Work order processing method, apparatus, device, and readable storage medium

Abstract

The application discloses a work order processing method, device and equipment and a readable storage medium. After obtaining a to-be-processed work order, first, first prompt information for processing the work order and indicating summarizing the content of the work order is determined, then, based on the first prompt information, the summary content of the work order is obtained; then, based on the summary content of the work order, whether there is a similar question to the summary content of the work order in a basic knowledge base is searched, and a first search result is obtained; finally, according to the first search result, the work order is processed, and a reply corresponding to the work order is obtained. Based on the scheme, automatic work order processing can be realized, and more efficient and high-quality government affairs question answering can be realized.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and more specifically, to a work order processing method, apparatus, device, and readable storage medium. Background Technology

[0002] In government Q&A scenarios, the traditional method of processing work orders is mainly manual. As the number and types of work orders increase, the traditional manual work order processing method can no longer achieve more efficient and higher-quality government Q&A.

[0003] Therefore, how to provide a method for processing work orders to achieve more efficient and higher-quality government affairs Q&A has become a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0004] In view of the above problems, this application proposes a work order processing method, apparatus, equipment, and readable storage medium. The specific solution is as follows:

[0005] A work order processing method, the method comprising:

[0006] Get the work orders to be processed;

[0007] A first prompt message is determined, which is used to instruct the summary of the contents of the work order;

[0008] Based on the first prompt information, the summary content of the work order is obtained;

[0009] Based on the summary content of the work order, a search is conducted in the basic knowledge base to determine if a first question exists. The first question is a question in the basic knowledge base that is similar to the summary content of the work order, and a first search result is obtained.

[0010] Based on the first search result, the work order is processed to obtain the corresponding response.

[0011] Optionally, if the first search result is yes, then processing the work order based on the first search result to obtain the response corresponding to the work order includes:

[0012] Get the answer to the first question;

[0013] A second prompt message is determined, which is used to instruct the answer to the first question to be rewritten;

[0014] Based on the second prompt information, a response to the work order is obtained.

[0015] Optionally, if the first search result is negative, then processing the work order based on the first search result to obtain a response corresponding to the work order includes:

[0016] A third prompt message is determined, which is used to instruct the work order to be classified by domain.

[0017] Based on the third prompt information, the domain category of the work order is obtained;

[0018] The system retrieves whether a second question exists in the professional knowledge base corresponding to the domain category of the work order, and the second question is a question similar to the work order in the professional knowledge base, thus obtaining a second search result.

[0019] Based on the second search result, the work order is processed to obtain the corresponding response.

[0020] Optionally, if the second search result is yes, then processing the work order based on the second search result to obtain the response corresponding to the work order includes:

[0021] Obtain the answer to the second question;

[0022] A fourth prompt message is determined, which is used to instruct the answer to the second question to be rewritten;

[0023] Based on the fourth prompt information, a response to the work order is obtained.

[0024] Optionally, if the second search result is negative, then processing the work order based on the second search result to obtain a response corresponding to the work order includes:

[0025] The fifth prompt message is determined, which is used to instruct the generation of a question-and-answer pair corresponding to the work order based on the government document;

[0026] Based on the fifth prompt information, a question-and-answer pair corresponding to the work order is obtained;

[0027] Obtain the answer from the question-and-answer pair corresponding to the work order;

[0028] A sixth prompt message is determined, which is used to instruct the rewriting of the answer in the question-and-answer pair corresponding to the work order;

[0029] Based on the sixth prompt information, a response to the work order is obtained.

[0030] Optionally, after generating a question-and-answer pair corresponding to the work order based on the fifth prompt information, the method further includes:

[0031] The basic knowledge base and / or the professional knowledge base are updated using the question-and-answer pairs corresponding to the work order.

[0032] Optionally, before obtaining the work order to be processed, the method further includes:

[0033] The seventh prompt message is determined, which is used to instruct the processing of government documents and generate multiple question-and-answer pairs, wherein each question-and-answer pair includes a question and a corresponding answer, and the questions in each question-and-answer pair are all related to the government documents;

[0034] Based on the seventh prompt information, multiple question-answer pairs are obtained;

[0035] Using the aforementioned question-and-answer pairs, a basic knowledge base and a professional knowledge base are constructed.

[0036] Optionally, after receiving the response corresponding to the work order, the method further includes:

[0037] Based on the work order, the contact person and contact information are determined;

[0038] The corresponding reply for the work order is sent back to the contact person using the aforementioned contact method.

[0039] A work order processing device, the device comprising:

[0040] The work order acquisition unit is used to acquire work orders to be processed.

[0041] The first prompt information determining unit is used to determine the first prompt information, which is used to indicate the summary of the content of the work order;

[0042] The content summary unit is used to obtain the summary content of the work order based on the first prompt information;

[0043] The first retrieval unit is used to search the basic knowledge base for a first question based on the summary content of the work order, wherein the first question is a question in the basic knowledge base that is similar to the summary content of the work order, and to obtain a first retrieval result.

[0044] The processing unit is used to process the work order based on the first search result and obtain the response corresponding to the work order.

[0045] Optionally, if the first search result is yes, then the processing unit includes:

[0046] The first answer acquisition unit is used to acquire the answer corresponding to the first question;

[0047] The second prompt information determining unit is used to determine the second prompt information, which is used to instruct the answer to the first question to be rewritten.

[0048] The first work order response determination unit is used to obtain a response to the work order based on the second prompt information.

[0049] Optionally, if the first search result is negative, the processing unit includes:

[0050] The third prompt information determination unit is used to determine the third prompt information, which is used to indicate the domain classification of the work order.

[0051] The domain category determination unit is used to determine the domain category of the work order based on the third prompt information;

[0052] The second retrieval unit is used to search the professional knowledge base corresponding to the domain category of the work order to see if there is a second question, the second question being a question similar to the work order in the professional knowledge base, and to obtain a second retrieval result;

[0053] The second work order response determination unit is used to process the work order based on the second search result to obtain the response corresponding to the work order.

[0054] Optionally, if the second search result is yes, then the second work order response confirmation unit includes:

[0055] The second answer acquisition unit is used to acquire the answer to the second question.

[0056] The fourth prompt information determining unit is used to determine the fourth prompt information, which is used to instruct the answer corresponding to the second question to be rewritten;

[0057] The third work order response confirmation unit is used to obtain a response to the work order based on the fourth prompt information.

