Customer complaint early warning method, electronic equipment and storage medium

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a customer complaint early warning method, electronic equipment and a storage medium. The method comprises the steps of obtaining to-be-detected customer complaint data; inputting the to-be-detected customer complaint data into a pre-trained customer complaint early warning model to obtain customer complaint early warning information, the customer complaint early warning information comprising multi-level customer complaint types, content abstracts and customer complaint levels; wherein the training process of the customer complaint early warning model comprises the steps of obtaining historical customer complaint data marked with historical customer complaint early warning information, taking the historical customer complaint data as training data, and performing fine adjustment on a preset base large language model in a low-rank adaptive mode to obtain the customer complaint early warning model. The training data is marked according to the preset customer complaint early warning information structure, so that the completeness and the accuracy of the customer complaint early warning information output by the customer complaint early warning model are relatively high; and the customer complaint early warning model is relatively short in development period and relatively low in development cost.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and particularly to a customer complaint warning method, an electronic device, and a storage medium. Background Art

[0002] With the wide application of voice and online customer service systems in various service industries, they can well assist customer service personnel in responding to and managing customer problems and needs. Conducting complaint warnings on the reactions of customers during the customer service process, and timely perceiving customers' dissatisfaction with products and services to prevent the expansion of negative impacts, can improve the customer experience and protect the corporate brand reputation.

[0003] Therefore, customer complaint warning is an important function of the customer service system. In the customer complaint warning methods of related technologies, it is mainly implemented based on manually preset empirical rules or traditional machine learning methods, and there are technical problems such as incomplete warning information content and low accuracy. Summary of the Invention

[0004] The present disclosure provides a customer complaint warning method, an electronic device, and a storage medium to improve the integrity and accuracy of customer complaint warning information.

[0005] According to a first aspect of the present disclosure, there is provided a customer complaint warning method, the method including:

[0006] Obtaining customer complaint data to be detected;

[0007] Inputting the customer complaint data to be detected into a pre-trained customer complaint warning model to obtain customer complaint warning information, the customer complaint warning information including multi-level customer complaint types, content summaries, and customer complaint levels;

[0008] Wherein, the training process of the customer complaint warning model includes: obtaining historical customer complaint data labeled with historical customer complaint warning information, using the historical customer complaint data as training data, and fine-tuning a preset base large language model in a low-rank adaptation manner to obtain the customer complaint warning model.

[0009] Optionally, the obtaining historical customer complaint data labeled with historical customer complaint warning information includes:

[0010] Obtaining historical customer complaint voice data, and extracting text content and emotion information from the historical customer complaint voice data;

[0011] Generating emotion tags based on the emotion information;

[0012] Based on the text content and emotion tags, labeling the historical customer complaint warning information of the historical customer complaint voice data according to a preset customer complaint warning information structure to obtain the historical customer complaint data.

[0013] Optionally, marking the historical customer complaint warning information of the historical customer complaint voice data according to a preset customer complaint warning information structure based on the text content and emotion tags to obtain the historical customer complaint data, including:

[0014] Marking the historical customer complaint warning information of the historical customer complaint voice data according to a preset customer complaint warning information structure based on the text content and emotion tags to obtain the first marked data;

[0015] Inputting the first marked data into a preset large language model, and performing amplification processing on the first marked data through the preset large language model to obtain the first customer complaint amplified data, where the first customer complaint amplified data is the historical customer complaint data.

[0016] Optionally, obtaining the historical customer complaint data marked with customer complaint warning information includes:

[0017] Obtaining historical customer complaint text data;

[0018] Marking the customer complaint warning information of the historical customer complaint text data according to a preset customer complaint warning information structure based on the text content of the historical customer complaint text data to obtain the historical customer complaint data.

[0019] Optionally, marking the customer complaint warning information of the historical customer complaint text data according to a preset customer complaint warning information structure based on the text content of the historical customer complaint text data to obtain the historical customer complaint data includes:

[0020] Marking the customer complaint warning information of the historical customer complaint text data according to a preset customer complaint warning information structure based on the text content of the historical customer complaint text data to obtain the second marked data;

[0021] Inputting the second marked data into a preset large language model, and performing amplification processing on the second marked data through the preset large language model to obtain the second customer complaint amplified data, where the second customer complaint amplified data is the historical customer complaint data.

[0022] Optionally, the method further includes:

[0023] When the customer complaint type included in the customer complaint warning information is a preset target customer complaint type, if the customer complaint data to be detected is customer complaint voice data, then using the customer complaint data to be detected as new historical customer complaint voice data, and extracting text content and emotion information from the new historical customer complaint voice data until a new customer complaint warning model is obtained.

[0024] Optionally, the method further includes:

[0025] When the type of customer complaint included in the customer complaint warning information is a preset target customer complaint type, if the customer complaint data to be detected is customer complaint text data, the customer complaint data to be detected is used as new historical customer complaint text data, and based on the text content of the new historical customer complaint text data, the new historical customer complaint text data is labeled according to a preset customer complaint warning information structure until a new customer complaint warning model is obtained.

