Information processing method and device, equipment and medium

Abstract

The embodiment of the invention relates to an information processing method and device, equipment and a medium. The method comprises the steps that to-be-processed target customer complaint information is acquired; performing semantic understanding processing on the target customer complaint information to obtain a target simplified text corresponding to the target customer complaint information; performing key information extraction processing on the target customer complaint information to obtain key information in the target customer complaint information; generating a target query statement based on the target simplified text; and according to the target query statement and the key information, retrieving from a preset question classification library to obtain a target question category to which the target customer complaint information belongs. According to the mode, manual classification is not needed, the cost is greatly saved, the classification efficiency is improved, the key information in the target customer complaint information and the simplified text capable of presenting the semantics of the target customer complaint information are integrated for retrieval, and the accuracy of the question category obtained through retrieval is effectively guaranteed.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technologies, and in particular, to an information processing method, apparatus, device, and medium. Background Art

[0002] In service industries such as home improvement services, there may be a phenomenon that customers file complaints due to dissatisfaction with the services. The ways for customers to file complaints are relatively diverse. For example, customers may directly express dissatisfaction in a communication group, make a complaint call, or fill out a complaint work order through an application. In related technologies, after obtaining customer complaint information, the service provider will process the customer complaint information manually to determine the problem category to which the customer complaint information belongs, so as to provide a targeted solution subsequently. However, the existing method of manually identifying the problem category to which the customer complaint information belongs not only takes time and effort, has low efficiency, and requires high costs, but also there may be situations where the classification is inaccurate due to lack of professional knowledge, making it difficult to ensure the accuracy of the classification result. Summary of the Invention

[0003] In order to solve the above technical problems or at least partially solve the above technical problems, the present disclosure provides an information processing method, apparatus, device, and medium.

[0004] An embodiment of the present disclosure provides an information processing method, the method including: obtaining target customer complaint information to be processed; performing semantic understanding processing on the target customer complaint information to obtain a target simplified text corresponding to the target customer complaint information; and performing key information extraction processing on the target customer complaint information to obtain key information in the target customer complaint information; generating a target query statement based on the target simplified text; and retrieving, from a preset problem classification library according to the target query statement and the key information, a target problem category to which the target customer complaint information belongs.

[0005] Optionally, the performing semantic understanding processing on the target customer complaint information to obtain a target simplified text corresponding to the target customer complaint information includes: using a preset first network model to perform semantic understanding processing on the target customer complaint information to obtain at least one target simplified text corresponding to the target customer complaint information; the performing key information extraction processing on the target customer complaint information to obtain key information in the target customer complaint information includes: using a preset second network model to perform key information extraction processing on the target customer complaint information to obtain key information in the target customer complaint information; the key information includes one or more of a problem object, a problem manifestation, and a problem cause.

[0006] Optionally, generating a target query statement based on the target simplified text includes: obtaining a first query statement based on the target simplified text; generating multiple second query statements based on the first query statement by using at least one preset expression transformation strategy; and obtaining a target query statement based on the multiple second query statements.

[0007] Optionally, when there are multiple types of the expression transformation strategies, obtaining a target query statement based on the multiple second query statements includes: obtaining a mixing ratio corresponding to multiple types of the expression transformation strategies, and determining the number of statement extractions corresponding to multiple types of the expression transformation strategies respectively based on the mixing ratio; and extracting a target query statement from multiple second query statements corresponding to multiple types of the expression transformation strategies respectively based on the number of statement extractions.

[0008] Optionally, retrieving a target problem category to which the target customer complaint information belongs from a preset problem classification library according to the target query statement and the key information includes: retrieving multiple candidate problem categories corresponding to the target customer complaint information from the preset problem classification library by using a preset third network model according to the target query statement and the key information; and determining the target problem category to which the target customer complaint information belongs from the multiple candidate problem categories by using a preset fourth network model.

[0009] Optionally, determining the target problem category to which the target customer complaint information belongs from the multiple candidate problem categories by using a preset fourth network model includes: obtaining service information corresponding to the target customer complaint information; sorting the multiple candidate problem categories based on the service information to obtain the top N candidate problem categories with the highest relevance to the service information, where N is a preset integer value; and determining the target problem category to which the target customer complaint information belongs from the top N candidate problem categories by using a preset fourth network model.

[0010] Optionally, the problem classification library contains multiple problem category blocks, and each problem category block contains a problem category label and at least one simplified text example corresponding to the problem category label; the step of retrieving, from a preset problem classification library, multiple candidate problem categories corresponding to the target customer complaint information according to the target query statement and the key information includes: retrieving multiple candidate category blocks from the preset problem classification library according to the target query statement and the key information by using a preset third network model; wherein each target query statement corresponds to a specified number of candidate category blocks, and the candidate category blocks contain simplified text examples semantically related to the target query statement, and / or at least some information in the key information; and obtaining multiple candidate problem categories corresponding to the target customer complaint information based on the problem category labels in the candidate category blocks.

[0011] Optionally, the problem classification library is constructed through the following steps: obtaining customer complaint samples and problem category labels corresponding to the customer complaint samples; wherein the number of the customer complaint samples is multiple; performing a simplification process on the customer complaint samples to obtain simplified text examples of the customer complaint samples; and constructing a problem classification library based on the problem category labels and the simplified text examples corresponding to the customer complaint samples.

[0012] Optionally, the step of constructing a problem classification library based on the problem category labels and the simplified text examples corresponding to the customer complaint samples includes: obtaining data blocks corresponding to the customer complaint samples based on the problem category labels and the simplified text examples corresponding to the customer complaint samples; generating problem category blocks based on the data blocks respectively corresponding to the multiple customer complaint samples; wherein each problem category block contains a problem category label and at least one simplified text example corresponding to the problem category label; the problem category labels and / or the simplified text examples in different problem category blocks are different, and the character size of the problem category block is within a preset range; and constructing a problem classification library based on the problem category blocks.

[0013] Optionally, the method further includes: retrieving, from a preset solution library, a target solution corresponding to the target problem category based on the target problem category to which the target customer complaint information belongs.

[0014] Optionally, the method further includes: generating a customized processing suggestion corresponding to the target customer complaint information by using a preset fifth network model based on the target problem category and the target solution.

[0015] Optionally, the method further includes: when it is determined that the target problem category is inaccurate, obtaining the standard problem category to which the target customer complaint information belongs; wherein, the standard problem category is a problem category specified manually; and updating the problem classification library based on the standard problem category to which the target customer complaint information belongs and the target simplified text.