[0058] Optionally, if the second search result is negative, the second work order response confirmation unit includes:

[0059] The fifth prompt information determination unit is used to determine the fifth prompt information, which is used to instruct the generation of a question-and-answer pair corresponding to the work order based on the government document;

[0060] The question-answer pair determination unit is used to obtain the question-answer pair corresponding to the work order based on the fifth prompt information;

[0061] The question-and-answer pair answer acquisition unit is used to acquire the answer from the question-and-answer pair corresponding to the work order;

[0062] The sixth prompt information determining unit is used to determine the sixth prompt information, which is used to instruct the answer in the question-answer pair corresponding to the work order to be rewritten;

[0063] The fourth work order response confirmation unit is used to obtain a response to the work order based on the sixth prompt information.

[0064] Optionally, the device further includes:

[0065] The knowledge base update unit is used to update the basic knowledge base and / or the professional knowledge base using the question-and-answer pair corresponding to the work order after generating the question-and-answer pair corresponding to the work order based on the fifth prompt information.

[0066] Optionally, the device further includes:

[0067] The knowledge base construction unit is used to determine a seventh prompt message before obtaining the work order to be processed. The seventh prompt message is used to instruct the processing of government documents and generate multiple question-answer pairs, wherein each question-answer pair includes a question and a corresponding answer, and the questions in each question-answer pair are all related to the government documents; based on the seventh prompt message, multiple question-answer pairs are obtained; and using the multiple question-answer pairs, a basic knowledge base and a professional knowledge base are constructed.

[0068] Optionally, the device further includes:

[0069] The feedback unit is used to determine the contact person and contact information based on the work order after receiving the reply corresponding to the work order; and to send the reply corresponding to the work order back to the contact person using the contact information.

[0070] A work order processing device, comprising a memory and a processor;

[0071] The memory is used to store programs;

[0072] The processor is used to execute the program and implement the various steps of the work order processing method described above.

[0073] A readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the various steps of the work order processing method described above.

[0074] By employing the above technical solution, this application discloses a work order processing method, apparatus, device, and readable storage medium. After obtaining a work order to be processed, firstly, a first prompt message indicating the summary of the work order's content is determined. Then, based on the first prompt message, the summary content of the work order is obtained. Next, based on the summary content, a search is performed in a basic knowledge base to determine if there are any questions similar to the summary content of the work order, resulting in a first search result. Finally, based on the first search result, the work order is processed to obtain the corresponding response. Based on this solution, automated work order processing can be achieved, thereby enabling more efficient and higher-quality government affairs Q&A. Attached Figure Description

[0075] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the scope of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0076] Figure 1 This is a flowchart illustrating a work order processing method disclosed in an embodiment of this application;

[0077] Figure 2 This is a schematic diagram of the overall workflow for work order processing disclosed in an embodiment of this application;

[0078] Figure 3 This is a schematic diagram of the structure of a work order processing device disclosed in an embodiment of this application;

[0079] Figure 4 This is a hardware structure block diagram of a work order processing device disclosed in an embodiment of this application. Detailed Implementation

[0080] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0081] In government affairs Q&A scenarios, the traditional way of processing work orders is mainly to process them manually.

[0082] The manual processing flow for work orders is as follows: First, the content of the work order is summarized. Then, a basic knowledge base is searched based on the summary. If a similar resolved issue is found in the basic knowledge base, the corresponding answer is used to compile a response. If no similar resolved issue is found in the basic knowledge base, the work order is categorized by domain (including industry and commerce, education, medical care, social security, housing provident fund, housing and construction, taxation, etc.) to obtain the work order's domain category. The work order is then assigned to the corresponding government department. Staff of that government department search the professional knowledge base. If a similar resolved issue is found in the professional knowledge base, the corresponding answer is used to compile a response. If no similar resolved issue is found in the department's professional knowledge base, government documents are consulted to compile the response. The quality of the manually summarized work order content is often unsatisfactory, resulting in low search accuracy and affecting the efficiency and effectiveness of work order processing. Therefore, the traditional manual work order processing method is time-consuming, labor-intensive, and its effectiveness is greatly affected by the individual's skill level.

[0083] To address the problems of traditional manual work order processing being time-consuming, labor-intensive, and highly dependent on human ability in terms of effectiveness, the inventors of this case conducted research and discovered that:

[0084] In recent years, with the rapid development of AI (Artificial Intelligence) technology, AI technology has also been applied in the field of government affairs. In government Q&A scenarios, there are currently two main ways of processing work orders involving AI technology, as follows:

[0085] The first method is a rule-based work order processing approach.

[0086] This method mainly relies on grammatical rules, shallow semantic parsing, and vocabulary resources to mine citizens' intentions and extract key information from work orders, and finally query business rules to perform various operations and responses.

[0087] The second method is a work order processing approach based on deep learning.

[0088] This method primarily involves a separate work order summary model that outputs a summary of the work orders; then, a separate retrieval model searches the basic knowledge base based on this summary. If the search is successful, an answer is prepared and provided; if the search fails, a separate work order classification model is used to classify the work orders and dispatch them to the relevant government departments. Government staff at these departments then use the retrieval model to search the professional knowledge base. If the search is successful, an answer is prepared and provided; if the search fails, government documents are consulted to prepare an answer and provide a response.

[0089] However, both of the above methods have certain problems. The first method is highly customized but has poor generalization, cannot handle situations outside the rules, and the returned response is not stable enough. The second method involves the serial operation of multiple models and is interspersed with a lot of manual operation, resulting in a low model fault tolerance. In addition, work orders generally have characteristics such as long dialogue content (more than 50% of work orders exceed 1024 characters) and inaccurate telephone voice recognition (the same transaction may be recognized inconsistently, such as recognizing "Xiao Li" as "Xiao Li"). Using traditional deep learning methods to summarize the content of work orders often yields unsatisfactory summaries, resulting in low retrieval accuracy and thus affecting the efficiency and effectiveness of work order processing.

[0090] Therefore, the current work order processing methods cannot achieve more efficient and higher-quality government affairs Q&A.

[0091] In view of the above problems, the inventors of this case finally proposed a work order processing method that can achieve more efficient and higher quality government affairs Q&A.

