] Human-computer interaction module: used to receive NLP large model text analysis results and warning information, intuitively display them through the user interface, and provide diagnostic analysis reports and management suggestions for managers;

[0015] Central Processing Unit: used for data exchange and coordinated control with all modules, using a multi-core processor with built-in data processing algorithms and communication protocol stacks.

[0016] Preferably, the first key text includes the number of correctly output texts of product information, the number of correctly output texts of predicted product information, the number of correctly output texts of product shipping status, the number of correctly output texts of predicted product shipping status, the number of correctly output texts of product logistics delivery time, and the number of correctly output texts of predicted product logistics delivery time; the second key text includes the total number of text output requests, the response time of each text output request, the total number of output texts, the total time required to process text, the total time from text input to output, and the total time of text output delay; the third key text includes the total number of text input characters, the number of incorrect characters recognized in text output, the total number of text input words, the number of incorrect words recognized in text output, the text correct recognition speed, and the result output speed.

[0017] Preferably, the formula of the NLP large model text recognition error coefficient model is as follows:

[0018] ,in, REC Expressed as the first text coefficient, Pcq Indicates the number of correctly output texts for product information. Ppcq Represents the number of correctly output texts for predicted product information, Scq Indicates the correct output text quantity for product delivery status, Pscq Indicates the number of correct output texts for the predicted product delivery status, Dcq Indicates the number of correct text outputs for product logistics delivery time. Pdcq Indicates the number of correctly output texts for the predicted product logistics delivery time.

[0019] Preferably, the mathematical formula calculation steps of the NLP large model text output efficiency coefficient model are as follows:

[0020] A1: Calculate the average response time of the text output of the NLP large model. The calculation model is as follows:

[0021] ,in, Art It is represented by the average response time of text output of NLP large model, and n is the total number of text output requests. rt i is represented as the response time of the i-th text output request, and the response time of each text output request is represented as rt 1, rt 2,……, rt n Second;

[0022] A2: Calculate the second text coefficient. The calculation model is as follows:

[0023] ,in,OEC Expressed as the second text coefficient, Tn Represented as the total amount of text output, Pt Represents the total time required to process the text, Tt It is expressed as the total time from text input to output. Dt Expressed as the total time of text output delay.

[0024] Preferably, the formula of the accuracy coefficient model of the NLP large model result output is as follows:

[0025] ,in, OAC Expressed as the third text coefficient, Tic Indicates the total number of characters entered for the text. Oen Indicates the number of characters recognized incorrectly for text output. Tiw Represents the total number of words in the text input. Own Represents the number of words recognized incorrectly in text output, Cv It is expressed as the speed of correct text recognition. Ov Expressed as the resulting output speed.

[0026] Preferably, the specific manner in which the adaptive adjustment module compares and determines the first text coefficient, the second text coefficient and the third text coefficient with the coefficient text threshold value preset in the database is:

[0027] Extract the first text coefficient and obtain the first text coefficient change fluctuation coefficient formula: , if the fluctuation coefficient of the first text coefficient change is greater than the first text coefficient threshold , it is judged that there is an error in the text recognition of the NLP large model, the abnormal result is sent to the management personnel, and the adjustment and error correction link is initiated. Otherwise, it indicates that there is no error in the text recognition of the NLP large model;

[0028] Extract the second text coefficient and obtain the second text coefficient change fluctuation coefficient formula: , if the second text coefficient change fluctuation coefficient is less than the second text coefficient threshold , then there is an abnormality in the text output efficiency of the NLP large model, the abnormal result is sent to the management personnel, and the adjustment and error correction link is initiated; otherwise, it indicates that there is no abnormality in the text output efficiency of the NLP large model;

[0029] Extract the third text coefficient and obtain the third text coefficient change fluctuation coefficient formula: , if the third text coefficient change fluctuation coefficient is less than the third text coefficient threshold , it is judged that there is an abnormality in the accuracy of the output of the NLP large model results, the abnormal results are sent to the management personnel and users, and the adjustment and error correction link is initiated. Otherwise, it indicates that there is no abnormality in the accuracy of the output of the NLP large model results.