[0016] An information processing apparatus according to an embodiment of the present disclosure includes: an information acquisition module configured to acquire target customer complaint information to be processed; an information processing module configured to perform semantic understanding processing on the target customer complaint information to obtain a target simplified text corresponding to the target customer complaint information, and perform key information extraction processing on the target customer complaint information to obtain key information in the target customer complaint information; a statement generation module configured to generate a target query statement based on the target simplified text; and a category retrieval module configured to retrieve, from a preset problem classification library, a target problem category to which the target customer complaint information belongs according to the target query statement and the key information.

[0017] An electronic device according to an embodiment of the present disclosure includes: a processor; a memory for storing executable instructions executable by the processor; and the processor is configured to read the executable instructions from the memory and execute the instructions to implement the information processing method provided in the embodiment of the present disclosure.

[0018] A computer-readable storage medium according to an embodiment of the present disclosure stores a computer program, and the computer program is used to execute the information processing method provided in the embodiment of the present disclosure.

[0019] A computer program product according to an embodiment of the present disclosure includes a computer program, and the computer program, when executed by a processor, implements the information processing method provided in the embodiment of the present disclosure.

[0020] The above technical solution provided by the embodiment of the present disclosure can perform semantic understanding processing and key information extraction processing on the target customer complaint information to be processed, obtain a target simplified text corresponding to the target customer complaint information and key information in the target customer complaint information. On this basis, a target query statement can be generated using the target simplified text, and according to the target query statement and the key information, a target problem category to which the target customer complaint information belongs can be retrieved from a preset problem classification library. The above method not only eliminates the need for manual classification, greatly saving costs and improving classification efficiency, but also retrieves by comprehensively considering the key information in the target customer complaint information and the simplified text that can present the semantics of the target customer complaint information, effectively ensuring the accuracy of the retrieved problem category.

[0021] It should be understood that the content described in this part is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The accompanying drawings herein are incorporated into and constitute a part of this specification, showing embodiments consistent with the present disclosure and, together with the specification, are used to explain the principles of the present disclosure.

[0023] In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure or in the prior art, the following will briefly introduce the accompanying drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0024] Figure 1 It is a schematic flowchart of an information processing method provided by an embodiment of the present disclosure;

[0025] Figure 2 It is a schematic diagram of data block processing provided by an embodiment of the present disclosure;

[0026] Figure 3 It is a schematic flowchart of a method for constructing a problem classification library provided by an embodiment of the present disclosure;

[0027] Figure 4 It is a schematic flowchart of an information processing process provided by an embodiment of the present disclosure;

[0028] Figure 5 It is a schematic structural diagram of an information processing device provided by an embodiment of the present disclosure;

[0029] Figure 6 It is a schematic structural diagram of an electronic device provided by an embodiment of the present disclosure. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0030] In order to be able to more clearly understand the above objects, features, and advantages of the present disclosure, the following will further describe the solutions of the present disclosure. It should be noted that, without conflict, the embodiments of the present disclosure and the features in the embodiments can be combined with each other.

[0031] In the following description, many specific details are set forth in order to fully understand the present disclosure, but the present disclosure can also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only a part of the embodiments of the present disclosure, rather than all of the embodiments.

[0032] Figure 1The flowchart of an information processing method provided by an embodiment of the present disclosure. This method can be executed by an information processing device, which can be implemented by software and / or hardware and is generally integrated in an electronic device. As Figure 1 shown, the method mainly includes the following steps S102 to S108:

[0033] Step S102, obtain target customer complaint information to be processed. The present disclosure does not limit the presentation form of the target customer complaint information. For example, the target customer complaint information can be the conversation content between the customer and the service provider, or it can be a customer complaint work order submitted by the customer. The target customer complaint information contains the description content of the customer's dissatisfaction with the current business. The present disclosure does not limit the type of business. For example, the current business can be a home improvement business.

[0034] Step S104, perform semantic understanding processing on the target customer complaint information to obtain a target simplified text corresponding to the target customer complaint information; and perform key information extraction processing on the target customer complaint information to obtain the key information in the target customer complaint information.

[0035] It can be understood that the target customer complaint information provided by the user may contain relatively long content, which may also contain words that are ineffective for semantic understanding, such as stop words. In this embodiment of the present disclosure, the target customer complaint information can be first cleaned, and then semantic understanding processing can be performed based on the cleaned target customer complaint information to obtain the target simplified text. The target simplified text can be a summary text or a concise text of the target customer complaint information, which can present the semantics of the target customer complaint information more concisely. In addition, in order to prevent some key information from being omitted when simplifying the target customer complaint information, the present disclosure can perform key information extraction processing on the basis of the target customer complaint information to obtain the key information. The key information includes, but is not limited to, one or more of the problem object, problem manifestation, and problem cause. Among them, the problem object is the object where the problem occurs. Taking the home improvement business as an example, it can indicate which main material or installation process has a problem, or indicate that the basic decoration construction (such as demolition / hydropower / woodworking / painting / others) has a problem. The problem manifestation can be understood as the form in which the problem appears, which is used to indicate what problems currently exist, such as construction period delay, garbage stacking, etc.; the problem cause is also the reason for the problem manifestation, such as untimely worker arrangement, extension without notice of the reason, failure to clean up in time, etc. In addition, the key information can also include information such as the person handling the problem, and information related to the business. For example, in the home improvement business, the key information can also include information such as the space where the problem is located, which is used to indicate the space where the problem occurs, such as the bedroom, kitchen, etc.

[0036] Step S106, generate a target query statement based on the target simplified text.

[0037] Embodiments of the present disclosure can construct a target query statement based on a target simplified text. The target query statement is the statement used as the retrieval basis. In some embodiments, when the target simplified text is a concise text with the number of characters within a preset range, the target simplified text can be directly used as the target query statement. In other embodiments, the target simplified text can be further processed. For example, the expression form of the target simplified text can be changed, or the target simplified text can be adjusted to a target query statement with a specified format. The specified format can be a format that is convenient for the large model used to retrieve problem categories from the problem classification library to process, or a format that matches the data format in the problem classification library, so as to facilitate more accurate and convenient retrieval based on the target query statement subsequently.

[0038] Step S108, retrieve the target problem category to which the target customer complaint information belongs from a preset problem classification library according to the target query statement and the key information. Exemplarily, the problem classification library can include multiple problem category blocks, and each problem category block includes a problem category label and at least one simplified text sample corresponding to the problem category label. The problem types included in the problem category labels can be flexibly set according to the actual business, such as including classification labels for indicating the problem object, classification labels for indicating the problem manifestation, classification labels for indicating the problem cause, classification labels for indicating the problem scenario (such as safety scenarios, quality scenarios, etc.), and the like, which are not limited herein. Correspondingly, the finally obtained target problem category can include a problem object category, a problem manifestation category, a problem cause category, a problem scenario category, etc.