[0092] The following embodiments will be used to describe the work order processing method provided in this application.

[0093] Reference Figure 1 , Figure 1 This is a flowchart illustrating a work order processing method disclosed in an embodiment of this application. The method can be applied to terminal devices with work order processing capabilities (such as smartphones, computers, robots, etc., with work order processing programs installed). The method may include the following steps:

[0094] Step S101: Obtain the work orders to be processed.

[0095] In this application, once the public completes a communication through the government hotline, a work order can be generated based on the communication data. There can be one or more work orders pending processing.

[0096] Step S102: Determine the first prompt information, which is used to indicate that the content of the work order be summarized.

[0097] In this application, the first prompt information may be any one or more of audio information, text information, or image information, and this application does not impose any limitations on this.

[0098] As one possible implementation, information in a fixed format can be preset, and this preset fixed format information includes preset slots. In this application, the preset fixed format information can be obtained, and the content of the work order to be processed can be embedded in the preset slots in the preset fixed format information to obtain the first prompt information.

[0099] For ease of understanding, for example, the first prompt message could be: "Please summarize the following work order content, including key information from the work order content, and no more than 64 characters. The work order content is..."<A:XXXXX,B:XXXX,…> "in,"<A:XXXXX,B:XXXX,…> "This refers to the content embedded in the preset slot."

[0100] Step S103: Based on the first prompt information, obtain the summary content of the work order.

[0101] For ease of understanding, let's assume the first prompt is: "Please summarize the following work order content, including key information from the work order, in no more than 64 characters. The work order content is..."<A:XXXXX,B:XXXX,…> If the summary content of the work order is "", then the summary content of the work order can be a text of no more than 64 characters.

[0102] It should be noted that the specific implementation method of this step will be described in detail through the following embodiments, and will not be described here.

[0103] Step S104: Based on the summary content of the work order, search the basic knowledge base to see if the first question exists, and obtain the first search result.

[0104] In this application, the first question is a question similar to the summary content of the work order in the basic knowledge base. The basic knowledge base is constructed based on the processing results of the resolved work orders and contains multiple question-answer pairs, each of which includes a question and a corresponding answer.

[0105] In this application, the work order processing program can retrieve a first search result from a basic knowledge base based on the summary content of the work order. There are two types of first search results: one indicates that a similar resolved problem has been found in the basic knowledge base, matching the summary content of the work order; the other indicates that no similar resolved problem has been found in the basic knowledge base.

[0106] Step S105: Based on the first search result, process the work order to obtain the corresponding reply.

[0107] In this application, different first search results determine the processing of the work order and the way to obtain the corresponding response to the work order. The specific details will be explained in detail through the following embodiments, and will not be elaborated here.

[0108] It should be noted that after receiving a response to the work order, a contact person and contact information can be determined based on the work order, and the response can be sent back to the contact person using the contact information. The contact person can be a member of the public, and the contact information can be telephone, email, etc.; this application does not impose any limitations on this.

[0109] This application discloses a work order processing method, apparatus, device, and readable storage medium. After obtaining a work order to be processed, firstly, a first prompt message is determined to indicate the processing of the work order by summarizing its content. Then, based on the first prompt message, the summary content of the work order is obtained. Next, based on the summary content, a search is performed in a basic knowledge base to determine if there are any questions similar to the summary content of the work order, resulting in a first search result. Finally, based on the first search result, the work order is processed to obtain the corresponding response. Based on this scheme, automated work order processing can be achieved, thereby enabling more efficient and higher-quality government affairs Q&A.

[0110] The inventors of this case discovered that large models (with hundreds of billions of parameters or more) possess massive knowledge learning and generalization capabilities. Applying them to the work order processing workflow can help achieve more efficient and higher-quality government Q&A.

[0111] Therefore, in this application, when obtaining the summary content of the work order based on the first prompt information, a large model can be used.

[0112] As one possible implementation method, in this application, the first prompt information can be input into a large model, the large model generates reply information based on the first prompt information, and the reply information generated by the large model is obtained as the summary content of the work order. The large model is a large model that is currently in mature application, such as the PaLM (Pathways Language Model), the Pangu large model, and the Xinghuo cognitive large model.

[0113] While large-scale models possess massive knowledge learning and generalization capabilities, they may not achieve satisfactory generation results for domains they haven't explored before. Therefore, the quality of work order summaries obtained using currently mature large-scale models is often unsatisfactory. This application identifies training work orders and corresponding summary content, and trains a currently mature large-scale model to improve the quality of the work order summaries obtained using this model. It should be noted that when training the currently mature large-scale model using training work orders and corresponding summary content, the loss function can be LM loss.

[0114] Alternatively, the first prompt information can be input into a large model, which generates a response based on the first prompt information. The response generated by the large model is then used as the summary content of the work order. The large model is a large model trained using the training work orders and the corresponding summary content.

[0115] In this embodiment, since the large model can alleviate the problem of poor quality of summary content caused by long work order content or errors to a certain extent, it can improve the quality of summary content and lay the foundation for improving the effect of subsequent processing steps.

[0116] As mentioned in the above embodiments, different first search results determine the processing of the work order and the way the corresponding response is obtained. In another embodiment of this application, a specific implementation of step 105, which processes the work order based on the first search result and obtains the corresponding response, is described in detail. This implementation is applicable when the first search result is "yes". A "yes" first search result indicates that a resolved problem similar to the summary content of the work order has been found in the basic knowledge base. This implementation may include the following steps:

[0117] Step S201: Obtain the answer to the first question.

[0118] Step S202: Determine the second prompt information, which is used to instruct the answer to the first question to be rewritten.

[0119] In this application, the second prompt information may be any one or more of audio information, text information, or image information, and this application does not impose any limitations on this.

[0120] As one possible implementation, information in a fixed format can be preset, and this fixed format information includes preset slots. In this application, the preset fixed format information can be obtained, and answers to solved problems similar to the summary content of the work order can be embedded in the preset slots of the preset fixed format information to obtain the second prompt information.

[0121] For ease of understanding, for example, the second prompt message can be "Please rewrite the following answer to meet the response format requirements of the work order. The answer to be rewritten is <XXXX>", where "<XXXX>" is the answer corresponding to the solved problem in the basic knowledge base embedded in the preset slot that is similar to the summary content of the work order.