[0030] Preferably, the NLP large model comprehensive evaluation index is calculated as follows:

[0031] ,in, CMEI Expressed as the comprehensive evaluation index of the NLP large model, REC Expressed as the first text coefficient, OEC Expressed as the second text coefficient, OAC Expressed as the third text coefficient, Other influencing factors expressed as a comprehensive evaluation index.

[0032] Preferably, the specific method of comparative analysis of the comprehensive index analysis and judgment module is:

[0033] Obtain the comprehensive evaluation index of the NLP large model and compare it with the preset comprehensive evaluation index threshold. If the comprehensive evaluation index of the NLP large model is greater than the preset comprehensive evaluation index threshold, it indicates that there is an abnormality in the NLP large model. Generate an analysis report for the abnormal data and send it to the corresponding management personnel for review and issue a warning signal. Otherwise, it indicates that there is no abnormality in the NLP large model.

[0034] Technical effects and advantages of the present invention:

[0035] 1. The present invention adopts artificial intelligence algorithm to monitor and analyze model performance in real time, and can automatically adjust the error correction text according to the text data analysis results of the model; the dynamic adaptive capability enables the system to quickly adapt to new data, ensuring the stability and continuous improvement of model performance; at the same time, it also reduces the need for manual intervention and improves analysis efficiency and accuracy.

[0036] 2. Compared with the traditional analysis system that relies on manually set rules and algorithms, the present invention adopts intelligent analysis means and can automatically learn and optimize analysis strategies; it not only improves the accuracy and efficiency of analysis, but also enables the system to process more complex and changeable data; the intelligent analysis and decision-making capabilities make the present system have broader application prospects and development potential in the fields of e-commerce intelligent customer service;

[0037] 3. The present invention collects NLP big model data from the database, analyzes the first text coefficient, the second text coefficient and the third text coefficient, and then analyzes the comprehensive evaluation index of the NLP big model. By introducing artificial intelligence adaptive technology, the performance of the NLP big model can be comprehensively and accurately evaluated, which helps to improve the accuracy, efficiency, stability and user experience of the model, and provides more comprehensive guidance for the optimization of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The present invention is further described using the accompanying drawings, but the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative work.

[0039] Figure 1 It is a schematic diagram of the overall structure of the system of the present invention.

[0040] Figure 2 It is a schematic diagram of the method flow structure of the present invention. DETAILED DESCRIPTION

[0041] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0042] See also Figure 1 As shown, the present invention provides an NLP large model analysis system based on artificial intelligence adaptation, including an NLP large model text acquisition module, a first key text processing module, a second key text processing module, a third key text processing module, an adaptive adjustment module, a comprehensive index analysis and judgment module, a human-computer interaction module, a database module and a central processing unit.

[0043] The output end of the NLP large model data acquisition module is communicatively connected to the input end of the first key text processing module, the output end of the first key text processing module is communicatively connected to the second key text processing module, the output end of the second key text processing module is communicatively connected to the input end of the third key text processing module, the output end of the third key text processing module is communicatively connected to the input end of the adaptive adjustment module, the output end of the adaptive adjustment module is communicatively connected to the input end of the comprehensive index analysis and judgment module and the database module respectively, and the output end of the comprehensive index analysis and judgment module is communicatively connected to the input end of the human-computer interaction module and the database module respectively;

[0044] The database module is used to collect the target NLP large model text information to obtain an NLP large model text information database, and store the coefficient text threshold and the comprehensive evaluation index threshold in the database;

[0045] The NLP large model text acquisition module is used to obtain the first key text, the second key text and the third key text from the database module;

[0046] In this embodiment, it should be specifically noted that the first key text includes the number of correctly output texts of product information, the number of correctly output texts of predicted product information, the number of correctly output texts of product delivery status, the number of correctly output texts of predicted product delivery status, the number of correctly output texts of product logistics delivery time, and the number of correctly output texts of predicted product logistics delivery time;

[0047] The second key text includes the total number of text output requests, the response time of each text output request, the total number of texts output, the total time required to process the text, the total time from text input to output, and the total time of text output delay;

[0048] The third key text includes the total number of characters in the text input, the number of characters with incorrect text output recognition, the total number of words in the text input, the number of words with incorrect text output recognition, the speed of correct text recognition, and the speed of result output.