[0039] It should be noted that the target query statement is constructed based on the simplified text of the target customer complaint information, so it presents the semantics of the target customer complaint information. The key information is directly extracted from the target customer complaint information. Embodiments of the present disclosure can perform retrieval based on two dimensions of the target query statement and the key information. Compared with retrieving using a single dimension, the retrieval accuracy can be greatly improved. It can be understood that the common retrieval method in the related art is to retrieve only based on keywords and does not analyze in combination with the specific context, which is prone to the phenomenon of inaccurate classification. Moreover, when a new scenario appears, if there is no appropriate keyword matching logic, it may lead to abnormal retrieval. However, embodiments of the present disclosure will introduce a target query statement that can present the semantics of the target customer complaint information. Even if the specific scenario changes, such as the place where the problem occurs changes from the bedroom to the kitchen, but the problem description is similar and the semantics are close, effective retrieval can still be performed. In addition, embodiments of the present disclosure will further combine the key information for retrieval on the basis of the target query statement, which can also effectively avoid the loss of key information in the target simplified text, thus affecting the retrieval result.

[0040] In summary, the above method provided by the embodiments of the present disclosure not only eliminates the need for manual classification, greatly saving costs and improving classification efficiency, but also retrieves by comprehensively considering the key information in the target customer complaint information and the simplified text that can present the semantics of the target customer complaint information, effectively ensuring the accuracy of the problem categories obtained by the retrieval.

[0041] In some embodiments, the step of performing semantic understanding processing on the target customer complaint information in step S104 to obtain the target simplified text corresponding to the target customer complaint information can be executed in the following manner: Use a preset first network model to perform semantic understanding processing on the target customer complaint information to obtain at least one target simplified text corresponding to the target customer complaint information. Embodiments of the present disclosure can utilize the strong processing ability of the network model and adopt the network model to perform semantic understanding processing on the target customer complaint information. Embodiments of the present disclosure do not limit the specific structure of the first network model. Exemplarily, the first network model can be a large language model. Additionally, for subsequent applications, one or more target simplified texts can be generated using the first network model. The character number ranges corresponding to different target simplified texts are different. For example, the target simplified text can be a summary text or a concise sentence. The number of first network models can be one or more. If there is one, different target simplified texts can be generated by modifying the model prompt words of the first network model. For example, a summary text and a concise sentence can be generated. If there are multiple, different first network models can be used to perform semantic understanding processing on the target customer complaint information respectively, so as to generate different target simplified texts. One or more of the structures, parameters, and prompt words of different first network models are different. It can be set flexibly specifically and is not limited here. If multiple target simplified texts are obtained, a target query statement can be comprehensively constructed based on the multiple target simplified texts, or a suitable text (such as a concise sentence) can be selected from the multiple target simplified texts to construct the target query statement, and other summary texts and other target simplified texts can be used as a reference for selecting the final retrieval result or provided to the user for reference.

[0042] For ease of understanding, assume that the target customer complaint information is an interactive conversation between the customer and the service provider. The communication content is rather lengthy and will not be shown here. After being simplified, the summarized text can be something like "The service provider advised the homeowner to view the cabinet plan through an online meeting in the evening, but the homeowner said they had something to do that day and couldn't attend, and asked about other times. The two parties finally agreed to hold the meeting at 9:30 am the next day. The homeowner asked for the meeting link before the appointed time but didn't get a timely response. After the appointed time, the homeowner found that the service provider didn't show up and couldn't contact the service provider by phone. The homeowner was dissatisfied with this, emphasizing that the project delay caused by the service provider not showing up on time was unacceptable, and demanded that the service provider make remedies as soon as possible." The streamlined statement can be something like "The service provider didn't attend the meeting as agreed and couldn't be contacted, causing dissatisfaction among the homeowners and demanding remedies for the project delay." Through the above method, the lengthy target customer complaint information can be simplified, making it more convenient for subsequent processing.

[0043] In some embodiments, for the step S104 of extracting key information from the target customer complaint information to obtain the key information in the target customer complaint information, the following method can be referred to: Use a preset second network model to extract key information from the target customer complaint information to obtain the key information in the target customer complaint information; the key information includes one or more of the problem object, problem manifestation, and problem cause. The key information can refer to the relevant descriptions above and will not be elaborated here. Similarly, the embodiments of the present disclosure do not limit the second network model, and the second network model can also be a large language model. The second network model can be the same as or different from the first network model. It should be emphasized that the embodiments of the present disclosure do not perform key information extraction on the basis of the target simplified text, but directly extract key information from the original target customer complaint information, so as to avoid the loss of key information due to the simplification of the target customer complaint information, thereby affecting problem judgment.

[0044] In some embodiments, for the above step S106, that is, the step of generating the target query statement based on the target simplified text, the following steps A to C can be referred to:

[0045] Step A, obtain the first query statement based on the target simplified text. In practical applications, the target simplified text can be directly used as the first query statement, or the target simplified text can be processed such as format change based on the data format in the problem classification library or the format convenient for processing by the retrieval model, so as to convert it into a first query statement with a specified format. In the case where there are multiple target simplified texts, the first query statement can be constructed by integrating multiple target simplified texts, or the most streamlined text among the multiple target simplified texts can be used as the first query statement, which is not limited here.

[0046] Step B: Based on the first query statement, generate multiple second query statements by using at least one preset expression transformation strategy.

[0047] The embodiments of the present disclosure do not limit the expression transformation strategy. Exemplarily, the number of expression transformation strategies is multiple. For example, multiple expression transformation strategies may include a first transformation strategy and a second transformation strategy. The first transformation strategy is to describe the first query statement from multiple perspectives to obtain multiple second query statements. For example, the first query statement is: The service provider did not attend the meeting at the agreed time and could not be contacted, resulting in dissatisfaction of the property owner and a request to remedy the project schedule delay. Then, the multiple second query statements constructed by the first transformation strategy may be: 1) The service provider did not attend the meeting on time and could not be contacted, and the property owner was dissatisfied and requested to remedy the project schedule delay. 2) Due to the service provider not attending at the agreed time and being unable to be contacted, the property owner expressed dissatisfaction and requested to make up for the project schedule delay. 3) The service provider did not attend the meeting on time and could not be contacted, resulting in dissatisfaction of the property owner and a request to accelerate the project schedule. 4) The property owner requested to remedy the project schedule loss caused by the delay due to the service provider not attending the meeting at the agreed time and being unable to be contacted. 5) The service provider did not attend the meeting on time and could not be contacted, and the property owner was dissatisfied and requested to remedy the project schedule problem caused by the delay. Among them, the first transformation strategy may be an extended query strategy. The second transformation strategy may be to construct query content based on the keywords in the first query statement, such as changing the keywords to synonyms, or adjusting the expression based on the relative relationship between multiple keywords. For example, if A is located on the left of B, it can be changed to B is located on the right of A, and a new statement is constructed based on multiple keywords. Among them, the second transformation strategy may be a self-query strategy. The first transformation strategy makes multiple-angle descriptions based on the known first query statement, while the second transformation strategy does not need to process the first query statement, but only needs to know the keywords in the first query statement and use the keywords to construct statements. In the above way, multiple second query statements can be obtained for each expression transformation strategy, and the number of second query statements that can be obtained by multiple expression transformation strategies will be even more.