[0122] Step S202: Based on the second prompt information, obtain a response to the work order.

[0123] For ease of understanding, for example, suppose the second prompt message can be "Please rewrite the following answer to meet the requirements of the work order reply format, the answer to be rewritten is <XXXX>", then the reply to the work order is text that meets the requirements of the work order reply format.

[0124] The inventors of this case discovered that a large model can simultaneously accommodate tens of thousands of downstream tasks. Based on this, in this application, when a response to the work order is obtained based on the second prompt information, the large model can also be used.

[0125] As one possible implementation method, in this application, the second prompt information can be input into the large model, the large model generates reply information based on the second prompt information, and the reply information generated by the large model is obtained as the reply corresponding to the work order. The large model is a large model that is currently in mature application, such as the PaLM (Pathways Language Model), the Pangu large model, and the Xinghuo cognitive large model.

[0126] While large-scale models possess massive knowledge learning and generalization capabilities, they may not achieve satisfactory generation results for domains they haven't explored before. Therefore, the quality of responses to work orders obtained using these mature models is often unsatisfactory. This application addresses this by identifying answers from a training knowledge base and their corresponding rewritten text, and then training the mature large-scale model to improve the quality of responses. It should be noted that when training the mature large-scale model using answers from the training knowledge base and their corresponding rewritten text, the loss function can be LM loss.

[0127] Alternatively, the second prompt information can be input into a large model, which generates a response based on the second prompt information. The response generated by the large model is then used as the response to the work order. The large model is a large model trained using the answers in the training knowledge base and the corresponding rewritten text.

[0128] As mentioned in the above embodiments, different first search results determine the processing of the work order and the way the corresponding response is obtained. In another embodiment of this application, a more detailed description is given of another specific implementation method for step 105, which processes the work order based on the first search result to obtain the corresponding response. This implementation method is applicable when the first search result is negative. A negative first search result indicates that no resolved issues similar to the summary content of the work order were found in the basic knowledge base. This implementation method may include the following steps:

[0129] Step S301: Determine the third prompt information, which is used to instruct the work order to be classified by domain.

[0130] In this application, the third prompt information can be any one or more of audio information, text information, or image information, and this application does not impose any limitations on it.

[0131] As one possible implementation, information in a fixed format can be preset, and this preset fixed format information includes preset slots. In this application, the preset fixed format information can be obtained, and the content of the work order to be processed can be embedded in the preset slots in the preset fixed format information to obtain the third prompt information.

[0132] For ease of understanding, for example, the third prompt message could be: "Please determine the domain category of the following work orders within the following domain category range, which includes domain category 1, domain category 2, ..., domain category n. The content of the work order is..."<A:XXXXX,B:XXXX,…> "in,"<A:XXXXX,B:XXXX,…> "This refers to the content of the work order to be processed embedded in the preset slot."

[0133] Step S302: Based on the third prompt information, obtain the domain category of the work order.

[0134] For ease of understanding, exemplarily, suppose the third prompt message can be: "Please determine the domain category of the following work orders within the following domain category range, including domain category 1, domain category 2, ..., domain category n. The content of the work order is..."<A:XXXXX,B:XXXX,…> If the work order is classified as "", then the domain category of the work order can be domain category 2.

[0135] The inventors of this case discovered that a large model can simultaneously accommodate tens of thousands of downstream tasks. Based on this, in this application, when obtaining the domain category of the work order based on the third prompt information, the large model can also be used.

[0136] As one possible implementation method, in this application, the third prompt information can be input into the large model, the large model generates response information based on the third prompt information, and the response information generated by the large model is obtained as the domain category of the work order. The large model is a large model that is currently in mature application, such as the PaLM (Pathways Language Model), the Pangu large model, and the Xinghuo cognitive large model.

[0137] While large-scale models possess massive knowledge learning and generalization capabilities, they may not achieve satisfactory generation results for domains they haven't encountered before. Therefore, the quality of domain categories for work orders obtained using currently mature large-scale models is often unsatisfactory. This application identifies the work orders used for training and their corresponding domain categories, and trains the currently mature large-scale model, thereby improving the quality of domain categories for work orders obtained using the large-scale model. It should be noted that when training the currently mature large-scale model using the training work orders and their corresponding domain categories, the loss function can be LM loss.

[0138] Alternatively, the third prompt information can be input into a large model, which generates a response based on the third prompt information. The response generated by the large model is then used as the domain category of the work order. The large model is a large model trained using training work orders and corresponding domain categories.

[0139] Step S303: Search the professional knowledge base corresponding to the domain category of the work order to see if there is a second question, the second question being a question similar to the work order in the professional knowledge base, and obtain a second search result.

[0140] In this application, the professional knowledge base is constructed based on the processing results of work orders that have been resolved by this government agency. It contains multiple question-answer pairs, each of which includes a question and its corresponding answer.

[0141] In this application, the work order processing program can retrieve a professional knowledge base corresponding to the domain category of the work order to obtain a second search result. There are two types of second search results: one indicates that a similar resolved issue was found in the professional knowledge base, and the other indicates that no similar resolved issue was found in the professional knowledge base.

[0142] Step S304: Based on the second search result, process the work order to obtain the corresponding reply.

[0143] In this application, different second search results determine the processing of the work order and the way the corresponding response to the work order is obtained. The specific details will be explained in detail through the following embodiments, and will not be described here.

[0144] As mentioned in the above embodiments, different second search results determine the processing of the work order and the way to obtain the corresponding response. In another embodiment of this application, a specific implementation of step S304, which processes the work order based on the second search result and obtains the corresponding response, is described in detail. This implementation is applicable when the second search result is "yes," indicating that a similar resolved issue has been found in the professional knowledge base. This implementation may include the following steps:

[0145] Step S401: Obtain the answer to the second question;

[0146] Step S402: Determine the fourth prompt information, which is used to instruct the answer to the second question to be rewritten.

[0147] In this application, the fourth prompt information can be any one or more of audio information, text information, or image information, and this application does not impose any limitations on it.

[0148] As one possible implementation, information in a fixed format can be preset, and this fixed format information includes preset slots. In this application, the preset fixed format information can be obtained, and answers to solved problems similar to the summary content of the work order can be embedded in the preset slots of the preset fixed format information to obtain the fourth prompt information.

