Water affair industry knowledge question-answering method and system based on large model technology

Abstract

The invention belongs to the technical field of computers. The invention provides a water affair industry knowledge question-answering method and system based on a large model technology. The method comprises the steps that a basic service layer receives a water affair industry problem input by a user and transmits the water affair industry problem to a model calling layer; the model calling layer calls a first question and answer model to evaluate the answering difficulty of the water affair industry question; if the answer difficulty obtained through evaluation is lower than a difficulty threshold value, the first question and answer model answers the water affair industry question; and if the answer difficulty obtained by evaluation is higher than a difficulty threshold, calling a plurality of second question and answer models in the basic model layer to answer the water affair industry question. Through cooperation of the local first question and answer model and other commercial large models, accurate answering of the water affair industry questions is realized.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to a water industry knowledge question and answer method and system based on large model technology. Background Art

[0002] The large-model knowledge question-answering system is built based on artificial intelligence and big data technologies, aiming to provide efficient and accurate knowledge question-answering services. The system mainly builds a large-scale knowledge base based on mature large-model technology, and realizes automatic understanding and answering of user questions through technologies such as NLP (natural language processing). The large-model knowledge question-answering system involves multiple key technologies, including knowledge representation and modeling, problem understanding and parsing, knowledge retrieval and fusion, answer generation and evaluation, etc. These technologies together constitute the core of the large-model knowledge question-answering system, enabling it to achieve rapid and accurate responses to complex questions.

[0003] Although the existing technology has made significant progress in large-scale knowledge question answering systems, there are still some limitations. For example, the understanding and answers to some complex questions may not be accurate enough, especially for the water industry, which involves multiple categories of questions such as raw water, water supply, sewage, water environment, construction and development, technology and services. It is more difficult to answer.

[0004] In summary, the current single-source Q&A large model is unable to effectively deal with some complex water industry knowledge, resulting in the inability to provide customers with a good Q&A experience, and urgently needs to be improved. Summary of the invention

[0005] In this regard, the present invention provides a water industry knowledge question and answer method, system, electronic device, computer storage medium and computer program product based on large model technology to solve the above technical problems.

[0006] The present invention discloses a water industry knowledge question-answering method based on a large model technology, the method comprising the following steps:

[0007] The basic service layer receives water industry problems input by users, and transmits the water industry problems to the model calling layer;

[0008] The model calling layer calls the first question-answering model to evaluate the difficulty of answering the water industry question; wherein the first question-answering model is embedded in the model calling layer;

[0009] If the difficulty of the answer determined by the evaluation is lower than the difficulty threshold, the first question-answering model will answer the water industry question;

[0010] If the difficulty of the obtained answer is higher than the difficulty threshold, several second Q&A models in the basic model layer are called to answer the water industry question.

[0011] Optionally, the model call layer evaluates the difficulty of the answer of the first Q&A model to the water industry question, including:

[0012] The model call layer performs semantic reduction processing on the water industry question to obtain a reduced water industry question, and asks the first Q&A model according to the reduced water industry question;

[0013] The first Q&A model queries the water industry knowledge database based on the reduced water industry question and outputs a trial answer content;

[0014] Semantic parsing is performed on the trial answer content to obtain the content length and the number of content categories, and the answer difficulty is calculated according to the content length and the number of content categories.

[0015] Optionally, calculating the answer difficulty according to the content length and the number of content categories includes:

[0016] The answer difficulty is calculated based on the content length, the number of content categories and the following formula:

[0017]

[0018] In the formula, d is the answer difficulty to the water industry question, s is the content length corresponding to the trial answer content, is the average value of the content lengths corresponding to each trial answer content recorded; u is the number of content categories corresponding to the trial answer content, is the average value of the number of content categories corresponding to each trial answer content recorded; a1, a2, b1, b2 are all constants, and a1 < a2, b1 < b2, and s > a1, u > n1, α1, α2 are weighting coefficients, and α1 < α2.

[0019] Optionally, calling several second Q&A models in the basic model layer to answer the water industry question includes:

[0020] Several second Q&A models are determined according to the answer difficulty, and the number of calls of the second Q&A model is positively correlated with the answer difficulty;

[0021] Receive the first answer content of each second Q&A model to the water industry question, perform semantic analysis on each first answer content to obtain each semantic content, and perform semantic fusion on each semantic content to obtain a second answer content;

[0022] The second answer content is fed back to the basic service layer, and the basic service layer outputs it to the user.