[0048] Step C: Obtain the target query statement based on multiple second query statements.

[0049] In the case where the types of expression transformation strategies are multiple, the second query statements corresponding to each expression transformation strategy can be mixed to obtain the target query statement. Specifically, Step C can be executed with reference to the following Step C1 and Step C2:

[0050] Step C1, obtain the mixing ratios corresponding to multiple expression transformation strategies, and determine the number of sentences to be extracted for each of the multiple expression transformation strategies based on the mixing ratios. The mixing ratios can be flexibly set according to requirements. Suppose 10 target query sentences are needed. The mixing ratio of the first transformation strategy to the second transformation strategy can be set to 4:6. That is, the number of sentences to be extracted for the first transformation strategy is 4, and the number of sentences to be extracted for the second transformation strategy is 6. It should be noted that the above is only an example and is not limited here.

[0051] Step C2, based on the number of sentences to be extracted, extract target query sentences from multiple second query sentences corresponding to each of the multiple expression transformation strategies.

[0052] The multiple target query sentences obtained in the above manner can comprehensively present the semantics of the target customer complaint information from multiple perspectives and in multiple ways, and can greatly improve the recall rate of the problem classification results of the target customer complaint information.

[0053] After obtaining the target query sentences, retrieval can be performed in the problem classification library based on the target query sentences. To ensure that the problem classification results can be obtained more efficiently and reliably, the reasonable construction of the problem classification library is also very crucial. An embodiment of the present disclosure provides a construction method for the problem classification library. Exemplarily, the problem classification library is constructed through the following steps 1 to 3:

[0054] Step 1, obtain customer complaint samples and the problem class target tags corresponding to the customer complaint samples; where the number of customer complaint samples is multiple. In practical applications, a large number of customer complaint samples can be collected. The samples can be manually labeled with the problem categories to which they belong, that is, the customer complaint samples carry problem class target tags. The target class target tags can include, for example, tags for indicating the problem object, tags for indicating the problem manifestation, tags for indicating the problem cause, etc., which can be flexibly set specifically and are not limited here.

[0055] Step 2, perform a simplification process on the customer complaint samples to obtain simplified text examples of the customer complaint samples. In practical applications, the customer complaint samples can be cleaned, etc. First, remove characters that are ineffective for semantic understanding, such as pause words and special characters, and then semantic understanding processing can be performed on them using, for example, a network model to obtain simplified text examples of the customer complaint samples.

[0056] Step 3, construct a problem classification library based on the problem class target tags corresponding to the customer complaint samples and the simplified text examples. In some specific implementation examples, Step 3 can be executed with reference to the following Steps 3.1 to 3.3:

[0057] Step 3.1, based on the problem category target tags and simplified text examples corresponding to the customer complaint samples, obtain the data blocks corresponding to the customer complaint samples. Exemplarily, for a customer complaint sample, the problem category target tags and simplified text examples corresponding to it can form a data block (also called a data shard) in a specified format.

[0058] For example, a data block can contain the following content: Problem object: construction site. Problem manifestation: garbage is piled up messily & in the wrong position. Problem cause: construction quality problem. During the construction process, the construction site was poorly managed, resulting in the garbage being piled up chaotically and in the wrong position, affecting the normal life and work of neighbors and the property management. This may be because the construction workers failed to comply with the garbage disposal regulations or lacked relevant training. Simplified text example: There is garbage or sundries piled up in the fire passage or public space. The above is only a basic example, and other tags such as tags indicating the problem scenario can also be included in actual applications.

[0059] Step 3.2, based on the data blocks corresponding to multiple customer complaint samples respectively, generate problem category blocks; wherein, the problem category blocks contain problem category target tags and at least one simplified text example corresponding to the problem category target tags; the problem category target tags and / or simplified text examples in different problem category blocks are different, and the character size of the problem category blocks is within a preset range.

[0060] In actual applications, two or more data blocks can be merged according to the size of the data blocks. For example, data blocks with the same problem category target tags but different simplified text examples can be merged to obtain problem category blocks, that is, simplified text examples of the same classification can be set in one problem category block. In addition, if the classification is relatively complex and there are many corresponding simplified text examples for a certain problem category, the data block can also be split. For example, a data block contains multiple simplified text examples, resulting in a relatively large data block. Therefore, the data block can be split into at least two sub-blocks, and each sub-block is used as a problem category block. In other words, the data block can be split into two identical classification items (problem category target tags), and each classification item corresponds to different simplified text examples. For easy understanding, reference can be made to Figure 2A schematic diagram of data block processing is shown. The merging process and the splitting process are illustrated. In the merging process, assume that one data block contains the problem category target tag A and the simplified text example 1, and another data block contains the problem category target tag A and the simplified text example 2. Then the two data blocks can be merged to obtain a problem category block containing the problem category target tag A and the simplified text examples 1-2. In the splitting process, assume that a larger data block contains the problem category target tag C and the simplified text examples 4-6. Then it can be split to obtain two problem category blocks, one of which contains the problem category target tag C and the simplified text examples 4-5, and the other contains the problem category target tag C and the simplified text example 6. Through the above method, multiple problem category blocks with a specified size can be obtained, which is more convenient for subsequent retrieval processing. The specific implementation of step 3.2 above can refer to the retrieval enhancement generation technology and will not be elaborated here.

[0061] Step 3.3, based on the problem category blocks, construct a problem classification library. Multiple problem category blocks can constitute a problem classification library. By constructing the problem classification library in the above way, it helps to improve the recall rate of the problem category blocks and the retrieval hit rate.