[0149] For ease of understanding, for example, the fourth prompt message can be "Please rewrite the following answer to meet the response format requirements of the work order. The answer to be rewritten is <XXXX>". Here, "<XXXX>" is the answer corresponding to the solved problem in the professional knowledge base embedded in the preset slot that is similar to the summary content of the work order.

[0150] Step S402: Based on the fourth prompt information, obtain the reply to the work order.

[0151] For ease of understanding, for example, suppose the fourth prompt message can be "Please rewrite the following answer to meet the work order's reply format requirements. The answer to be rewritten is <XXXX>". Then the work order's reply will be text that meets the work order's reply format requirements.

[0152] The inventors of this case discovered that a large model can simultaneously accommodate tens of thousands of downstream tasks. Based on this, in this application, when obtaining a response to the work order based on the fourth prompt information, the large model can also be used.

[0153] As one possible implementation method, in this application, the fourth prompt information can be input into the large model, the large model generates reply information based on the fourth prompt information, and the reply information generated by the large model is obtained as the reply corresponding to the work order. The large model is a large model that is currently in mature application, such as the PaLM (Pathways Language Model), the Pangu large model, and the Xinghuo cognitive large model.

[0154] Although large-scale models possess massive knowledge learning and generalization capabilities, they may not achieve satisfactory generation results for domains they haven't encountered before. Therefore, the quality of responses to work orders obtained using currently mature large-scale models is often unsatisfactory. Since the answers in the professional knowledge base have similar formats to those in the basic knowledge base, as an alternative implementation method, the fourth prompt information can be input into the large-scale model. The large-scale model generates response information based on this fourth prompt information, and the response information generated by the large-scale model is obtained as the response to the work order. The large-scale model is a model trained using answers from the basic knowledge base used for training, along with the corresponding rewritten text.

[0155] Although the answers in the professional knowledge base are similar in format to those in the basic knowledge base, subtle differences may exist. These differences could lead to imperfect quality in the responses to work orders obtained by the large model trained using the answers from the basic knowledge base and the corresponding rewritten text. Therefore, this application identifies the answers from the professional knowledge base used for training and uses the corresponding rewritten text to train the already mature and widely used large model, thereby improving the quality of the responses to work orders obtained by the large model. It should be noted that when training the already mature and widely used large model using the answers from the professional knowledge base used for training and the corresponding rewritten text, the loss function can be LM loss.

[0156] Alternatively, the fourth prompt information can be input into a large model, which generates a response based on the fourth prompt information. The response generated by the large model is then used as the response to the work order. The large model is a large model trained using the answers in the professional knowledge base used for training and the corresponding rewritten text.

[0157] As mentioned in the above embodiments, different second search results determine the processing of the work order and the way the corresponding response is obtained. In another embodiment of this application, a more detailed description is given of another specific implementation method for step S304, which processes the work order based on the second search result and obtains the corresponding response. This implementation method is applicable when the second search result is negative, indicating that no similar resolved issues as the work order were found in the professional knowledge base. This implementation method may include the following steps:

[0158] Step S501: Determine the fifth prompt information, which is used to instruct the generation of a question-and-answer pair corresponding to the work order based on the government document.

[0159] In this application, the fifth prompt information may be any one or more of audio information, text information, or image information, and this application does not impose any limitations on it.

[0160] As one possible implementation, information in a fixed format can be preset, and the information in the fixed format includes preset slots. In this application, the information in the fixed format can be obtained, and government documents can be embedded in the preset slots in the information in the fixed format to obtain the fifth prompt information.

[0161] For ease of understanding, for example, the fifth prompt could be: "Please generate a question-and-answer pair based on the following paragraph or article, requiring the question to be relevant to the paragraph and the answer to the corresponding question. The paragraph is..."<XXX…> "Among them, "<XXXX>" refers to government documents embedded in the preset slot.

[0162] Step S502: Based on the fifth prompt information, obtain the question-and-answer pair corresponding to the work order generated from the government document.

[0163] For ease of understanding, let's assume, for example, that the fifth prompt could be: "Please generate a question-and-answer pair based on the following paragraph or article, requiring the question to be relevant to the paragraph and the answer to the corresponding question. The paragraph is..."<XXX…> Then, the question-and-answer pair corresponding to the work order can be Q: XXXX; A: XXXX.

[0164] The inventors of this case discovered that a large model can simultaneously accommodate tens of thousands of downstream tasks. Based on this, in this application, when obtaining the question-and-answer pair corresponding to the work order generated from the government document based on the fifth prompt information, it can also be achieved with the help of a large model.

[0165] As one possible implementation method, in this application, the fifth prompt information can be input into the big model, the big model generates response information based on the fifth prompt information, and the response information generated by the big model is obtained as a question-and-answer pair corresponding to the work order generated based on the government document. The big model is a big model that is currently in mature application, such as the PaLM (Pathways Language Model), the Pangu Big Model, and the Xinghuo Cognitive Big Model.

[0166] While large-scale models possess massive knowledge learning and generalization capabilities, they may not achieve satisfactory generation results for domains they haven't encountered before. Therefore, the quality of question-and-answer pairs generated from government documents using currently mature large-scale models is often unsatisfactory. This application identifies training work orders, government documents, and question-and-answer pairs generated from these documents to train a currently mature large-scale model, thereby improving the quality of the question-and-answer pairs generated from government documents using this model. It should be noted that when training the currently mature large-scale model using training work orders, government documents, and question-and-answer pairs generated from these documents, the loss function can be LM loss.

[0167] As another possible implementation, the fifth prompt information can be input into the large model, and the large model can generate response information based on the fifth prompt information. The response information generated by the large model can be obtained as a question-and-answer pair corresponding to the work order generated based on the government document. The large model is a large model trained using the training work order, the government document, and the question-and-answer pair corresponding to the training work order generated based on the government document.

[0168] Step S503: Obtain the answer from the question-and-answer pair corresponding to the work order;

[0169] Step S504: Determine the sixth prompt information, which is used to instruct the answer in the question-and-answer pair corresponding to the work order to be rewritten.

[0170] In this application, the sixth prompt information can be any one or more of audio information, text information, or image information, and this application does not impose any limitations on it.

[0171] As one possible implementation, information in a fixed format can be preset, and this preset fixed format information includes preset slots. In this application, the preset fixed format information can be obtained, and the answer in the question-and-answer pair corresponding to the work order generated based on the government document can be embedded in the preset slot of the preset fixed format information to obtain the sixth prompt information.