[0049] The first key text processing module is used to import the first key text obtained by the NLP large model text acquisition module into the NLP large model text recognition error coefficient model to obtain the first text coefficient;

[0050] In this embodiment, it should be specifically explained that the NLP large model text recognition error coefficient model formula is as follows:

[0051] ,in, REC Expressed as the first text coefficient, Pcq Indicates the number of correctly output texts for product information. Ppcq Represents the number of correctly output texts for predicted product information, Scq Indicates the correct output text quantity for product delivery status, Pscq Indicates the number of correct output texts for the predicted product delivery status, Dcq Indicates the number of correct text outputs for product logistics delivery time. Pdcq Indicates the number of correctly output texts for the predicted product logistics delivery time.

[0052] In this embodiment, it should be specifically explained that in the e-commerce intelligent customer service scenario, the questions users ask mainly include product information, product delivery time, and product logistics delivery time. Intelligent customer service can efficiently and correctly reply to users, which is helpful in improving the conversion rate of store products.

[0053] In this embodiment, it should be specifically explained that in the formula It is expressed as root mean square error RMSE, which represents the square root of the mean of the square of the difference between the predicted value and the true value. The smaller the root mean square error RMSE is, the better the prediction is. NLP The smaller the gap between the predicted value and the true value of the large model,NLP The higher the prediction accuracy of the large model, the smaller the NLP large model text recognition error coefficient, indicating that NLP When the large model is applied to the e-commerce intelligent customer service reply, the text recognition error is small and the recognition accuracy of the system is high; the larger the root mean square error RMSE, the better the prediction NLP The larger the gap between the predicted value and the true value of the large model, NLP The lower the prediction accuracy of the large model, the larger the NLP large model text recognition error coefficient, indicating that NLP When large models are applied to e-commerce intelligent customer service responses, the text recognition error is large and the system's recognition accuracy is low.

[0054] The second key text processing module is used to import the second key text obtained by the NLP large model text acquisition module into the NLP large model text output efficiency coefficient model to obtain the second text coefficient;

[0055] In this embodiment, it should be specifically explained that the steps for calculating the mathematical formula of the NLP large model text output efficiency coefficient model are as follows:

[0056] A1: Calculate the average response time of the text output of the NLP large model. The calculation model is as follows:

[0057] ,in, Art It is represented by the average response time of text output of NLP large model, and n is the total number of text output requests. rt i is represented as the response time of the i-th text output request, and the response time of each text output request is represented as rt 1, rt 2,……, rt n Second;

[0058] A2: Calculate the second text coefficient. The calculation model is as follows:

[0059] ,in, OEC Expressed as the second text coefficient, Tn Represented as the total amount of text output, Pt Represents the total time required to process the text, Tt It is expressed as the total time from text input to output. Dt Expressed as the total time of text output delay.

[0060] In this embodiment, it should be specifically explained that the second text coefficient is obtained by analyzing the text output parameters of the NLP large model. Adaptive adjustment based on the analysis results helps to improve the text output efficiency of the NLP large model, which means that the system can process the text input by the user and generate a reply in a shorter time; in the e-commerce intelligent customer service scenario, faster response speed can significantly improve the user experience, reduce user waiting time, and improve user satisfaction and loyalty.

[0061] The third key text processing module is used to import the third key text obtained by the NLP large model text acquisition module into the NLP large model result output accuracy coefficient model to obtain the third text coefficient;

[0062] In this embodiment, it should be specifically explained that the formula of the accuracy coefficient model of the NLP large model result output is as follows:

[0063] ,in, OAC Expressed as the third text coefficient, Tic Indicates the total number of characters entered for the text. Oen Indicates the number of characters recognized incorrectly for text output. Tiw Represents the total number of words in the text input. Own Represents the number of words recognized incorrectly in text output, Cv It is expressed as the speed of correct text recognition. Ov Expressed as the resulting output speed.

[0064] In this embodiment, it should be specifically explained that the third text coefficient is obtained by analyzing the output parameters of the NLP large model results. Adaptive adjustment based on the analysis results helps to improve the accuracy of the NLP large model result output. In the e-commerce intelligent customer service scenario, the high accuracy of the intelligent customer service system reduces the demand for manual customer service, thereby reducing the labor costs of e-commerce companies. By improving the accuracy of system result output, disputes and complaints caused by misunderstandings or incorrect answers are reduced, thereby reducing the error costs of companies.