[0023] Optionally, the method further comprises:

[0024] The model calling layer also generates a number of similar water industry questions based on the water industry questions whose answer difficulty is higher than the difficulty threshold, and outputs the similar water industry questions to each of the second question-answering models;

[0025] Receive the third answer content of each of the second question-answer models, and associate the similar water industry questions with each of the third answer contents one by one to form a piece of training data, and each piece of the training data constitutes a training data set;

[0026] The first question-answering model is trained using the training dataset.

[0027] Optionally, the method further comprises:

[0028] When the model calling layer identifies that the proportion of the water industry questions whose answer difficulty is higher than the difficulty threshold is higher than the proportion threshold, it generates a number of similar water industry questions based on the water industry questions whose answer difficulty is higher than the difficulty threshold, and outputs the similar water industry questions to each of the second question-and-answer models.

[0029] The present invention also discloses a water industry knowledge question-answering system based on large model technology, the system comprising a processing module and a storage module, the processing module executes the computer code in the storage module to implement the following method steps:

[0030] The basic service layer receives water industry problems input by users, and transmits the water industry problems to the model calling layer;

[0031] The model calling layer calls the first question-answering model to evaluate the difficulty of answering the water industry question; wherein the first question-answering model is embedded in the model calling layer;

[0032] If the difficulty of the answer determined by the evaluation is lower than the difficulty threshold, the first question-answering model will answer the water industry question;

[0033] If the difficulty of the answer determined by the assessment is higher than the difficulty threshold, several second question-answering models in the basic model layer are called to answer the water industry question.

[0034] The present invention also discloses an electronic device, comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement any of the above methods.

[0035] The present invention also discloses a computer storage medium, wherein the computer-readable storage medium stores a computer program, and the computer program is executed by a processor to implement any of the above methods.

[0036] The present invention also discloses a computer program product, which can implement any of the above methods when the computer program product is executed by a processor.

[0037] After receiving the water industry question input by the user, the present invention does not directly answer the question, but first controls the local first question and answer model to evaluate the difficulty of the answer. When the answer difficulty is low, it is preferred to answer through the local first question and answer model, which can improve the answer efficiency; when the answer difficulty is high, it switches to use other large commercial models to answer the water industry question, thereby ensuring the accuracy of the answer content to the water industry question based on the multi-source question and answer model. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0039] Figure 1 It is a flow chart of a water industry knowledge question and answer method based on a large model technology disclosed in an embodiment of the present invention;

[0040] Figure 2 It is a schematic diagram of the three-layer architecture of the water industry knowledge question-and-answer system disclosed in an embodiment of the present invention;

[0041] Figure 3 It is a structural diagram of a water industry knowledge question and answer system based on large model technology disclosed in an embodiment of the present invention. DETAILED DESCRIPTION

[0042] The following is a description of the implementation of the present application by specific specific embodiments. People familiar with the technology can easily understand other advantages and effects of the present application from the contents disclosed in this specification. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work are within the scope of protection of the present application.

[0043] In addition, the technical features involved in the different embodiments of the present application described below can be combined with each other as long as they do not conflict with each other.

[0044] like Figure 1 , Figure 2 As shown, the embodiment of the present invention discloses a water industry knowledge question and answer method based on a large model technology, and the method comprises the following steps:

[0045] The basic service layer receives water industry problems input by users, and transmits the water industry problems to the model calling layer;

[0046] The model calling layer calls the first question-answering model to evaluate the difficulty of answering the water industry question; wherein the first question-answering model is embedded in the model calling layer;

[0047] If the difficulty of the answer determined by the evaluation is lower than the difficulty threshold, the first question-answering model will answer the water industry question;

[0048] If the difficulty of the answer determined by the assessment is higher than the difficulty threshold, several second question-answering models in the basic model layer are called to answer the water industry question.

[0049] The solution of the present invention is based on a three-layer structure of a water industry knowledge question and answer system, which includes a basic model layer, a model call layer and a basic service layer. Among them, the basic model layer is used to receive water industry questions input by users, and can be embedded with AI services, PDF parsing services, QA parsing services, etc., to achieve accurate parsing and understanding of various types of water industry questions input by users; the model call layer is the core processing layer of the present invention, which includes a processor and a corresponding first question and answer model, and the first question and answer model is a local model embedded in the model call layer; the basic model layer is provided with an interface component that is connected to multiple second question and answer models, and the second question and answer model is a large commercial model, such as the Tongyi Qianwen series model, the Wenxin Yiyan series model, the Zhipu Qingyan series model, etc.