[0026] Optionally, the method further includes:

[0027] Obtain the customer complaint type from the customer complaint warning information;

[0028] When it is detected that the customer complaint type does not belong to the existing customer complaint types, label the customer complaint warning information of the customer complaint data to be detected, and use the customer complaint data to be detected labeled with the customer complaint warning information as training data, and fine-tune the preset base large language model again through the low-rank adaptation method to obtain a new customer complaint warning model.

[0029] According to a second aspect of the present disclosure, an electronic device is provided. The electronic device includes: a memory and a processor, a computer program is stored on the memory, and when the processor executes the program, the method as described above is implemented.

[0030] According to a third aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored, and when the program is executed by a processor, the above method of the present disclosure is implemented.

[0031] According to a fourth aspect of the present disclosure, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the above method of the present disclosure is implemented.

[0032] The customer complaint warning method, electronic device and storage medium provided by the embodiments of the present disclosure, the customer complaint warning model obtained by fine-tuning the preset base large language model can be better applied to the customer service system. Since the training data is labeled according to a preset customer complaint warning information structure, the customer complaint warning information structure includes multi-level customer complaint types, content summaries and customer complaint levels. Therefore, the integrity and accuracy of the customer complaint warning information output by the customer complaint warning model are relatively high; and the customer complaint warning model is obtained by fine-tuning the preset base large language model. Therefore, the development cycle of the customer complaint warning model is short and the development cost is low. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] In the following description of the exemplary embodiments in conjunction with the drawings, more details, features and advantages of the present disclosure are disclosed. In the drawings:

[0034] Figure 1Flowchart of the overall technical solution of the customer complaint warning method provided by an exemplary embodiment of the present disclosure;

[0035] Figure 2 Flowchart of the customer complaint warning method provided by an exemplary embodiment of the present disclosure;

[0036] Figure 3 Block diagram of the structure of an electronic device provided by an exemplary embodiment of the present disclosure;

[0037] Figure 4 Block diagram of the structure of a computer system provided by an exemplary embodiment of the present disclosure. Detailed implementation manners

[0038] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although some embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. On the contrary, these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not used to limit the protection scope of the present disclosure.

[0039] It should be understood that the various steps recorded in the method embodiments of the present disclosure can be executed in different orders and / or in parallel. In addition, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this regard.

[0040] As used herein, the term "including" and its variations are open-ended, that is, "including but not limited to". The term "based on" is "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description. It should be noted that the concepts such as "first" and "second" mentioned in the present disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order of the functions executed by these devices, modules or units or their interdependent relationships.

[0041] It should be noted that the modifications of "one" and "plural" mentioned in the present disclosure are illustrative rather than restrictive. Those skilled in the art should understand that unless otherwise clearly stated in the context, it should be understood as "one or more".

[0042] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only for illustrative purposes and are not used to limit the scope of these messages or information.

[0043] It is understandable that before using the technical solutions disclosed in the embodiments of the present disclosure, the types, scope of use, 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.

[0044] For example, in response to receiving an active request from a user, a prompt message is sent to the user to clearly prompt the user that the operation requested to be performed 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, application, server, or storage medium that performs the operation of the technical solution of the present disclosure according to the prompt message.

[0045] As an optional but non-limiting implementation, in response to receiving an active request from the user, the method of sending a prompt message to the user may be, for example, a pop-up window, in which the prompt message may be presented in text. 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. It is understandable that the above notification and the process of obtaining user authorization are only illustrative and do not constitute a limitation on the implementation of the present disclosure. Other methods that meet relevant laws and regulations may also be applied to the implementation of the present disclosure.

[0046] As voice and online customer service systems are widely used in various service industries, they can help customer service personnel respond to and manage customer issues and needs. Complaint warnings based on customer responses during the customer service process can timely detect customer dissatisfaction with products and services to prevent negative impacts from expanding, thereby improving customer experience and protecting corporate brand reputation.

[0047] Therefore, customer complaint warning is an important function of the customer service system. The customer complaint warning methods in related technologies are mainly based on manually preset empirical rules or traditional machine learning methods, which have technical problems such as incomplete warning information content and low accuracy. In related technologies, the customer complaint warning function of the customer service system has problems such as long development cycle and high cost. How to improve the content integrity and accuracy of customer complaint warning information and reduce costs at the same time has become a technical problem that needs to be solved urgently in this field.

[0048] The disclosed embodiments are intended to solve the problems of low integrity and accuracy of customer complaint warning information of voice and online customer complaints, long system development cycle and high development cost, so as to achieve the application effect of complete and accurate customer complaint warning information, smaller investment and better effect.

[0049] Compared with traditional speech recognition technology, large-model speech recognition results (including speech speed, intonation, emotion and content) are more accurate. Vertical large language models have both certain common sense capabilities and specific field text analysis capabilities, and are often used to analyze and solve specific tasks.