[0172] For ease of understanding, for example, the sixth prompt message can be "Please rewrite the following answer to meet the response format requirements of the work order. The answer to be rewritten is <XXXX>", where "<XXXX>" is the answer in the question-and-answer pair corresponding to the work order, which is generated based on the government document and embedded in the preset slot.

[0173] Step S505: Based on the sixth prompt information, obtain the reply to the work order.

[0174] For ease of understanding, for example, suppose the sixth prompt message can be "Please rewrite the following answer to meet the requirements of the work order reply format, the answer to be rewritten is <XXXX>", then the reply to the work order is text that meets the requirements of the work order reply format.

[0175] The inventors of this case discovered that a large model can simultaneously accommodate tens of thousands of downstream tasks. Based on this, in this application, when obtaining a response to the work order based on the sixth prompt information, the large model can also be used.

[0176] As one possible implementation method, in this application, the sixth prompt information can be input into the large model, the large model generates reply information based on the sixth prompt information, and the reply information generated by the large model is obtained as the reply corresponding to the work order. The large model is a large model that is currently in mature application, such as the PaLM (Pathways Language Model), the Pangu large model, and the Xinghuo cognitive large model.

[0177] Although large-scale models possess massive knowledge learning and generalization capabilities, they may not achieve satisfactory generation results for domains they haven't encountered before. Therefore, the quality of responses to work orders obtained using currently mature large-scale models is often unsatisfactory. Since the formats of answers in the professional knowledge base, the basic knowledge base, and the answers in the question-and-answer pairs generated from government documents corresponding to the work order are similar, as an alternative implementation method, the sixth prompt information can be input into the large-scale model. The large-scale model generates response information based on the sixth prompt information, and the response information generated by the large-scale model is obtained as the response to the work order. The large-scale model can be a large-scale model trained using answers from the basic knowledge base used for training and corresponding rewritten text, or a large-scale model trained using answers from the professional knowledge base used for training and corresponding rewritten text.

[0178] Although the answers in the professional knowledge base, the basic knowledge base, and the answers in the question-and-answer pairs generated from government documents corresponding to the work order have similar formats, subtle differences may exist. These differences could lead to imperfect quality of the work order responses obtained using either the answers from the basic knowledge base used for training, or the large model trained using the answers from the professional knowledge base used for training, and the corresponding rewritten text. Therefore, this application determines that the answers in the question-and-answer pairs generated from government documents corresponding to the work order, along with the corresponding rewritten text, are used to train the currently mature and widely used large model, thereby improving the quality of the work order responses obtained using the large model. It should be noted that when training the currently mature and widely used large model using the answers in the question-and-answer pairs generated from government documents corresponding to the work order, along with the corresponding rewritten text, the loss function can be LM loss.

[0179] Alternatively, the sixth prompt information can be input into a large model, which generates a response based on the sixth prompt information. The response generated by the large model is then used as the response to the work order. The large model is a large model trained using the answers in the question-and-answer pairs corresponding to the work order generated from government documents and the corresponding rewritten text.

[0180] To better understand the technical solution of this application, refer to Figure 2 , Figure 2 This is a schematic diagram of the overall process of work order processing disclosed in an embodiment of this application.

[0181] In traditional solutions, both the basic knowledge base and the professional knowledge base need to be built manually, which is time-consuming, labor-intensive, and its effectiveness is greatly affected by human capabilities. Therefore, this application also discloses an implementation method for building the basic knowledge base and the professional knowledge base. Before obtaining the work orders to be processed, the basic knowledge base and the professional knowledge base can be built using this method. This implementation method may include the following steps:

[0182] Step S601: Determine the seventh prompt information, which is used to instruct the government document to be processed and generate multiple question-answer pairs, wherein each question-answer pair includes a question and a corresponding answer, and the questions in each question-answer pair are all related to the government document.

[0183] In this application, the first prompt information may be any one or more of audio information, text information, or image information, and this application does not impose any limitations on this.

[0184] As one possible implementation, information in a fixed format can be preset, and this preset fixed format information includes preset slots. In this application, the preset fixed format information can be obtained, and the content of the work order to be processed can be embedded in the preset slots in the preset fixed format information to obtain the first prompt information.

[0185] For ease of understanding, for example, the seventh prompt could be: "Please generate several question-and-answer pairs based on the following paragraphs or articles, requiring that the questions are related to the paragraphs and that the answers can address the corresponding questions. The paragraphs are..."<XXX…> "in,"<XXX…> "This refers to the content embedded in the preset slot."

[0186] Step S602: Based on the seventh prompt information, obtain multiple question-answer pairs.

[0187] For ease of understanding, let's assume, for example, that the seventh prompt could be: "Please generate several question-and-answer pairs based on the following paragraphs or articles, requiring that the questions are related to the paragraphs and that the answers can address the corresponding questions. The paragraphs are..."<XXX…> "in,"<XXX…> Then, multiple question-answer pairs can be Q1: XXXX; A1: XXXX; Q2: XXXX; A2: XXXX; ...

[0188] Considering the massive knowledge learning capacity of large models, in this application, when multiple question-answer pairs are obtained based on the seventh prompt information, it can also be achieved with the help of large models.

[0189] As one possible implementation method, in this application, the seventh prompt information can be input into the large model, the large model generates response information based on the seventh prompt information, and the response information generated by the large model is obtained as multiple question-and-answer pairs generated based on government documents. The large model is a large model that is currently in mature application, such as the PaLM (Pathways Language Model), the Pangu large model, and the Xinghuo cognitive large model.

[0190] While large-scale models possess massive knowledge learning and generalization capabilities, they may not achieve satisfactory generation results for domains they haven't encountered before. Therefore, the quality of multiple question-answer pairs generated from government documents using currently mature large-scale models is often unsatisfactory. This application identifies training government documents and their corresponding question-answer pairs, and trains a currently mature large-scale model, thereby improving the quality of the question-answer pairs generated from these models. It should be noted that when training the currently mature large-scale model using training government documents and their corresponding question-answer pairs, the loss function can be LM loss.