[0050] Based on the above three-layer architecture, the specific implementation method of the scheme of the present invention is as follows: the user inputs specific question information and / or corresponding attachment content in the basic service layer, and the basic service layer performs semantic analysis on the input content to determine the user's water industry knowledge problem, that is, the water industry problem, and transmits the water industry problem to the model calling layer. After the model calling layer receives the water industry problem, it calls the first question and answer model embedded locally to evaluate the difficulty of answering the water industry problem. If the difficulty of the answer obtained by the evaluation is lower than the difficulty threshold, it indicates that the local first question and answer model can deal with the water industry problem. At this time, the first question and answer model can be controlled to answer the water industry problem. If the difficulty of the answer obtained by the evaluation is higher than the difficulty threshold, it indicates that the local first question and answer model is difficult to deal with the water industry problem. At this time, it is necessary to call several second question and answer models in the basic model layer, that is, the large commercial model, to answer the water industry problem.

[0051] Therefore, after receiving the water industry question input by the user, the present invention does not directly answer it, but first controls the local first question and answer model to evaluate the difficulty of the answer. When the answer difficulty is low, it is preferred to answer it through the local first question and answer model, which can improve the answer efficiency; when the answer difficulty is high, it switches to use other large commercial models to answer the water industry question, thereby ensuring the accuracy of the answer content to the water industry question based on the multi-source question and answer model.

[0052] Optionally, the model calling layer calls the first question-answering model to evaluate the difficulty of answering the water industry question, including:

[0053] The model calling layer performs semantic simplification on the water industry question to obtain a simplified water industry question, and asks the first question-answering model questions based on the simplified water industry question;

[0054] The first question-answering model queries the water industry knowledge database based on the simplified water industry question and outputs a trial answer content;

[0055] The test answer content is semantically parsed to obtain the content length and the number of content categories, and the answer difficulty is calculated based on the content length and the number of content categories.

[0056] In these embodiments, after the model call layer receives the water industry problems forwarded by the basic service layer, the water industry problems are first processed by streamlining. The streamlining process refers to processing such as summarizing the detailed water industry problems and extracting discrete keywords, so as to obtain streamlined water industry problems. The streamlined water industry problems are used to query the first question-and-answer model. The first question-and-answer model queries the water industry knowledge database through the built-in decision-making mechanism, and makes decisions and outputs corresponding trial answer content. It should be noted that since the streamlined water industry problems only contain some broad content compared with the water industry problems before processing, the trial answer content output by it will include more content with low relevance. Then, semantic parsing is performed on the trial answer content, and the content length and the number of content categories of the trial answer content can be obtained. The content length refers to the total length of the information involved in the trial answer, such as the number of characters, and the number of content categories refers to the number of categories of the content with low relevance included in the trial answer. Finally, the answer difficulty is calculated based on the analyzed content length and the number of content categories.

[0057] It should be noted that generally, when the content length included in the trial answer content is longer and the number of content categories is larger, it indicates that the basic information involved in the water industry problem is more, the processing difficulty of the first question-and-answer model is greater, and it is more likely to make mistakes, that is, it is more likely to answer off-topic. At the same time, due to the differences in the description methods of the water industry problems by users, it may cause the content length and the number of content categories included in the trial answer content of the first question-and-answer model to deviate from the true intention corresponding to the water industry problem, resulting in an incorrect judgment of the answer difficulty. Therefore, by streamlining the original water industry problems, the present invention can appropriately increase the content length and the number of content categories of the trial answer content output by the first question-and-answer model, which is conducive to more easily analyzing and determining its answer difficulty.

[0058] Optionally, calculating the answer difficulty according to the content length and the number of content categories includes:

[0059] The answer difficulty is calculated based on the content length, the number of content categories, and the following formula:

[0060]

[0061] In the formula, d is the answer difficulty for the water industry problem, s is the content length corresponding to the trial answer content, is the average value of the content lengths corresponding to each recorded trial answer content; u is the number of content categories corresponding to the trial answer content, is the average value of the number of content categories corresponding to each recorded trial answer content; a1, a2, b1, and b2 are all constants, and a1 < a2, b1 < b2, and s > a1, u>b1, α1 and α2 are weighting coefficients, and α1<α2.

[0062] In some embodiments, the content length and the number of content categories determined above are substituted into the above calculation formula to obtain the difficulty of answering the water industry question by the first question-answering model.

[0063] The above formula contains two parts, namely, analyzing the answer difficulty corresponding to the content length and the number of content categories respectively, and then integrating and calculating them in a weighted manner. Since more content categories are more correlated with answer difficulty than more content length, that is, the more irrelevant content is included in the answer, the more difficult it is to answer the water industry question; and the length of the answer content is affected by many factors, so its weight is set slightly lower. Among them, α1+α2=1.

[0064] in, It is the average value of the content length and number of content categories corresponding to the trial answer content obtained by statistical analysis of the corresponding historical data recorded.