[0050] The present disclosure uses high-performance large model speech recognition technology to extract conversation content and speaker emotion tags from customer service conversation voices; for specific customer service applications, it designs a customer complaint warning information structure (including multi-level customer complaint types, content summaries, customer complaint levels, etc.); according to the designed customer complaint warning information structure, it annotates historical customer complaint data with historical customer complaint warning information, and uses the historical customer complaint data marked with historical customer complaint warning information as training data to fine-tune a preset large model through a low-rank adaptation method to obtain a customer complaint warning model. And the customer complaint warning model is applied to the customer service system. After the customer service system obtains the customer complaint data to be detected, the customer complaint data to be detected is input into the customer complaint warning model, and the customer complaint warning model outputs customer complaint warning information, where the customer complaint warning information includes multi-level customer complaint types, content summaries, and customer complaint levels. It can be seen that the customer complaint warning information obtained through the present disclosure is complete in content and has a high accuracy rate; and the customer complaint warning model is obtained by fine-tuning an open-source large model. Therefore, the training cycle of the customer complaint warning model is short and the cost is low.

[0051] Moreover, in actual applications, if the customer complaint type output by the customer complaint warning model is a preset target customer complaint type, or the customer complaint type output by the customer complaint warning model is not an existing customer complaint type, the data to be detected can also be used as training data to continue training the customer complaint warning model, thereby continuously optimizing the training data of the customer complaint warning model and continuously iterating the customer complaint warning model, which helps to improve the content integrity and accuracy rate of the customer complaint warning information.

[0052] For the sake of clear description of the solution, the complete technical solution of the embodiments of the present disclosure will be elaborated in detail below. Among them, the complete technical solution of the present invention includes the training process and application process of the customer complaint warning model.

[0053] As Figure 1 shown, the customer complaint warning method provided by the embodiments of the present disclosure may include the following steps:

[0054] S110, determine whether the customer service conversation content is customer complaint voice data or customer complaint text data. If it is customer complaint voice data, execute step S120. If it is customer complaint text data, directly execute step S130.

[0055] S120, use high-performance large model speech recognition technology to extract the conversation text and speaker emotion tags of the customer complaint voice data.

[0056] S130, annotate the customer complaint voice data and customer complaint text data according to the preset customer complaint warning information structure.

[0057] Specifically, for customer complaint voice data, based on the dialogue text and emotion tags extracted by S120, data annotation can be performed according to a pre-set customer complaint warning information structure (including multi-level customer complaint types, content summaries, and customer complaint levels).

[0058] For customer complaint text data, based on the text content of the customer complaint text data, data annotation can be performed according to a pre-set customer complaint warning information structure (including multi-level customer complaint types, content summaries, and customer complaint levels).

[0059] Among them, the multi-level customer complaint types can be tree-structured customer complaint types. By setting multi-level customer complaint types, the specific customer complaint types can be determined more accurately and in detail. The content summary can intuitively display the core content of the customer complaint warning information. There can be multiple customer complaint levels, such as level one, level two, level three, etc., which can intuitively reflect the urgency of customer service personnel handling customer complaint cases. If the customer complaint level is high, it means that customer service personnel need to prioritize handling customer complaint cases.

[0060] S140, based on S130, use a high-performance large language model to augment the customer complaint voice data and customer complaint text data to obtain customer complaint augmented data.

[0061] Among them, the customer complaint augmented data can include customer complaint voice data, customer complaint text data, newly added customer complaint voice data obtained by augmenting the customer complaint voice data, and newly added customer complaint voice data obtained by augmenting the customer complaint text data. A high-performance large language model refers to a deep learning model trained with a large amount of text data, which has powerful natural language understanding and generation capabilities. In this step, a pre-set high-performance large language model can be determined among many high-performance large language models, and the customer complaint voice data and customer complaint text data can be augmented through the pre-set high-performance large language model. The purpose of augmenting the customer complaint voice data and customer complaint text data is to obtain more training data, so that in the following steps, the customer complaint warning model can be trained with more training data.

[0062] S150, use the customer complaint augmented data as training data and fine-tune the pre-set base large language model through low-rank adaptation to obtain the customer complaint warning model.

[0063] Among them, low-rank adaptation is a method for fine-tuning large pre-trained language models. Its core idea is to reduce the number of parameters to be trained by adding a low-rank matrix to the model weight matrix, thereby improving training efficiency and reducing resource consumption.

[0064] The base large language model, also known as the foundation model, is a large language model trained from scratch. It generally trains on a vast amount of text and code data in a self-supervised training manner, possessing basic capabilities and knowledge. It is a general model, also known as a pre-trained large model. Model fine-tuning refers to when training a model on a new task or domain, only updating the low-rank adaptation layer while keeping the weights of the pre-trained model fixed. This allows the model to effectively learn task-specific information without changing its general knowledge.

[0065] The customer complaint warning model obtained by fine-tuning the preset base large language model can be better applied to the customer service system. Since the training data is labeled according to the preset customer complaint warning information structure, which includes multi-level customer complaint types, content summaries, and customer complaint levels, the integrity and accuracy of the customer complaint warning information output by the customer complaint warning model are relatively high. And since the customer complaint warning model is obtained by fine-tuning the preset base large language model, the development cycle of the customer complaint warning model is short and the development cost is low.