[0065] The adaptive adjustment module is used to compare the first text coefficient, the second text coefficient and the third text coefficient with the coefficient text threshold preset in the database to determine whether an abnormality occurs, and to start an adjustment and error correction link for the abnormal data;

[0066] In this embodiment, it should be specifically explained that the specific manner in which the adaptive adjustment module compares and determines the first text coefficient, the second text coefficient, and the third text coefficient with the coefficient text threshold value preset in the database is:

[0067] Extract the first text coefficient and obtain the first text coefficient change fluctuation coefficient formula: , if the fluctuation coefficient of the first text coefficient change is greater than the first text coefficient threshold , it is judged that there is an error in the text recognition of the NLP large model, the abnormal result is sent to the management personnel, and the adjustment and error correction link is initiated. Otherwise, it indicates that there is no error in the text recognition of the NLP large model;

[0068] Extract the second text coefficient and obtain the second text coefficient change fluctuation coefficient formula: , if the second text coefficient change fluctuation coefficient is less than the second text coefficient threshold , then there is an abnormality in the text output efficiency of the NLP large model, the abnormal result is sent to the management personnel, and the adjustment and error correction link is initiated; otherwise, it indicates that there is no abnormality in the text output efficiency of the NLP large model;

[0069] Extract the third text coefficient and obtain the third text coefficient change fluctuation coefficient formula: , if the third text coefficient change fluctuation coefficient is less than the third text coefficient threshold , it is judged that there is an abnormality in the accuracy of the output of the NLP large model results, the abnormal results are sent to the management personnel and users, and the adjustment and error correction link is initiated. Otherwise, it indicates that there is no abnormality in the accuracy of the output of the NLP large model results.

[0070] In this embodiment, it should be specifically explained that the adjustment and error correction steps are as follows:

[0071] 1. Strengthen text preprocessing and optimize text preprocessing steps, including removing noise data, performing text segmentation, part-of-speech tagging, etc., to improve the accuracy and efficiency of text recognition.

[0072] 2. Introduce more advanced algorithms in the error correction process, such as deep learning-based error detection and correction algorithms, to more accurately identify and correct errors in the text.

[0073] 3. Refine error classification and processing. Refine the classification according to the error type (such as grammatical errors, spelling errors, semantic errors, etc.), and design different processing strategies for different types of errors. For example, for spelling errors, spelling check algorithms can be used to correct them; for semantic errors, reasoning and corrections need to be combined with context and domain knowledge.

[0074] 4. Strengthen the construction of the knowledge base, continuously improve and update the knowledge base, and ensure that the model can accurately reference relevant information when generating responses. At the same time, strengthen the semantic association and structured processing of the knowledge base to improve the semantic understanding ability of the model.

[0075] The comprehensive index analysis and judgment module is used to import the first text coefficient, the second text coefficient and the third text coefficient adjusted by the adaptive adjustment module into the comprehensive evaluation index model to obtain the comprehensive evaluation index of the NLP large model, and compare and analyze it with the comprehensive evaluation index threshold preset in the database module to obtain a judgment result, and transmit the judgment result to the human-computer interaction module;

[0076] In this embodiment, it should be specifically stated that the calculation of the comprehensive evaluation index of the NLP large model is as follows:

[0077] ,in, CMEI Expressed as the comprehensive evaluation index of the NLP large model, REC Expressed as the first text coefficient, OEC Expressed as the second text coefficient, OAC Expressed as the third text coefficient, Other influencing factors expressed as a comprehensive evaluation index.

[0078] In this embodiment, it should be specifically explained that the specific method of comparative analysis by the comprehensive index analysis and judgment module is:

[0079] Obtain the comprehensive evaluation index of the NLP large model and compare it with the preset comprehensive evaluation index threshold. If the comprehensive evaluation index of the NLP large model is greater than the preset comprehensive evaluation index threshold, it indicates that there is an abnormality in the NLP large model. Generate an analysis report for the abnormal data and send it to the corresponding management personnel for review and issue a warning signal. Otherwise, it indicates that there is no abnormality in the NLP large model.

[0080] The human-computer interaction module is used to receive the NLP large model text analysis results and warning information, intuitively display them through the user interface, and provide diagnostic analysis reports and management suggestions for managers;

[0081] The central processor is used for data exchange and coordinated control with all modules, and adopts a multi-core processor with built-in data processing algorithms and communication protocol stacks.