[0062] For easy understanding, reference can be made to Figure 3 The flow schematic diagram of a method for constructing a problem classification library shown below mainly includes the following steps S302 to step S316:

[0063] Step S302, obtain the customer complaint samples and the problem category target tags corresponding to the customer complaint samples.

[0064] Step S304, perform cleaning processing on the customer complaint samples. Specifically, the pause words, special characters, etc. that do not affect the semantics in the customer complaint samples can be removed through the cleaning processing. This step is an optional item.

[0065] Step S306, perform simplification processing on the cleaned customer complaint samples to obtain simplified text examples. The simplified text examples should at least contain concise statements, and may also contain summary paragraphs, etc., which are not restricted here.

[0066] Step S308, based on the problem category target tags corresponding to the customer complaint samples and the simplified text examples, obtain the data blocks corresponding to the customer complaint samples.

[0067] Step S310, determine whether the size of the data block is within the preset interval range. If not, execute step S312. If so, execute step S314.

[0068] Step S312, perform re-blocking processing on the data block. The specific implementation can refer to the relevant description in step 3.2 above and will not be elaborated here.

[0069] Step S314: Obtain the problem category block. The problem category block includes problem category tags and at least one simplified text example corresponding to the problem category tags.

[0070] Step S316: Based on the problem category block, construct a problem classification library. Additionally, in practical applications, the data in the problem classification library can be vectorized to obtain a vector library corresponding to the problem classification library for subsequent retrieval processing.

[0071] The specific implementation methods of the above steps can refer to the relevant content described above and will not be elaborated here. Through the above method, a problem classification library convenient for retrieval can be obtained. The problem classification library not only includes problem category tags but also corresponding simplified text examples, which is more convenient for subsequent retrieval. For example, the semantic similarity between the target query statement and the simplified text examples can be compared to ensure the comprehensiveness and reliability of the retrieval and reduce the retrieval difficulty.

[0072] The method provided by the embodiments of the present disclosure further includes: when it is determined that the target problem category is inaccurate, obtaining the standard problem category to which the target customer complaint information belongs; where the standard problem category is a problem category specified manually; updating the problem classification library based on the standard problem category to which the target customer complaint information belongs and the target simplified text. In practical applications, after retrieving the target problem category to which the target customer complaint information belongs through the network model, the target problem category can also be provided to the user, and feedback information from the user regarding the target problem category can be obtained. When the feedback information indicates that the target problem category is inaccurate, the standard problem category modified manually by the user can be obtained, and the standard problem category to which the target customer complaint information belongs and the target simplified text can be stored in the problem classification library, that is, the recognition error problem in individual scenarios can be improved through the way of human feedback learning, and the problem classification library can be gradually optimized to improve the subsequent retrieval accuracy.

[0073] In some embodiments, step S108, that is, the step of retrieving the target problem category to which the target customer complaint information belongs from the preset problem classification library according to the target query statement and key information, can be executed with reference to the following steps a to b:

[0074] Step a: According to the target query statement and key information, use a preset third network model to retrieve multiple candidate problem categories corresponding to the target customer complaint information from the preset problem classification library. The embodiments of the present disclosure do not limit the third network model. The third network model can also be a large language model. The third network model can be the same as or different from the foregoing first network model or second network model. In some specific implementation examples, the problem classification library includes multiple problem category blocks, and the problem category block includes problem category tags and at least one simplified text example corresponding to the problem category tags. Based on this, step a can be executed with reference to the following steps a1 to a2:

[0075] Step a1, according to the target query statement and the key information, use the preset third network model to retrieve multiple candidate category blocks from the preset question classification library; wherein each target query statement corresponds to a specified number of candidate category blocks, and the candidate category blocks contain simplified text samples that are semantically related to the target query statement, and / or the candidate category blocks contain at least part of the key information.

[0076] In actual applications, model prompt information can be pre-set, and the model prompt information is used to indicate how the third network model retrieves multiple candidate category blocks from the preset question classification library based on the target query statement and key information. For example, the model prompt information can be "You need to classify the following scenarios. Classification requires reference to the professional knowledge of {question classification library}, where you need to pay attention to whether the retrieved results contain {key information}. If it does not contain key information, you need to determine whether {target query statement} conforms to the description of {question classification library}. If it does not conform, please re-search and repeat the above process." It should be noted that the above is only a brief example and should not be regarded as a limitation. In actual applications, the model prompt information can be flexibly set according to needs, such as instructing the third network model in the model prompt information to first search based on the target query statement, and then filter the search results based on the key information to obtain the candidate category block. When the third network model searches based on the target query statement and key information, it can use retrieval algorithms such as vector distance algorithm. For details, please refer to the relevant technology and will not be repeated here. In practical applications, each target query statement can retrieve M candidate category blocks, and the value of M is 3, so multiple (assuming 5) target query statements can eventually obtain 15 candidate category blocks.

[0077] Step a2, based on the problem category tags in the candidate category blocks, obtain multiple candidate problem categories corresponding to the target customer complaint information. It can be understood that the problem category tags in each candidate category block can be used as candidate problem categories. If the problem category tags in multiple candidate category blocks partially overlap, multiple different candidate problem categories can be finally obtained by deduplication and other methods.

[0078] Step b, based on multiple candidate problem categories, using a preset fourth network model to determine the target problem category to which the target customer complaint information belongs from multiple candidate problem categories. The present disclosure embodiment does not limit the fourth network model, and the fourth network model may also be a large language model. The fourth network model may be the same as or different from the aforementioned first network model, second network model, or third network model. In order to ensure the accuracy of the target problem category finally obtained, in some specific implementation examples, step b may be performed with reference to the following steps b1 to b3:

[0079] Step b1: Obtain the business information corresponding to the target customer complaint information. Such business information can be, for example, the customer's order information.

[0080] Step b2: Sort multiple candidate problem categories based on the business information to obtain the top N candidate problem categories with the highest relevance to the business information; N is a preset integer value. For example, if the customer order indicates that the current stage is the water and electricity construction period, then the candidate problem categories related to the water and electricity construction period can be preferred, while the candidate problem categories related to other construction periods such as the painting construction period can be ranked later. After re - sorting the multiple candidate problem categories based on the relevance between the candidate problem categories and the business information, the top N candidate problem categories can be selected. The value of N can be flexibly set according to requirements and is not limited here. In this way, it helps to further improve the accuracy of problem recognition in the follow - up, and the finally recognized problem category truly conforms to the current business scenario.