[0191] As another possible implementation, the seventh prompt information can be input into the large model, and the large model can generate response information based on the seventh prompt information. The response information generated by the large model can be obtained as multiple question-and-answer pairs generated based on government documents. The large model is a large model trained using training government documents and multiple question-and-answer pairs corresponding to the training government documents.

[0192] Step S603: Utilize the multiple question-answer pairs to construct a basic knowledge base and a professional knowledge base.

[0193] After obtaining multiple question-answer pairs, the basic knowledge base and professional knowledge base can be constructed using these multiple question-answer pairs.

[0194] In this embodiment, since the question-answer pair generation capability of the large model far exceeds that of ordinary deep learning models, the quality of the basic knowledge base and professional knowledge base built based on the capabilities of the large model will also be greatly improved.

[0195] In traditional solutions, updating the basic knowledge base and professional knowledge base also requires manual construction, which is time-consuming, labor-intensive, and its effectiveness is greatly affected by human capabilities. Therefore, this application also discloses an implementation method for updating the basic knowledge base and professional knowledge base. This implementation method is applied to updating the basic knowledge base and / or the professional knowledge base by generating a question-and-answer pair corresponding to the work order based on the fifth prompt information and then using the question-and-answer pair corresponding to the work order.

[0196] By updating the basic knowledge base and / or the professional knowledge base, it is possible to further ensure more efficient and higher-quality government affairs Q&A.

[0197] Based on the above embodiments, only one large model is needed to complete the processing of work orders, thus enabling more efficient and higher-quality government affairs Q&A.

[0198] It should also be noted that, in the above embodiments, the training of the large model can be performed simultaneously or separately. When training the large model, training data can be constructed by concatenating several input-output examples into the input, so that the large model can learn a hidden paradigm that maps from the input to the output.

[0199] The work order processing apparatus disclosed in the embodiments of this application is described below. The work order processing apparatus described below can be referred to in correspondence with the work order processing method described above.

[0200] Reference Figure 3 , Figure 3 This is a schematic diagram of a work order processing device disclosed in an embodiment of this application. Figure 3 As shown, the work order processing device may include:

[0201] Work order acquisition unit 11 is used to acquire work orders to be processed;

[0202] The first prompt information determining unit 12 is used to determine the first prompt information, which is used to indicate the summary of the content of the work order;

[0203] Content summary unit 13 is used to obtain the summary content of the work order based on the first prompt information;

[0204] The first retrieval unit 14 is used to search the basic knowledge base for a first question based on the summary content of the work order, wherein the first question is a question in the basic knowledge base that is similar to the summary content of the work order, and to obtain a first retrieval result.

[0205] The processing unit 15 is used to process the work order according to the first search result and obtain the reply corresponding to the work order.

[0206] Optionally, if the first search result is yes, then the processing unit includes:

[0207] The first answer acquisition unit is used to acquire the answer corresponding to the first question;

[0208] The second prompt information determining unit is used to determine the second prompt information, which is used to instruct the answer corresponding to the first question to be rewritten;

[0209] The first work order response determination unit is used to obtain a response to the work order based on the second prompt information.

[0210] Optionally, if the first search result is negative, the processing unit includes:

[0211] The third prompt information determination unit is used to determine the third prompt information, which is used to indicate the domain classification of the work order.

[0212] The domain category determination unit is used to determine the domain category of the work order based on the third prompt information;

[0213] The second retrieval unit is used to search the professional knowledge base corresponding to the domain category of the work order to see if there is a second question, the second question being a question similar to the work order in the professional knowledge base, and to obtain a second retrieval result;

[0214] The second work order response determination unit is used to process the work order based on the second search result to obtain the response corresponding to the work order.

[0215] Optionally, if the second search result is yes, then the second work order response confirmation unit includes:

[0216] The second answer acquisition unit is used to acquire the answer to the second question.

[0217] The fourth prompt information determining unit is used to determine the fourth prompt information, which is used to instruct the answer corresponding to the second question to be rewritten;

[0218] The third work order response confirmation unit is used to obtain a response to the work order based on the fourth prompt information.

[0219] Optionally, if the second search result is negative, the second work order response confirmation unit includes:

[0220] The fifth prompt information determination unit is used to determine the fifth prompt information, which is used to instruct the generation of a question-and-answer pair corresponding to the work order based on the government document;

[0221] The question-answer pair determination unit is used to obtain the question-answer pair corresponding to the work order based on the fifth prompt information;

[0222] The question-and-answer pair answer acquisition unit is used to acquire the answer from the question-and-answer pair corresponding to the work order;

[0223] The sixth prompt information determining unit is used to determine the sixth prompt information, which is used to instruct the answer in the question-answer pair corresponding to the work order to be rewritten;

[0224] The fourth work order response confirmation unit is used to obtain a response to the work order based on the sixth prompt information.

[0225] Optionally, the device further includes:

[0226] The knowledge base update unit is used to update the basic knowledge base and / or the professional knowledge base using the question-and-answer pair corresponding to the work order after generating the question-and-answer pair corresponding to the work order based on the fifth prompt information.

[0227] Optionally, the device further includes:

[0228] The knowledge base construction unit is used to determine a seventh prompt message before obtaining the work order to be processed. The seventh prompt message is used to instruct the processing of government documents and generate multiple question-answer pairs, wherein each question-answer pair includes a question and a corresponding answer, and the questions in each question-answer pair are all related to the government documents; based on the seventh prompt message, multiple question-answer pairs are obtained; and using the multiple question-answer pairs, a basic knowledge base and a professional knowledge base are constructed.

[0229] Optionally, the device further includes:

[0230] The feedback unit is used to determine the contact person and contact information based on the work order after receiving the reply corresponding to the work order; and to send the reply corresponding to the work order back to the contact person using the contact information.

[0231] Reference Figure 4 , Figure 4 A hardware structure block diagram of a work order processing device provided in this application embodiment is shown below. Figure 4 The hardware structure of the work order processing device may include: at least one processor 1, at least one communication interface 2, at least one memory 3, and at least one communication bus 4.