[0082] See also Figure 2 As shown, the present invention provides an NLP large model analysis method based on artificial intelligence adaptation, comprising the following steps: S1: establishing a database, S2: collecting text data, S3: processing a first key text, S4: processing a second key text, S5: processing a third key text, S6: performing adaptive adjustment, S7: performing comprehensive index analysis and judgment, and S8: performing human-computer interaction.

[0083] S1: Establishing a database: Collecting the target NLP large model text information to obtain an NLP large model text information database, and storing the coefficient text threshold and the comprehensive evaluation index threshold in the database;

[0084] S2: Collecting text data: obtaining a first key text, a second key text, and a third key text from a database module;

[0085] S3: Processing the first key text: importing the first key text obtained by the NLP large model text acquisition module into the NLP large model text recognition error coefficient model to obtain the first text coefficient;

[0086] S4: Processing the second key text: importing the second key text obtained by the NLP large model text acquisition module into the NLP large model text output efficiency coefficient model to obtain the second text coefficient;

[0087] S5: Processing the third key text: importing the third key text obtained by the NLP large model text acquisition module into the NLP large model result output accuracy coefficient model to obtain the third text coefficient;

[0088] S6: Performing adaptive adjustment: comparing the first text coefficient, the second text coefficient and the third text coefficient with the coefficient text threshold preset in the database to determine whether an abnormality occurs, and starting an adjustment and error correction link for the abnormal data;

[0089] S7: Perform comprehensive index analysis and judgment: import the first text coefficient, the second text coefficient and the third text coefficient adjusted by the adaptive adjustment module into the comprehensive evaluation index model to obtain the comprehensive evaluation index of the NLP large model, and compare and analyze it with the comprehensive evaluation index threshold preset in the cloud database module to obtain a judgment result;

[0090] S8: Conduct human-computer interaction: Receive NLP large model text analysis results and warning information, display them intuitively through the user interface, and provide managers with diagnostic analysis reports and management suggestions.

[0091] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

[0092] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. The NLP large model analysis system based on artificial intelligence adaptation is characterized by: include: Database module: used to collect the target NLP large model text information to obtain the NLP large model text information database, and store the coefficient text threshold and the comprehensive evaluation index threshold in the database; NLP large model text collection module: used to obtain the first key text, the second key text and the third key text from the database module; The first key text includes the number of correctly output texts of product information, the number of correctly output texts of predicted product information, the number of correctly output texts of product shipment status, the number of correctly output texts of predicted product shipment status, the number of correctly output texts of product logistics delivery time, and the number of correctly output texts of predicted product logistics delivery time; the second key text includes the total number of text output requests, the response time of each text output request, the total number of output texts, the total time required to process texts, the total time from text input to text output, and the total time of text output delay; the third key text includes the total number of text input characters, the number of text output recognition error characters, the total number of text input words, the number of text output recognition error words, the text correct recognition speed, and the result output speed; The first key text processing module is used to import the first key text obtained by the NLP large model text acquisition module into the NLP large model text recognition error coefficient model to obtain the first text coefficient; The second key text processing module is used to import the second key text obtained by the NLP large model text acquisition module into the NLP large model text output efficiency coefficient model to obtain the second text coefficient; The third key text processing module is used to import the third key text obtained by the NLP large model text acquisition module into the NLP large model result output accuracy coefficient model to obtain the third text coefficient; Adaptive adjustment module: used to compare the first text coefficient, the second text coefficient and the third text coefficient with the coefficient text threshold preset in the database to determine whether an abnormality occurs, and to start an adjustment and error correction link for the abnormal data; The specific manner in which the adaptive adjustment module compares and determines the first text coefficient, the second text coefficient, and the third text coefficient with the coefficient text threshold value preset in the database is: Extract the first text coefficient and obtain the first text coefficient change fluctuation coefficient formula: , if the fluctuation coefficient of the first text coefficient change is greater than the first text coefficient threshold , it is judged that there is an error in the text recognition of the NLP large model, the abnormal result is sent to the management personnel, and the adjustment and error correction link is initiated. Otherwise, it indicates that there is no error in the text recognition of the NLP large model; Extract the second text coefficient and obtain the second text coefficient change fluctuation coefficient formula: , if the second text coefficient change fluctuation coefficient is less than the second text coefficient threshold , then there is an abnormality in the text output efficiency of the NLP large model, the abnormal result is sent to the management personnel, and the adjustment and error correction link is initiated; otherwise, it indicates that there is no abnormality in the text output efficiency of the NLP large model; Extract the third text coefficient and obtain the third text coefficient change fluctuation coefficient formula: , if the third text coefficient change fluctuation coefficient is less than the third text coefficient threshold , it is judged that there is an abnormality in the accuracy of the output of the NLP large model results, and the abnormal results are sent to the management personnel and users, and the adjustment and error correction link is initiated. Otherwise, it indicates that there is no abnormality in the accuracy of the output of the NLP large model results; Comprehensive index analysis and judgment module: used to import the first text coefficient, the second text coefficient and the third text coefficient adjusted by the adaptive adjustment module into the comprehensive evaluation index model, obtain the comprehensive evaluation index of the NLP large model, and compare and analyze it with the comprehensive evaluation index threshold preset in the database module to obtain the judgment result, and transmit the judgment result to the human-computer interaction module; Human-computer interaction module: used to receive NLP large model text analysis results and warning information, intuitively display them through the user interface, and provide diagnostic analysis reports and management suggestions for managers; Central Processing Unit: used for data exchange and coordinated control with all modules, using a multi-core processor with built-in data processing algorithms and communication protocol stacks.