[0066] S160, practically apply the customer complaint warning model obtained in step S150.

[0067] Specifically, input the customer complaint data to be detected into the customer complaint warning model obtained in step S150, and the customer complaint warning model outputs customer complaint warning information. The output customer complaint warning information includes multi-level customer complaint types, content summaries, and customer complaint levels.

[0068] S170, use a high-performance closed-source large language model and manual sampling to review the customer complaint data to be detected in the actual application.

[0069] S180, determine whether the customer complaint type in the customer complaint warning information is the preset target customer complaint type. If so, return to execute step S110.

[0070] S190, determine whether the customer complaint type in the customer complaint warning information is a new customer complaint type. If so, return to execute step S110.

[0071] Specifically, after practically applying the customer complaint warning model, it is also possible to review the customer complaint data in the actual application. If the customer complaint type is the preset target customer complaint type or the customer complaint type is a new customer complaint type, then add the customer complaint data in the actual application to the training data and train the customer complaint warning model to obtain a new customer complaint warning model. Thus, the purpose of continuously optimizing the training data and iterating the customer complaint warning model is achieved, which helps to improve the accuracy of the customer complaint warning model.

[0072] The technical solution provided by the embodiments of the present disclosure can better apply the customer complaint warning model obtained by fine-tuning the preset base large language model to the customer service system. Since the training data is labeled according to the preset customer complaint warning information structure, which includes multi-level customer complaint types, content summaries, and customer complaint levels, the integrity and accuracy of the customer complaint warning information output by the customer complaint warning model are relatively high; and the customer complaint warning model is obtained by fine-tuning the preset base large language model, so the development cycle of the customer complaint warning model is short and the development cost is low.

[0073] Moreover, after the customer complaint warning model is actually applied, the customer complaint data in the actual application can also be audited. If the customer complaint type is the preset target customer complaint type or the customer complaint type is a new customer complaint type, then the customer complaint data in the actual application is also added to the training data to train the customer complaint warning model until a new customer complaint warning model is obtained. Thus, the purpose of continuously optimizing the training data and iterating the customer complaint warning model is achieved, which helps to improve the accuracy rate of the customer complaint warning model.

[0074] After elaborating on the overall technical solution of the embodiments of the present disclosure in detail, a customer complaint warning method provided by the embodiments of the present disclosure will be elaborated in detail below.

[0075] As Figure 2 shown, a customer complaint warning method provided by an embodiment of the present invention may include the following steps:

[0076] S210, obtain the customer complaint data to be detected.

[0077] Specifically, the customer complaint data to be detected may be customer service conversation data, which may be customer complaint voice data or customer complaint text data.

[0078] S220, input the customer complaint data to be detected into the pre-trained customer complaint warning model to obtain customer complaint warning information.

[0079] The customer complaint warning information includes multi-level customer complaint types, content summaries, and customer complaint levels.

[0080] Among them, the training process of the customer complaint warning model includes: obtaining historical customer complaint data labeled with historical customer complaint warning information, using the historical customer complaint data as training data, and fine-tuning the preset base large language model in a low-rank adaptation manner to obtain the customer complaint warning model.

[0081] Specifically, after obtaining the customer complaint data to be detected, the customer complaint data to be detected can be input into the pre-trained customer complaint warning model to obtain customer complaint warning information. Among them, the customer complaint warning information includes multi-level customer complaint types, content summaries, and customer complaint levels. For example, the multi-level customer complaint types can be tree-structured customer complaint types. By outputting the multi-level customer complaint types, the specific customer complaint types can be determined more accurately and in detail. The content summary can intuitively display the core content of the customer complaint warning information. There can be multiple customer complaint levels, such as level one, level two, level three, etc., which can intuitively reflect the urgency of customer service personnel to handle customer complaint cases. If the customer complaint level is high, then the customer complaint case can be processed preferentially.

[0082] It should be noted that the training process of the customer complaint warning model has been described in the above embodiments and will not be elaborated here.

[0083] The technical solution provided by the embodiments of the present disclosure can better apply the customer complaint warning model obtained by fine-tuning the preset base large language model to the customer service system. Since the training data is labeled according to the preset customer complaint warning information structure, which includes multi-level customer complaint types, content summaries, and customer complaint levels, the integrity and accuracy of the customer complaint warning information output by the customer complaint warning model are relatively high; and since the customer complaint warning model is obtained by fine-tuning the preset base large language model, the development cycle of the customer complaint warning model is short and the development cost is low.

[0084] In Figure 2 Based on the shown embodiments, in one implementation, obtaining historical customer complaint data labeled with historical customer complaint warning information may include the following steps, namely steps a1 to a3:

[0085] Step a1, obtain historical customer complaint voice data, and extract text content and emotion information from the historical customer complaint voice data.