[0232] In this embodiment of the application, the number of processor 1, communication interface 2, memory 3, and communication bus 4 is at least one, and processor 1, communication interface 2, and memory 3 communicate with each other through communication bus 4;

[0233] Processor 1 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0234] Memory 3 may include high-speed RAM, and may also include non-volatile memory, such as at least one disk storage device;

[0235] The memory stores a program, which the processor can call. The program is used for:

[0236] Get the work orders to be processed;

[0237] A first prompt message is determined, which is used to instruct the summary of the contents of the work order;

[0238] Based on the first prompt information, the summary content of the work order is obtained;

[0239] Based on the summary content of the work order, a search is conducted in the basic knowledge base to determine if a first question exists. The first question is a question in the basic knowledge base that is similar to the summary content of the work order, and a first search result is obtained.

[0240] Based on the first search result, the work order is processed to obtain the corresponding response.

[0241] Optionally, the refined and extended functions of the program can be found in the description above.

[0242] This application embodiment also provides a readable storage medium that can store a program suitable for execution by a processor, the program being used for:

[0243] Get the work orders to be processed;

[0244] A first prompt message is determined, which is used to instruct the summary of the contents of the work order;

[0245] Based on the first prompt information, the summary content of the work order is obtained;

[0246] Based on the summary content of the work order, a search is conducted in the basic knowledge base to determine if a first question exists. The first question is a question in the basic knowledge base that is similar to the summary content of the work order, and a first search result is obtained.

[0247] Based on the first search result, the work order is processed to obtain the corresponding response.

[0248] Optionally, the refined and extended functions of the program can be found in the description above.

[0249] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0250] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0251] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A work order processing method, characterized in that, The method includes: Get the work orders to be processed; A first prompt message is determined, which is used to instruct the summary of the contents of the work order; The first prompt information is input into the large model, and the large model generates reply information based on the first prompt information. The reply information generated by the large model is obtained as the summary content of the work order. The large model is a large model trained using the training work orders and the corresponding summary content. Based on the summary content of the work order, a search is conducted in the basic knowledge base to determine if a first question exists. The first question is a question in the basic knowledge base that is similar to the summary content of the work order, and a first search result is obtained. Based on the first search result, the work order is processed to obtain the corresponding response to the work order; Different first search results correspond to different ways of processing the work order to obtain the corresponding reply. If the first search result is yes, then processing the work order according to the first search result to obtain the corresponding reply includes: obtaining the answer to the first question; determining second prompt information, the second prompt information being used to instruct the answer to the first question to be rewritten; and obtaining the reply to the work order based on the second prompt information. If the first search result is no, then processing the work order according to the first search result to obtain the corresponding reply includes: determining third prompt information, the third prompt information being used to instruct the work order to be classified into different domains; obtaining the domain category of the work order based on the third prompt information; searching the professional knowledge base corresponding to the domain category of the work order to see if there is a second question, the second question being a question similar to the work order in the professional knowledge base, to obtain a second search result; and processing the work order according to the second search result to obtain the corresponding reply.

2. The method according to claim 1, characterized in that, If the second search result is yes, then processing the work order based on the second search result to obtain the corresponding response for the work order includes: Obtain the answer to the second question; A fourth prompt message is determined, which is used to instruct the answer to the second question to be rewritten; Based on the fourth prompt information, a response to the work order is obtained.

3. The method according to claim 1, characterized in that, If the second search result is negative, then processing the work order based on the second search result to obtain the corresponding response for the work order includes: The fifth prompt message is determined, which is used to instruct the generation of a question-and-answer pair corresponding to the work order based on the government document; Based on the fifth prompt information, a question-and-answer pair corresponding to the work order is obtained; Obtain the answer from the question-and-answer pair corresponding to the work order; A sixth prompt message is determined, which is used to instruct the rewriting of the answer in the question-and-answer pair corresponding to the work order; Based on the sixth prompt, a response to the work order is obtained.

4. The method according to claim 3, characterized in that, After generating the question-and-answer pair corresponding to the work order based on the fifth prompt information, the method further includes: The basic knowledge base and / or the professional knowledge base are updated using the question-and-answer pairs corresponding to the work order.

5. The method according to claim 1, characterized in that, Before obtaining the work order to be processed, the method further includes: The seventh prompt message is determined, which is used to instruct the processing of government documents and generate multiple question-and-answer pairs, wherein each question-and-answer pair includes a question and a corresponding answer, and the questions in each question-and-answer pair are all related to the government documents; Based on the seventh prompt information, multiple question-answer pairs are obtained; Using the aforementioned question-and-answer pairs, a basic knowledge base and a professional knowledge base are constructed.

6. The method according to any one of claims 1 to 5, characterized in that, After receiving the response corresponding to the work order, the method further includes: Based on the work order, the contact person and contact information are determined; The corresponding reply for the work order is sent back to the contact person using the aforementioned contact method.

7. A work order processing device, characterized in that, The device includes: The work order acquisition unit is used to acquire work orders to be processed. The first prompt information determining unit is used to determine the first prompt information, which is used to indicate the summary of the content of the work order; The content summary unit is used to input the first prompt information into the large model, the large model generates reply information based on the first prompt information, and obtains the reply information generated by the large model as the summary content of the work order. The large model is a large model trained using the training work order and the corresponding summary content. The first retrieval unit is used to search the basic knowledge base for a first question based on the summary content of the work order, wherein the first question is a question in the basic knowledge base that is similar to the summary content of the work order, and to obtain a first retrieval result. A processing unit is configured to process the work order according to the first search result to obtain a response corresponding to the work order; wherein different first search results correspond to different ways of processing the work order to obtain a response corresponding to the work order; if the first search result is yes, then processing the work order according to the first search result to obtain a response corresponding to the work order includes: obtaining the answer corresponding to the first question; determining second prompt information, the second prompt information being used to instruct the answer corresponding to the first question to be rewritten; and obtaining a response to the work order based on the second prompt information; if the first search result is no, then processing the work order according to the first search result to obtain a response corresponding to the work order includes: determining third prompt information, the third prompt information being used to instruct the work order to be classified into different domains; obtaining the domain category of the work order based on the third prompt information; searching the professional knowledge base corresponding to the domain category of the work order to see if there is a second question, the second question being a question similar to the work order in the professional knowledge base, and obtaining a second search result; and processing the work order according to the second search result to obtain a response corresponding to the work order.

8. A work order processing device, characterized in that, Including memory and processor; The memory is used to store programs; The processor is used to execute the program to implement each step of the work order processing method as described in any one of claims 1 to 6.

9. A readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements each step of the work order processing method as described in any one of claims 1 to 6.

Images