[0081] Step b3: Use the preset fourth network model to determine the target problem category to which the target customer complaint information belongs from the top N candidate problem categories. In practical applications, prompt information for guiding the fourth network model to screen the target problem category can be constructed. For example, it can be "You are a problem classification expert, and you need to classify the following scenario: {problem scenario}; when classifying, you need to refer to the following materials and give the most reasonable classification information based on the materials. The classification materials are as follows: {retrieved document fragment 1}, {retrieved document fragment 2}, {retrieved document fragment 3}; please return the classified result in a specific format". The above - mentioned retrieved document fragments correspond to the candidate problem categories, and a brief example is given with 3 as an example. The above prompt information is only an example for easy understanding and should not be regarded as a limitation. In practical applications, the expression of the model prompt information can be flexibly adjusted. The fourth network model can finally output the target problem category, such as indicating the problem object category, problem manifestation category, problem cause category, problem scenario category, etc. In addition, it can also output the problem classification thinking process or analysis process, which can also be generated in combination with the aforementioned concise statements. The embodiments of the present disclosure do not specifically limit the form and content of the model output result. The above - mentioned process of generating the target problem category can also be called the enhanced generation method, which helps to further improve the accuracy of the target problem category.

[0082] Furthermore, in order to enable users to clearly know how to reasonably solve problems, the embodiments of the present disclosure can also retrieve the target solution corresponding to the target problem category from the preset solution library based on the target problem category to which the target customer complaint information belongs. That is, a solution library can be additionally set. The solution library records the solutions corresponding to each problem category for convenient associated retrieval, so as to quickly lock the corresponding target solution when determining the target problem category.

[0083] On the basis of obtaining the target problem category to which the target customer complaint problem belongs and the corresponding target solution, the embodiments of the present disclosure can also generate customized processing suggestions corresponding to the target customer complaint information based on the target problem category and the target solution by using a preset fifth network model. The embodiments of the present disclosure do not limit the fifth network model, and the fifth network model can also be a large language model. The fifth network model can be the same as or different from the foregoing first network model, second network model, third network model, or fifth network model. It can be understood that both the target problem category and the target solution are established contents based on the database. To improve the user experience, based on the target problem category and the target solution, the fifth network model can be used to generate corresponding customized processing suggestions in combination with the actual target customer complaint information. For example, personalized contents such as the specific object name and location mentioned in the target customer complaint information can be specified in the customized processing suggestions to provide users with a complete personalized solution result.

[0084] For ease of understanding, the embodiments of the present disclosure also provide Figure 4 a schematic diagram of an information processing flow as shown, which shows that semantic understanding processing and key information extraction processing can be performed on the target customer complaint information, a first query statement can be constructed for the target simplified text, multiple second query statements can be constructed respectively through a preset expression transformation strategy (assuming there are X kinds in total), and processed by proportional mixing to obtain multiple target query statements. The multiple target query statements are combined with the key information to retrieve multiple candidate problem categories from the problem classification library, and then re-ranked in combination with the business information to obtain the top N candidate problem categories with the highest relevance to the business information. On this basis, enhanced generation processing is performed to obtain the target problem category. Among them, the enhanced generation processing is to infer and analyze the top N candidate problem categories by means of a model to obtain the target problem category. Further, the target solution corresponding to the target problem category can be retrieved from the solution library, and finally customized processing is performed on the basis of the target problem category and the target solution to obtain customized processing suggestions that can be provided to users. The above specific processing process can refer to the foregoing relevant content.

[0085] In addition, the above-mentioned processing can all be implemented with the help of large language models. Through the strong parsing and processing capabilities of large language models, the reliability of the results is guaranteed. Specifically, the information processing method provided by the embodiments of the present disclosure can classify problems based on a network model (such as a large language model). By constructing a problem classification library, a classification basis is provided for the classification of large language models. The cost is relatively controllable. And when analyzing problems in complex scenarios, the reasoning ability of large language models helps to infer potential problems. Moreover, when large language models perform problem classification, they can combine the context composed of context and knowledge, so as to more accurately analyze problem information. And when new data such as new scenarios appear, even if the problem classification library does not cover the relevant data, large language models will also refer to similar examples for problem positioning and analysis, obtain relatively accurate results, and help to further optimize and improve the problem classification library. And the iteration method is also relatively convenient, such as supplementing the corresponding data blocks to achieve new data entry.

[0086] Corresponding to the foregoing information processing method, the embodiments of the present disclosure further provide an information processing device. Figure 5 As shown in the structural schematic diagram of an information processing device provided by the embodiments of the present disclosure, the device can be implemented by software and / or hardware, and is generally integrated in an electronic device, such as Figure 5 shown, the information processing device includes:

[0087] An information acquisition module 502, configured to acquire target customer complaint information to be processed;

[0088] An information processing module 504, configured to perform semantic understanding processing on the target customer complaint information to obtain a target simplified text corresponding to the target customer complaint information; and perform key information extraction processing on the target customer complaint information to obtain key information in the target customer complaint information;

[0089] A statement generation module 506, configured to generate a target query statement based on the target simplified text;

[0090] A category retrieval module 508, configured to retrieve a target problem category to which the target customer complaint information belongs from a preset problem classification library according to the target query statement and the key information.

[0091] The above-mentioned device provided by the embodiments of the present disclosure not only eliminates the need for manual classification, greatly saves costs and improves classification efficiency, but also comprehensively retrieves based on the key information in the target customer complaint information and the simplified text that can present the semantics of the target customer complaint information, effectively ensuring the accuracy of the retrieved problem categories.

[0092] In some embodiments, the information processing module 504 is specifically configured to: perform semantic understanding processing on the target customer complaint information by using a preset first network model to obtain at least one target simplified text corresponding to the target customer complaint information; perform key information extraction processing on the target customer complaint information by using a preset second network model to obtain key information in the target customer complaint information; the key information includes one or more of a problem object, a problem manifestation, and a problem cause.

[0093] In some embodiments, the statement generation module 506 is specifically configured to: obtain a first query statement based on the target simplified text; generate multiple second query statements based on the first query statement by using at least one preset statement transformation strategy; obtain a target query statement based on the multiple second query statements.

[0094] In some embodiments, when there are multiple types of the statement transformation strategies, the statement generation module 506 is specifically configured to: obtain a mixing ratio corresponding to the multiple statement transformation strategies, and determine the respective statement extraction quantities corresponding to the multiple statement transformation strategies based on the mixing ratio; extract a target query statement from the multiple second query statements respectively corresponding to the multiple statement transformation strategies based on the statement extraction quantities.

[0095] In some embodiments, the category retrieval module 508 is specifically configured to: retrieve, from a preset problem classification library, multiple candidate problem categories corresponding to the target customer complaint information by using a preset third network model according to the target query statement and the key information; determine a target problem category to which the target customer complaint information belongs from the multiple candidate problem categories by using a preset fourth network model.