[0086] Step a2, generate emotion labels based on the emotion information.

[0087] Step a3, label the historical customer complaint warning information of the historical customer complaint voice data according to the preset customer complaint warning information structure based on the text content and emotion labels to obtain historical customer complaint data.

[0088] Specifically, since the high-performance large language model has a high speech recognition accuracy and can recognize the speech rate, intonation, emotion, and content of speech data, the text content and emotion information of historical customer complaint speech data can be extracted through the high-performance large language model, and emotion tags can be generated according to the emotion information. The emotion tags can be happy, sad, angry, etc., and the embodiments of the present disclosure do not make specific limitations on this. After obtaining the text content and emotion tags, the customer complaint type, content summary, and customer complaint level of the historical customer complaint speech data can be marked according to the preset customer complaint warning information structure, that is, label the training data. Thus, in actual applications, after inputting the customer complaint data to be detected into the customer complaint warning model, the customer complaint warning model can output customer complaint warning information including the customer complaint type, content summary, and customer complaint level, and the output customer complaint warning information content is more complete and the accuracy is higher.

[0089] Based on the above embodiments, as an implementation manner of the embodiments of the present disclosure, step a3 above, based on the text content and emotion tags, mark the historical customer complaint warning information of the historical customer complaint speech data according to the preset customer complaint warning information structure to obtain historical customer complaint data, which may include the following steps, namely steps a31 and a32:

[0090] Step a31, based on the text content and emotion tags, mark the historical customer complaint warning information of the historical customer complaint speech data according to the preset customer complaint warning information structure to obtain the first marked data.

[0091] Step a32, input the first marked data into the preset large language model, and perform amplification processing on the first marked data through the preset large language model to obtain the first customer complaint amplified data, and the first customer complaint amplified data is the historical customer complaint data.

[0092] The large language model refers to a deep learning model trained with a large amount of text data and has powerful natural language understanding and generation capabilities. To increase the training data of the customer complaint warning model, a preset large language model can be determined among many high-performance large language models, and the first marked data is amplified through the preset large language model to obtain the first customer complaint amplified data, and the first customer complaint amplified data is also the historical customer complaint data. In this way, the historical customer complaint data includes the first marked data and the first customer complaint amplified data, and by increasing the number of training data, the prediction accuracy of the trained customer complaint warning model is higher.

[0093] In Figure 1 Based on the embodiments shown, in one implementation manner, obtaining the historical customer complaint data marked with customer complaint warning information may include the following two steps, namely steps b1 and b2:

[0094] Step b1, obtain the historical customer complaint text data.

[0095] Step b2: Based on the text content of the historical customer complaint text data, label the customer complaint warning information of the historical customer complaint text data according to the preset customer complaint warning information structure to obtain historical customer complaint data.

[0096] Specifically, after obtaining the historical customer complaint text data, the text content of the historical customer complaint text data can be obtained. Based on the text content of the historical customer complaint text data, label the customer complaint type, content summary, and customer complaint level of the historical customer complaint text data according to the preset customer complaint warning information structure, that is, label the training data. Thus, in actual application, after inputting the customer complaint data to be detected into the customer complaint warning model, the customer complaint warning model can output the customer complaint warning information including the customer complaint type, content summary, and customer complaint level, and the content of the output customer complaint warning information is more complete and the accuracy is higher.

[0097] Based on the above embodiment, as an implementation manner of the present disclosure, step b2: Based on the text content of the historical customer complaint text data, label the customer complaint warning information of the historical customer complaint text data according to the preset customer complaint warning information structure to obtain historical customer complaint data, may include the following two steps, namely step b21 and step b22:

[0098] Step b21: Based on the text content of the historical customer complaint text data, label the customer complaint warning information of the historical customer complaint text data according to the preset customer complaint warning information structure to obtain the second labeled data.

[0099] Step b22: Input the second labeled data into the preset large language model, and perform amplification processing on the second labeled data through the preset large language model to obtain the second customer complaint amplified data, and the second customer complaint amplified data is the historical customer complaint data.

[0100] The large language model refers to a deep learning model trained with a large amount of text data and has powerful natural language understanding and generation capabilities. In order to increase the training data of the customer complaint warning model, a preset large language model can be determined from many high-performance large language models, and the second labeled data is amplified through the preset large language model to obtain the second customer complaint amplified data, and the second customer complaint amplified data is also the historical customer complaint data. In this way, the historical customer complaint data includes the second labeled data and the second customer complaint amplified data, and by increasing the quantity of the training data, the prediction accuracy of the trained customer complaint warning model is higher.

[0101] Based on the above embodiment, in one implementation manner, the customer complaint warning method may further include the following steps:

[0102] When the customer complaint type included in the customer complaint warning information is a preset target customer complaint type, if the customer complaint data to be detected is customer complaint voice data, the customer complaint data to be detected is used as new historical customer complaint voice data, and the text content and emotion information are extracted from the new historical customer complaint voice data until a new customer complaint warning model is obtained.