[0065] Optionally, calling several second question-answering models in the basic model layer to answer the water industry question includes:

[0066] Determine a plurality of second question-answering models according to the difficulty of the answer, wherein the number of calls of the second question-answering models is positively correlated with the difficulty of the answer;

[0067] Receiving the first answer content of each of the second question-answer models to the water industry question, performing semantic analysis on each of the first answer contents to obtain each semantic content, and performing semantic fusion on each of the semantic contents to obtain the second answer content;

[0068] The second answer content is fed back to the basic service layer, and the basic service layer outputs it to the user.

[0069] In some embodiments, when the local first question-and-answer model is difficult to accurately answer the water industry question, several second question-and-answer models in the scheduling basic model layer are selected to answer the water industry question. After receiving the first answer content of each second question-and-answer model to the water industry question, first perform semantic analysis on each first answer content to obtain each semantic content, and then perform semantic fusion on each semantic content to obtain the second answer content. After such processing, it is achieved that multiple first answer contents are fused into one second answer content, and the second answer content is used to represent each first answer content.

[0070] Among them, the number of the second question and answer models dispatched is positively correlated with the answer difficulty obtained above, that is, when the answer difficulty is higher, more second question and answer models are selected to participate in the answer. Such a setting is conducive to improving the accuracy of the answer to the water industry question; conversely, fewer second question and answer models are selected to participate in the answer. Such a setting is conducive to improving the efficiency of answering the water industry question and can also reduce the difficulty of fusion processing.

[0071] Optionally, the method further comprises:

[0072] The model calling layer also generates a number of similar water industry questions based on the water industry questions whose answer difficulty is higher than the difficulty threshold, and outputs the similar water industry questions to each of the second question-answering models;

[0073] Receive the third answer content of each of the second question-answer models, and associate the similar water industry questions with each of the third answer contents one by one to form a piece of training data, and each piece of the training data constitutes a training data set;

[0074] The first question-answering model is trained using the training dataset.

[0075] In some embodiments, the present invention also sets up an update training mechanism for the first question-answering model, that is, when the first question-answering model is difficult to accurately answer a certain water industry question, the model call layer generates several water industry questions similar to the water industry question, and outputs these similar water industry questions to each second question-answering model one by one, and associates the obtained third answer content with similar water industry questions one by one, thereby forming a set of training data sets. The training data set is then used to train the first question-answering model, thereby improving the first question-answering model's ability to answer this type of water industry question.

[0076] In addition, through the above method, there is no need to manually collect training data, which can effectively reduce the difficulty of maintaining the first question-answering model. As for the second question-answering model, since it is a large commercial model, it can be updated more quickly through the use of a large number of users. The present invention uses these large commercial models to achieve the update of the local first question-answering model.

[0077] Optionally, the method further comprises:

[0078] When the model calling layer identifies that the proportion of the water industry questions whose answer difficulty is higher than the difficulty threshold is higher than the proportion threshold, it generates a number of similar water industry questions based on the water industry questions whose answer difficulty is higher than the difficulty threshold, and outputs the similar water industry questions to each of the second question-and-answer models.

[0079] In some embodiments, the update training of the first question-and-answer model should be restrained, rather than immediately conducting update training as soon as a water industry question with a difficulty level higher than the difficulty threshold appears. In this regard, the present invention is configured to trigger the update training of the first question-and-answer model when the existence ratio of water industry questions with a difficulty level higher than the difficulty threshold is identified to be higher than the ratio threshold, that is, to generate a number of similar water industry questions based on water industry questions with a difficulty level higher than the difficulty threshold, and output the similar water industry questions to each second question-and-answer model. In addition, the statistical calculation of the existence ratio should also be within a set statistical period, such as one week.

[0080] like Figure 3 As shown, the embodiment of the present invention also discloses a water industry knowledge question and answer system based on large model technology, the system includes a processing module and a storage module, the processing module executes the computer code in the storage module to implement the following method steps:

[0081] The basic service layer receives water industry problems input by users, and transmits the water industry problems to the model calling layer;

[0082] The model calling layer calls the first question-answering model to evaluate the difficulty of answering the water industry question; wherein the first question-answering model is embedded in the model calling layer;

[0083] If the difficulty of the answer determined by the evaluation is lower than the difficulty threshold, the first question-answering model will answer the water industry question;

[0084] If the difficulty of the answer determined by the assessment is higher than the difficulty threshold, several second question-answering models in the basic model layer are called to answer the water industry question.

[0085] An embodiment of the present invention further discloses an electronic device, comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the method described in the above embodiment.