[0096] In some embodiments, the category retrieval module 508 is specifically configured to: obtain service information corresponding to the target customer complaint information; sort the multiple candidate problem categories based on the service information to obtain the top N candidate problem categories with the highest relevance to the service information; N is a preset integer value; determine a target problem category to which the target customer complaint information belongs from the top N candidate problem categories by using a preset fourth network model.

[0097] In some embodiments, the problem classification library includes a plurality of problem category blocks, and each problem category block includes a problem category tag and at least one simplified text example corresponding to the problem category tag; specifically, the category retrieval module 508 is configured to: retrieve a plurality of candidate category blocks from a preset problem classification library by using a preset third network model according to the target query statement and the key information; wherein, each target query statement corresponds to a specified number of candidate category blocks, and the candidate category blocks include simplified text examples semantically related to the target query statement, and / or at least some information in the key information; based on the problem category tags in the candidate category blocks, obtain a plurality of candidate problem categories corresponding to the target customer complaint information.

[0098] In some embodiments, the information processing device further includes a classification library construction module, configured to: obtain customer complaint samples and problem category tags corresponding to the customer complaint samples; wherein, the number of the customer complaint samples is multiple; perform a simplification process on the customer complaint samples to obtain simplified text examples of the customer complaint samples; construct a problem classification library based on the problem category tags and the simplified text examples corresponding to the customer complaint samples.

[0099] In some embodiments, the classification library construction module is specifically configured to: obtain data blocks corresponding to the customer complaint samples based on the problem category tags and the simplified text examples corresponding to the customer complaint samples; generate problem category blocks based on the data blocks respectively corresponding to the multiple customer complaint samples; wherein, each problem category block includes a problem category tag and at least one simplified text example corresponding to the problem category tag; the problem category tags and / or the simplified text examples in different problem category blocks are different, and the character size of the problem category block is within a preset range; construct a problem classification library based on the problem category blocks.

[0100] In some embodiments, the device further includes a solution retrieval module, configured to retrieve a target solution corresponding to the target problem category from a preset solution library based on the target problem category to which the target customer complaint information belongs.

[0101] In some embodiments, the device further includes a customization information generation module, configured to generate a customized processing suggestion corresponding to the target customer complaint information by using a preset fifth network model based on the target problem category and the target solution.

[0102] In some embodiments, the device further includes a classification library update module, configured to obtain the standard problem category to which the target customer complaint information belongs when it is determined that the target problem category is inaccurate; wherein, the standard problem category is a problem category specified manually; and update the problem classification library based on the standard problem category to which the target customer complaint information belongs and the target simplified text.

[0103] The information processing device provided by the embodiments of the present disclosure can execute the information processing method provided by any embodiment of the present disclosure, and has functional modules and beneficial effects corresponding to the execution of the method.

[0104] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the device embodiments described above can refer to the corresponding process in the method embodiments, and will not be described in detail here.

[0105] The embodiments of the present disclosure provide an electronic device, which includes: a storage device on which a computer program is stored; and a processing device configured to execute the computer program in the storage device to implement the steps of any one of the methods in the present disclosure.

[0106] Next, refer to Figure 6 , which shows a schematic structural diagram of an electronic device 600 suitable for implementing the embodiments of the present disclosure. The terminal device in the embodiments of the present disclosure may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistants), PADs (Tablet Computers), PMPs (Portable Multimedia Players), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 6 The electronic device shown is only an example, and should not impose any limitation on the functions and usage scope of the embodiments of the present disclosure.

[0107] As Figure 6 shown, the electronic device 600 may include a processing device (such as a central processing unit, a graphics processing unit, etc.) 601, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 602 or the program loaded from the storage device 608 into the random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the electronic device 600 are also stored. The processing device 601, the ROM 602, and the RAM 603 are connected to each other through a bus 604. The input / output (I / O) interface 605 is also connected to the bus 604.

[0108] Typically, the following devices can be connected to the I / O interface 605: an input device 606 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 607 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 608 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 609. The communication device 609 can allow the electronic device 600 to communicate with other devices wirelessly or wiredly to exchange data. Although Figure 6 the electronic device 600 with various devices is shown, it should be understood that it is not required to implement or have all the shown devices. More or fewer devices can be alternatively implemented or had.

[0109] Specifically, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program contains program codes for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 609, or installed from the storage device 608, or installed from the ROM 602. When the computer program is executed by the processing device 601, the above functions defined in the method of the embodiment of the present disclosure are executed.

[0110] In addition to the above methods and devices, an embodiment of the present disclosure can also be a computer program product, which includes computer program instructions that cause the processor to execute the method provided by the embodiment of the present disclosure when the processor runs. The computer program product can be written in any combination of one or more programming languages for executing the program codes of the operations of the embodiment of the present disclosure. The programming languages include object-oriented programming languages such as Java, C++, etc., and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program codes can be executed completely on the user computing device, partially on the user device, executed as an independent software package, partially on the user computing device and partially on a remote computing device, or completely executed on a remote computing device or server.

[0111] Furthermore, an embodiment of the present disclosure can also be a computer-readable storage medium, on which computer program instructions are stored, and the computer program instructions cause the processor to execute the information processing method provided by the embodiment of the present disclosure when the processor runs.

[0112] The computer-readable storage medium may adopt any combination of one or more readable media. The readable media may be a readable signal medium or a readable storage medium. The readable storage medium may include, for example, but is not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the readable storage medium include: an electrical connection having one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0113] Embodiments of the present disclosure further provide a computer program product, including a computer program / instructions, which when executed by a processor implement the information processing method in the embodiments of the present disclosure.

[0114] It can be understood that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, usage scopes, usage scenarios, etc. of the personal information involved in the present disclosure should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with relevant laws and regulations.

[0115] For example, in response to receiving an active request from the user, a prompt message is sent to the user to clearly prompt the user that the operation requested by the user will require obtaining and using the user's personal information. Thus, the user can autonomously choose whether to provide personal information to software or hardware such as an electronic device, an application program, a server, or a storage medium that performs the operations of the technical solutions of the present disclosure according to the prompt message.

[0116] As an optional but non-limiting implementation manner, the manner of sending a prompt message to the user in response to receiving an active request from the user may be, for example, a pop-up window manner, and the prompt message may be presented in text in the pop-up window. In addition, the pop-up window may also carry a selection control for the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0117] It can be understood that the above process of notifying and obtaining the user's authorization is only illustrative and does not constitute a limitation on the implementation manner of the present disclosure. Other manners that meet relevant laws and regulations can also be applied to the implementation manner of the present disclosure.