[0103] Specifically, the target customer complaint type can be an important customer complaint type among multiple customer complaint types. In order to enable the customer complaint warning model to more accurately predict the target customer complaint type in subsequent application processes, if the customer complaint data to be detected is customer complaint voice data, and after inputting the customer complaint voice data into the customer complaint warning model, the output customer complaint type is the target customer complaint type, then the customer complaint data to be detected can be used as new historical customer complaint voice data, and the text content and emotion information are extracted from the new historical customer complaint voice data. An emotion label is generated based on the emotion information; and based on the text content and the emotion label, the historical customer complaint warning information of the historical customer complaint voice data is marked according to the preset customer complaint warning information structure to obtain historical customer complaint data (new first marked data), and the obtained historical customer complaint data is used as training data; the new first marked data can also be amplified through a preset large language model to obtain first customer complaint amplified data, and the first customer complaint amplified data is also historical customer complaint data, that is, the first customer complaint amplified data is also used as training data. And the customer complaint warning model obtained through training is fine-tuned in a low-rank adaptation manner until a new customer complaint warning model is obtained.

[0104] It can be seen that through this implementation method, it is possible to continuously optimize the training data and continuously iterate the customer complaint warning model, thereby improving the prediction accuracy of the customer complaint warning model.

[0105] Based on the above embodiments, in one implementation manner, the customer complaint warning method may further include the following steps:

[0106] When the customer complaint type included in the customer complaint warning information is a preset target customer complaint type, if the customer complaint data to be detected is customer complaint text data, the customer complaint data to be detected is used as new historical customer complaint text data, and based on the text content of the new historical customer complaint text data, the new historical customer complaint text data is marked according to the preset customer complaint warning information structure until a new customer complaint warning model is obtained.

[0107] Specifically, the target customer complaint type can be the relatively important one among multiple customer complaint types. In order to enable the customer complaint warning model to more accurately predict the target customer complaint type in subsequent application processes, if the customer complaint data to be detected is customer complaint text data, and after inputting the customer complaint text data into the customer complaint warning model, the output customer complaint type is the target customer complaint type, then the customer complaint data to be detected can be used as new historical customer complaint text data, and based on the text content of the customer complaint data to be detected, the historical customer complaint warning information of the historical customer complaint text data can be marked according to the preset customer complaint warning information structure to obtain historical customer complaint data (new second marked data), and the obtained historical customer complaint data can be used as training data; it is also possible to perform amplification processing on the new second marked data through a preset large language model to obtain second customer complaint amplified data, and the second customer complaint amplified data is also historical customer complaint data, that is, the second customer complaint amplified data is also used as training data. And the customer complaint warning model obtained through training can be fine-tuned in a low-rank adaptation manner until a new customer complaint warning model is obtained.

[0108] It can be seen that through this implementation method, it is possible to continuously optimize the training data and continuously iterate the customer complaint warning model, thereby improving the prediction accuracy of the customer complaint warning model.

[0109] Based on the above embodiments, the customer complaint warning method may further include the following steps, namely step c1 and step c2:

[0110] Step c1, obtain the customer complaint type from the customer complaint warning information.

[0111] Step c2, in the case where it is detected that the customer complaint type does not belong to the existing customer complaint types, mark the customer complaint warning information of the customer complaint data to be detected, and use the customer complaint data to be detected marked with the customer complaint warning information as training data, and once again fine-tune the preset base large language model in a low-rank adaptation manner to obtain a new customer complaint warning model.

[0112] Specifically, the customer complaint warning information includes the type of customer complaint. Therefore, the type of customer complaint can be obtained from the customer complaint warning information. If it is detected that the type of customer complaint does not belong to the existing types of customer complaints, in order to improve the generalization ability of the model, the customer complaint warning model can also be fine-tuned using the data to be detected. Among them, the customer complaint data to be detected can be customer complaint voice data or customer complaint text data. The annotation methods for both customer complaint voice data and customer complaint text data are elaborated in detail in the above embodiments and will not be repeated here. And the customer complaint data to be detected marked with the customer complaint warning information is used as training data, and the preset base large language model is fine-tuned again through the low-rank adaptation method to obtain a new customer complaint warning model. Moreover, the labeled data can also be amplified by the preset large language model, and the amplified customer complaint data is also used as training data, and the preset base large language model is fine-tuned again through the low-rank adaptation method to obtain a new customer complaint warning model.

[0113] It can be seen that through this implementation method, it is possible to continuously optimize the training data and continuously iterate the customer complaint warning model, thereby improving the prediction accuracy of the customer complaint warning model.

[0114] The embodiments of the present disclosure also provide an electronic device, including: at least one processor; a memory for storing executable instructions of the at least one processor; wherein, the at least one processor is configured to execute the instructions to implement the above methods disclosed in the embodiments of the present disclosure.

[0115] Figure 3 It is a schematic structural diagram of an electronic device provided by an exemplary embodiment of the present disclosure. As Figure 3 shown, the electronic device 300 includes at least one processor 301 and a memory 302 coupled to the processor 301. The processor 301 can execute the corresponding steps in the above methods disclosed in the embodiments of the present disclosure.