2. The artificial intelligence-adaptive NLP large model analysis system according to claim 1, characterized in that: The formula of the NLP large model text recognition error coefficient model is as follows: ,in, REC Expressed as the first text coefficient, Pcq Indicates the number of correctly output texts for product information. Pcq Represents the number of correctly output texts for predicted product information, Sc Indicates the correct output text quantity for product delivery status, Psq Indicates the number of correct output texts for the predicted product delivery status, Dq Indicates the number of correct text outputs for product logistics delivery time. PcqI Indicates the number of correctly output texts for the predicted product logistics delivery time.

3. The artificial intelligence-adaptive NLP large model analysis system according to claim 1, characterized in that: The mathematical formula calculation steps of the NLP large model text output efficiency coefficient model are as follows: A1: Calculate the average response time of the text output of the NLP large model. The calculation model is as follows: ,in, Art It is represented by the average response time of text output of NLP large model, and n is the total number of text output requests. rt i is represented as the response time of the i-th text output request, and the response time of each text output request is represented as rt 1, rt 2,……, rt n Second; A2: Calculate the second text coefficient. The calculation model is as follows: ,in, OEC Expressed as the second text coefficient, Tn Represented as the total amount of text output, Pt Represents the total time required to process the text, Tt It is expressed as the total time from text input to output. Dt Expressed as the total time of text output delay.

4. The artificial intelligence adaptive NLP large model analysis system according to claim 1 is characterized by: The formula of the accuracy coefficient model of the NLP large model result output is as follows: ,in, OAC Expressed as the third text coefficient, Tic Indicates the total number of characters entered for the text. Oen Indicates the number of characters recognized incorrectly for text output. Tiw Represents the total number of words in the text input. Own Represents the number of words recognized incorrectly in text output, Cv It is expressed as the speed of correct text recognition. Ov Expressed as the resulting output speed.

5. The artificial intelligence adaptive NLP large model analysis system according to claim 1 is characterized by: The calculation of the comprehensive evaluation index of the NLP large model is as follows: ,in, CMEI Expressed as the comprehensive evaluation index of the NLP large model, REC Expressed as the first text coefficient, OEC Expressed as the second text coefficient, OAC Expressed as the third text coefficient, Other influencing factors expressed as a comprehensive evaluation index.

6. The artificial intelligence-adaptive NLP large model analysis system according to claim 1, characterized in that: The specific method of comparative analysis of the comprehensive index analysis and judgment module is as follows: Obtain the comprehensive evaluation index of the NLP large model and compare it with the preset comprehensive evaluation index threshold. If the comprehensive evaluation index of the NLP large model is greater than the preset comprehensive evaluation index threshold, it indicates that there is an abnormality in the NLP large model. Generate an analysis report for the abnormal data and send it to the corresponding management personnel for review and issue a warning signal. Otherwise, it indicates that there is no abnormality in the NLP large model.

Images