[0086] An embodiment of the present invention further discloses a computer storage medium, wherein the computer storage medium stores a computer program, and the computer program is executed by a processor to implement the method described in the above embodiment.

[0087] The embodiment of the present invention further discloses a computer program product, which can implement the method described in the above embodiment when the computer program product is executed by a processor.

[0088] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0089] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0090] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, or can be electrical, mechanical or other forms of connection.

[0091] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0092] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0093] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the various embodiments of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0094] The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. A water industry knowledge question-answering method based on large model technology, characterized by: The method comprises the following steps: The basic service layer receives water industry problems input by users, and transmits the water industry problems to the model calling layer; The model calling layer calls the first question-answering model to evaluate the difficulty of answering the water industry question; wherein the first question-answering model is embedded in the model calling layer; If the difficulty of the answer determined by the evaluation is lower than the difficulty threshold, the first question-answering model will answer the water industry question; If the difficulty of the answer determined by the assessment is higher than the difficulty threshold, several second question-answering models in the basic model layer are called to answer the water industry question.

2. The water industry knowledge question-answering method based on large model technology according to claim 1 is characterized by: The model calling layer calls the first question-answering model to evaluate the difficulty of answering the water industry question, including: The model calling layer performs semantic simplification on the water industry question to obtain a simplified water industry question, and asks the first question-answering model questions based on the simplified water industry question; The first question-answering model queries the water industry knowledge database based on the simplified water industry question and outputs a trial answer content; The test answer content is semantically parsed to obtain the content length and the number of content categories, and the answer difficulty is calculated based on the content length and the number of content categories.

3. The water industry knowledge question and answer method based on large model technology according to claim 2 is characterized by: The difficulty of answering is calculated according to the content length and the number of content categories, including: The difficulty of the answer is calculated based on the content length, the number of content categories and the following formula: Wherein, d is the difficulty of the answer to the water industry problem, s is the content length corresponding to the trial answer content, which is the average value of the content lengths corresponding to each recorded trial answer content; u is the number of content categories corresponding to the trial answer content, which is the average value of the number of content categories corresponding to each recorded trial answer content; a1, a2, b1, and b2 are all constants, and a1 < a2, b1 < b2, and s > a1, u > b1, α1 and α2 are weighting coefficients, and α1 < α2.

4. The water industry knowledge question-answering method based on large model technology according to claim 3 is characterized by: Several second question-answering models in the basic model layer are called to answer the water industry questions, including: Determine a plurality of second question-answering models according to the difficulty of the answer, wherein the number of calls of the second question-answering models is positively correlated with the difficulty of the answer; Receiving the first answer content of each of the second question-answer models to the water industry question, performing semantic analysis on each of the first answer contents to obtain each semantic content, and performing semantic fusion on each of the semantic contents to obtain the second answer content; The second answer content is fed back to the basic service layer, and the basic service layer outputs it to the user.

5. The water industry knowledge question-answering method based on large model technology according to claim 4 is characterized by: The method further comprises: The model calling layer also generates a number of similar water industry questions based on the water industry questions whose answer difficulty is higher than the difficulty threshold, and outputs the similar water industry questions to each of the second question-answering models; Receive the third answer content of each of the second question-answer models, and associate the similar water industry questions with each of the third answer contents one by one to form a piece of training data, and each piece of the training data constitutes a training data set; The first question-answering model is trained using the training dataset.

6. The water industry knowledge question-answering method based on large model technology according to claim 5 is characterized by: The method further comprises: When the model calling layer identifies that the proportion of the water industry questions whose answer difficulty is higher than the difficulty threshold is higher than the proportion threshold, it generates a number of similar water industry questions based on the water industry questions whose answer difficulty is higher than the difficulty threshold, and outputs the similar water industry questions to each of the second question-and-answer models.

7. A water industry knowledge question-answering system based on large model technology, the system comprising a processing module and a storage module, characterized in that: The processing module executes the computer code in the storage module to implement the following method steps: The basic service layer receives water industry problems input by users, and transmits the water industry problems to the model calling layer; The model calling layer calls the first question-answering model to evaluate the difficulty of answering the water industry question; wherein the first question-answering model is embedded in the model calling layer; If the difficulty of the answer determined by the evaluation is lower than the difficulty threshold, the first question-answering model will answer the water industry question; If the difficulty of the answer determined by the assessment is higher than the difficulty threshold, several second question-answering models in the basic model layer are called to answer the water industry question.

8. An electronic device comprising: At least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the method according to any one of claims 1 to 6.

9. A computer storage medium storing a computer program, characterized in that: The computer program is executed by a processor to implement the method according to any one of claims 1 to 6.

10. A computer program product, characterized in that: When the computer program product is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images