[0116] The above-mentioned processor 301 can also be referred to as a central processing unit (CPU). It can be an integrated circuit chip with the ability to process signals. Each step in the above-mentioned method disclosed in the embodiments of the present disclosure can be completed by the integrated logic circuit in the hardware of the processor 301 or instructions in the form of software. The above-mentioned processor 301 can be a general-purpose processor, a digital signal processor (DSP), an ASIC (Application Specific Integrated Circuit), an FPGA (field-programmable gate array), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present disclosure can be directly embodied as being executed and completed by the hardware decoding processor, or executed and completed by a combination of the hardware and software modules in the decoding processor. The software module can be located in the memory 302, such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory, or an electrically erasable programmable memory, a register, and other well-known storage media in the art. The processor 301 reads the information in the memory 302 and combines its hardware to complete the steps of the above-mentioned method.

[0117] In addition, when various operations / processes according to the present disclosure are implemented through software and / or firmware, a program constituting the software can be installed from a storage medium or a network into a computer system having a dedicated hardware structure, such as Figure 4 the computer system 1900 shown. When various programs are installed in the computer system, it can execute various functions, including the functions described above, etc. Figure 4 It is a block diagram of the structure of a computer system provided by an exemplary embodiment of the present disclosure.

[0118] The computer system 1900 is intended to represent various forms of digital electronic computer devices, such as, a laptop computer, a desktop computer, a workbench, a personal digital assistant, a server, a blade server, a mainframe computer, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, a personal digital processor, a cellular phone, a smart phone, a wearable device, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are only examples and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0119] As Figure 4As shown, computer system 1900 includes a computing unit 1901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1902 or a computer program loaded from a storage unit 1908 into a random access memory (RAM) 1903. In the RAM 1903, various programs and data required for the operation of the computer system 1900 can also be stored. The computing unit 1901, the ROM 1902, and the RAM 1903 are connected to each other via a bus 1904. An input / output (I / O) interface 1905 is also connected to the bus 1904.

[0120] Multiple components in the computer system 1900 are connected to the I / O interface 1905, including: an input unit 1906, an output unit 1907, a storage unit 1908, and a communication unit 1909. The input unit 1906 can be any type of device capable of inputting information into the computer system 1900. The input unit 1906 can receive input digital or character information and generate key signal inputs related to user settings and / or function controls of the electronic device. The output unit 1907 can be any type of device capable of presenting information and can include, but is not limited to, a display, a speaker, a video / audio output terminal, a vibrator, and / or a printer. The storage unit 1908 can include, but is not limited to, a magnetic disk, an optical disk. The communication unit 1909 allows the computer system 1900 to exchange information / data with other devices via a network such as the Internet and can include, but is not limited to, a modem, a network card, an infrared communication device, a wireless communication transceiver, and / or a chipset, such as a BluetoothTM device, a WiFi device, a WiMax device, a cellular communication device, and / or the like.

[0121] The computing unit 1901 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 1901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The computing unit 1901 executes the various methods and processes described above. For example, in some embodiments, the above-described methods disclosed in the embodiments of the present disclosure can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 1908. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device via the ROM 1902 and / or the communication unit 1909. In some embodiments, the computing unit 1901 can be configured to execute the above-described methods disclosed in the embodiments of the present disclosure in any other appropriate manner (e.g., by means of firmware).

[0122] An embodiment of the present disclosure also provides a computer-readable storage medium. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device can execute the above methods disclosed in the embodiments of the present disclosure.

[0123] The computer-readable storage medium in the embodiments of the present disclosure may be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device. The above computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specifically, the above computer-readable storage medium may include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0124] The above computer-readable medium may be included in the above electronic device; or may exist separately without being assembled into the electronic device.

[0125] An embodiment of the present disclosure also provides a computer program product, including a computer program, wherein when the computer program is executed by a processor, the above methods disclosed in the embodiments of the present disclosure are implemented.

[0126] In the embodiments of the present disclosure, computer program code for performing the operations of the present disclosure can be written in one or more programming languages or combinations thereof. The above programming languages include, but are not limited to, object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network (including a local area network (LAN) or a wide area network (WAN)), or can be connected to an external computer.

[0127] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architectures, functions, and operations of systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flowchart or block diagram may represent a module, a segment of a program, or a portion of code that contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions noted in the blocks may occur in a different order than that noted in the accompanying drawings. For example, two consecutive blocks shown may actually be executed substantially in parallel, or they may sometimes be executed in the reverse order, depending on the functions involved. It should also be noted that each block in the block diagram and / or flowchart, and combinations of blocks in the block diagram and / or flowchart, can be implemented by a dedicated hardware-based system that performs the specified functions or operations, or by a combination of dedicated hardware and computer instructions.

[0128] The modules, components, or units described in the embodiments of the present disclosure can be implemented in software or in hardware. Among them, the names of the modules, components, or units do not, in some cases, constitute a limitation on the modules, components, or units themselves.