[0118] It should be noted that in this document, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, such that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising an..." does not exclude the presence of additional identical elements in the process, method, article or device comprising said element.

[0119] The above are only specific embodiments of the present disclosure, enabling those skilled in the art to understand or implement the present disclosure. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present disclosure. Therefore, the present disclosure will not be limited to the embodiments described herein, but rather to the broadest scope consistent with the principles and novel features disclosed herein.

Claims

1. An information processing method, characterized in that, Including: Obtain target customer complaint information to be processed; Perform semantic understanding processing on the target customer complaint information to obtain a target simplified text corresponding to the target customer complaint information; And perform key information extraction processing on the target customer complaint information to obtain key information in the target customer complaint information; Generate a target query statement based on the target simplified text; Retrieve a target problem category to which the target customer complaint information belongs from a preset problem classification library according to the target query statement and the key information.

2. The method according to claim 1, characterized in that, The performing semantic understanding processing on the target customer complaint information to obtain a target simplified text corresponding to the target customer complaint information includes: Use a preset first network model to perform semantic understanding processing on the target customer complaint information to obtain at least one target simplified text corresponding to the target customer complaint information; The performing key information extraction processing on the target customer complaint information to obtain key information in the target customer complaint information includes: Use a preset second network model to perform key information extraction processing on the target customer complaint information to obtain key information in the target customer complaint information; the key information includes one or more of a problem object, a problem manifestation, and a problem cause.

3. The method according to claim 1, wherein The generating a target query statement based on the target simplified text includes: Obtain a first query statement based on the target simplified text; Based on the first query statement, use at least one preset expression transformation strategy to generate multiple second query statements; Obtain a target query statement based on the multiple second query statements.

4. The method according to claim 3, characterized in that, When there are multiple types of the expression transformation strategies, the obtaining a target query statement based on the multiple second query statements includes: Obtain a mixing ratio corresponding to multiple types of the expression transformation strategies, and determine the number of statement extractions corresponding to each of the multiple types of the expression transformation strategies based on the mixing ratio; Based on the number of statement extractions, extract a target query statement from multiple second query statements corresponding to each of the multiple types of the expression transformation strategies.

5. The method according to claim 1, wherein The retrieving a target problem category to which the target customer complaint information belongs from a preset problem classification library according to the target query statement and the key information includes: According to the target query statement and the key information, use a preset third network model to retrieve multiple candidate problem categories corresponding to the target customer complaint information from a preset problem classification library; Based on the multiple candidate problem categories, use a preset fourth network model to determine a target problem category to which the target customer complaint information belongs from the multiple candidate problem categories.

6. The method according to claim 5, wherein The based on the multiple candidate problem categories, using a preset fourth network model to determine a target problem category to which the target customer complaint information belongs from the multiple candidate problem categories includes: Obtain business information corresponding to the target customer complaint information; Rank the multiple candidate problem categories based on the business information to obtain the top N candidate problem categories with the highest relevance to the business information; N is a preset integer value; Use a preset fourth network model to determine a target problem category to which the target customer complaint information belongs from the top N candidate problem categories.

7. The method according to claim 5, characterized in that The problem classification library contains multiple problem category blocks, and each problem category block contains a problem category label and at least one simplified text example corresponding to the problem category label; Retrieving, from a preset problem classification library using a preset third network model according to the target query statement and the key information, multiple candidate problem categories corresponding to the target customer complaint information, including: Retrieving, from a preset problem classification library using a preset third network model according to the target query statement and the key information, multiple candidate category blocks; wherein, each target query statement corresponds to a specified number of candidate category blocks, and the candidate category blocks contain simplified text examples semantically related to the target query statement, and / or, at least some of the information in the key information; Based on the problem category labels in the candidate category blocks, obtaining multiple candidate problem categories corresponding to the target customer complaint information.

8. The method according to claim 1, wherein The problem classification library is constructed through the following steps: Obtaining customer complaint samples and problem category labels corresponding to the customer complaint samples; wherein, the number of customer complaint samples is multiple; Performing a simplification process on the customer complaint samples to obtain simplified text examples of the customer complaint samples; Constructing a problem classification library based on the problem category labels and simplified text examples corresponding to the customer complaint samples.

9. The method according to claim 8, characterized in that, The constructing a problem classification library based on the problem category labels and simplified text examples corresponding to the customer complaint samples includes: Obtaining data blocks corresponding to the customer complaint samples based on the problem category labels and simplified text examples corresponding to the customer complaint samples; Generating problem category blocks based on the data blocks respectively corresponding to multiple customer complaint samples; wherein, each problem category block contains a problem category label and at least one simplified text example corresponding to the problem category label; the problem category labels and / or simplified text examples in different problem category blocks are different, and the character size of the problem category block is within a preset range; Constructing a problem classification library based on the problem category blocks.

10. The method according to any one of claims 1 to 9, characterized in that, The method further includes: Retrieving, from a preset solution library, a target solution corresponding to the target problem category based on the target problem category to which the target customer complaint information belongs.

11. The method according to claim 10, characterized in that, The method further includes: Generating a customized processing suggestion corresponding to the target customer complaint information using a preset fifth network model based on the target problem category and the target solution.

12. The method according to claim 1, wherein The method further includes: When it is determined that the target problem category is inaccurate, obtaining a standard problem category to which the target customer complaint information belongs; wherein, the standard problem category is a problem category specified manually; Updating the problem classification library based on the standard problem category to which the target customer complaint information belongs and the target simplified text.

13. An information processing apparatus, characterized in that, Including: An information acquisition module, configured to acquire target customer complaint information to be processed; An information processing module, configured to perform semantic understanding processing on the target customer complaint information to obtain a target simplified text corresponding to the target customer complaint information; And performing key information extraction processing on the target customer complaint information to obtain key information in the target customer complaint information; A statement generation module, configured to generate a target query statement based on the target simplified text; A category retrieval module, configured to retrieve, from a preset question classification library, a target question category to which the target customer complaint information belongs according to the target query statement and the key information.

14. An electronic device, characterized in that, The electronic device includes: A storage device, on which a computer program is stored; A processing device, configured to execute the computer program in the storage device to implement the steps of the information processing method according to any one of claims 1-12.

15. A computer-readable storage medium, characterized in that, The storage medium stores a computer program, and the computer program is used to execute the information processing method according to any one of the above claims 1-12.

16. A computer program product, characterized in that, Including a computer program, which implements the information processing method according to any one of claims 1-12 when being executed by a processor.

Images