[0129] The functions described above can be performed, at least in part, by one or more hardware logic components. For example, without limitation, exemplary hardware logic components that can be used include: Field Programmable Gate Array (FPGA), Application Specific Integrated Circuit (ASIC), Application Specific Standard Product (ASSP), System on Chip (SOC), Complex Programmable Logic Device (CPLD), and so on.

[0130] The above description is only some embodiments of the present disclosure and an explanation of the applied technical principles. Those skilled in the art should understand that the scope of the disclosure involved in the present disclosure is not limited to the technical solutions formed by the specific combination of the above technical features, and should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosure concept. For example, technical solutions formed by mutually replacing the above features with technical features having similar functions (but not limited to) disclosed in the present disclosure.

[0131] Although some specific embodiments of the present disclosure have been described in detail by way of examples, those skilled in the art should understand that the above examples are only for illustration and not for limiting the scope of the present disclosure. Those skilled in the art should understand that the above embodiments can be modified without departing from the scope and spirit of the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A customer complaint early warning method, characterized in that: The method comprises: Obtain customer complaint data to be tested; Inputting the customer complaint data to be detected into a pre-trained customer complaint warning model to obtain customer complaint warning information, wherein the customer complaint warning information includes multi-level customer complaint types, content summaries, and customer complaint levels; Among them, the training process of the customer complaint warning model includes: obtaining historical customer complaint data marked with historical customer complaint warning information, using the historical customer complaint data as training data, and fine-tuning the preset base large language model through a low-rank adaptive method to obtain the customer complaint warning model.

2. The method according to claim 1, characterized in that The obtaining of historical customer complaint data marked with historical customer complaint warning information includes: Acquire historical customer complaint voice data, and extract text content and emotional information from the historical customer complaint voice data; generating an emotion label based on the emotion information; Based on the text content and the emotion label, the historical customer complaint warning information of the historical customer complaint voice data is annotated according to a preset customer complaint warning information structure to obtain the historical customer complaint data.

3. The method according to claim 2, characterized in that The method of labeling the historical customer complaint warning information of the historical customer complaint voice data based on the text content and the emotion tag according to a preset customer complaint warning information structure to obtain the historical customer complaint data includes: Based on the text content and the emotion tag, annotating the historical customer complaint warning information of the historical customer complaint voice data according to a preset customer complaint warning information structure to obtain first annotated data; The first annotated data is input into a preset large language model, and the first annotated data is amplified by the preset large language model to obtain first customer complaint amplified data, where the first customer complaint amplified data is the historical customer complaint data.

4. The method according to claim 1, characterized in that: The obtaining of historical customer complaint data marked with customer complaint warning information includes: Obtain historical customer complaint text data; Based on the text content of the historical customer complaint text data, the customer complaint warning information of the historical customer complaint text data is annotated according to a preset customer complaint warning information structure to obtain the historical customer complaint data.

5. The method according to claim 4, characterized in that The step of labeling the customer complaint warning information of the historical customer complaint text data based on the text content of the historical customer complaint text data according to a preset customer complaint warning information structure to obtain the historical customer complaint data includes: Based on the text content of the historical customer complaint text data, annotating the customer complaint warning information of the historical customer complaint text data according to a preset customer complaint warning information structure to obtain second annotated data; The second annotated data is input into a preset large language model, and the second annotated data is amplified by the preset large language model to obtain second customer complaint amplified data, where the second customer complaint amplified data is the historical customer complaint data.

6. The method according to claim 2, characterized in that The method further comprises: In the case where the customer complaint type included in the customer complaint warning information is a preset target customer complaint type, if the customer complaint data to be detected is customer complaint voice data, the customer complaint data to be detected is used as new historical customer complaint voice data, and text content and emotional information are extracted from the new historical customer complaint voice data until a new customer complaint warning model is obtained.

7. The method according to claim 4, characterized in that The method further comprises: In the case where the customer complaint type included in the customer complaint warning information is a preset target customer complaint type, if the customer complaint data to be detected is customer complaint text data, the customer complaint data to be detected is used as new historical customer complaint text data, and based on the text content of the new historical customer complaint text data, the new historical customer complaint text data is labeled according to the preset customer complaint warning information structure until a new customer complaint warning model is obtained.

8. The method according to any one of claims 1 to 7, characterized in that: The method further comprises: Obtaining the customer complaint type from the customer complaint warning information; When it is detected that the customer complaint type does not belong to the existing customer complaint type, the customer complaint warning information of the customer complaint data to be detected is marked, and the customer complaint data to be detected marked with the customer complaint warning information is used as training data. The preset base large language model is again fine-tuned through a low-rank adaptive method to obtain a new customer complaint warning model.

9. An electronic device, characterized in that: include: at least one processor; a memory for storing the at least one processor-executable instruction; The at least one processor is configured to execute the instructions to implement the method as claimed in any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that: When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform the method as claimed in any one of claims 1 to 8.

Images