Complaint information processing method, device and equipment applied to long text call

Abstract

Embodiments of the present disclosure disclose a complaint information processing method, device and equipment applied to long text call. A specific embodiment of the method comprises: performing change processing on target call text information to obtain user enhanced text information and customer service enhanced text information; performing fusion processing on the user enhanced text information and the customer service enhanced text information to obtain fused call text information; generating an emotional dialogue block information set according to the fused call text information and a preset dialogue selection condition; generating complaint reason information and early warning level information according to the emotional dialogue block information set, a preset complaint prompt word information set and a preset early warning level information set; generating complaint information according to the complaint reason information and the early warning level information; and in response to determining that the early warning level information of the complaint information represents a fifth early warning level of a target terminal device, shutting down the target terminal device. The embodiment reduces the hardware resource overhead and shortens the time consumption of complaint information generation.

Description

Technical Field

[0001] The embodiments disclosed herein relate to the field of computer technology, and more specifically to a method, apparatus, and device for processing customer complaint information in long text calls. Background Technology

[0002] By analyzing the text information of user complaint calls (for example, the text information of a complaint call could be related to hardware devices, such as ATMs), customer complaint information can be generated. This information can then be used to better clarify customer demands and address equipment malfunctions in a timely manner. Currently, the common approach to processing customer complaint information from long text calls is as follows: first, the complaint call text is randomly truncated; then, a large language model is used to analyze and process the truncated, excessively long complaint call text.

[0003] However, in practice, it has been found that when processing customer complaint information from long text calls using the above method, the following technical problems often arise: The maximum input length of large models is highly correlated with computer resources. Processing long call texts leads to excessive computer resource consumption and long processing times, resulting in poor timeliness in detecting hardware malfunctions. Consequently, failure to remotely shut down hardware in a timely manner can damage the hardware. Furthermore, randomly truncating excessively long complaint call texts can cause loss of contextual semantics, leading to lower accuracy in semantic analysis of the complete complaint call content. This can result in biased assessment of hardware status, causing incorrect shutdown of hardware and affecting user experience, or failure to shut down faulty devices in a timely manner, leading to device damage.

[0004] The information disclosed in this background section is only intended to enhance the understanding of the background of the inventive concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.

[0006] Some embodiments of this disclosure provide methods, apparatus, electronic devices, and computer-readable media for processing customer complaint information in long text calls to solve one or more of the technical problems mentioned in the background section above.

[0007] In a first aspect, some embodiments of this disclosure provide a method for processing customer complaint information applied to long text calls. The method includes: in response to receiving target call text information for a target terminal device, modifying the target call text information to obtain user-enhanced text information and customer service-enhanced text information; fusing the user-enhanced text information and the customer service-enhanced text information according to user call location information and customer service call location information corresponding to the target call text information to obtain fused call text information; generating an emotional dialogue block information set according to the fused call text information, preset constraint information, and preset dialogue selection conditions; generating customer complaint reason information and warning level information according to the emotional dialogue block information set, a preset customer complaint prompt word information set, and a preset warning level information set; generating customer complaint information corresponding to the target terminal device according to the customer complaint reason information and the warning level information; and closing the target terminal device in response to determining that the warning level information representing the target terminal device in the customer complaint information is the fifth warning level.

[0008] Secondly, some embodiments of this disclosure provide a customer complaint information processing apparatus for long text calls, comprising: a modification unit configured to modify the target call text information for a target terminal device in response to receiving target call text information for a target terminal device, thereby obtaining user-enhanced text information and customer service-enhanced text information; a fusion unit configured to fuse the user-enhanced text information and the customer service-enhanced text information based on user call location information and customer service call location information corresponding to the target call text information, thereby obtaining fused call text information; a first generation unit configured to generate an emotional dialogue block information set based on the fused call text information, preset constraint information, and preset dialogue selection conditions; a second generation unit configured to generate customer complaint reason information and warning level information based on the emotional dialogue block information set, a preset customer complaint prompt word information set, and a preset warning level information set; a third generation unit configured to generate customer complaint information corresponding to the target terminal device based on the customer complaint reason information and the warning level information; and a closing unit configured to close the target terminal device in response to determining that the warning level information representing the target terminal device in the customer complaint information is a fifth warning level.

[0009] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.

[0010] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the method described in any of the implementations of the first or second aspect.

[0011] The above-described embodiments of this disclosure have the following beneficial effects: A customer complaint information processing method applied to long text calls, according to some embodiments of this disclosure, can reduce computer resource consumption and shorten processing time, improve the timeliness of detecting hardware device malfunctions, and thus improve the accuracy of semantic analysis of the complete complaint call content and the user experience. The reasons for the large computer resource consumption, long processing time, poor timeliness of detecting hardware device malfunctions, low accuracy of semantic analysis of the complete complaint call content, and poor user experience are as follows: The maximum input length of a large model is highly correlated with computer resources. When processing long call texts, the computer resource consumption is excessive and the processing time is long, resulting in poor timeliness of detecting hardware device malfunctions, which may lead to hardware device damage due to failure to remotely shut down the hardware device in a timely manner. Furthermore, randomly truncating excessively long complaint call texts can cause missing contextual semantics, leading to low accuracy of semantic analysis of the complete complaint call content, resulting in inaccurate information about the hardware device's status, causing incorrect shutdown of the hardware device and affecting the user experience, or failure to shut down the device in a timely manner due to malfunction, leading to device damage. Based on this, some embodiments of the present disclosure of a customer complaint information processing method for long text calls firstly, in response to receiving target call text information for a target terminal device, modifying the target call text information to obtain user-enhanced text information and customer service-enhanced text information. This allows for enhancement processing of the target call text information, resulting in enhanced user-enhanced call text and enhanced customer service call text. Secondly, based on the user call location information and customer service call location information corresponding to the target call text information, fusing the user-enhanced text information and the customer service-enhanced text information to obtain fused call text information. This yields the fused call text. Then, based on the fused call text information, preset constraint information, and preset dialogue selection conditions, an emotional dialogue block information set is generated. This yields each emotional dialogue block. Next, based on the emotional dialogue block information set, a preset customer complaint prompt word information set, and a preset warning level information set, customer complaint reason information and warning level information are generated. This yields the customer complaint reason and warning level for each emotional dialogue block in the emotional dialogue block set. Finally, based on the customer complaint reason information and the warning level information, customer complaint information corresponding to the target terminal device is generated. Thus, the customer complaint information from the aforementioned call text can be obtained. Finally, in response to determining that the aforementioned customer complaint information represents the warning level information of the aforementioned target terminal device as the fifth warning level, the aforementioned target terminal device is shut down. This is because the process does not involve simply and crudely truncating the complaint call text and then using a large model to analyze and process the split complaint call text. Instead, the aforementioned call text information undergoes enhanced fusion processing to obtain fused call text information.Then, the aforementioned fused call text information is segmented to obtain a set of segmented emotional dialogue blocks. Finally, based on the aforementioned emotional dialogue block information set, the preset customer complaint prompt word information set, and the preset warning level information set, customer complaint information corresponding to the aforementioned call text information is obtained. Furthermore, the aforementioned target call text information can be segmented to obtain emotional dialogue block information sets for each emotional dialogue block with similar themes and emotions. This allows for segment-by-segment parsing of each emotional dialogue block with similar themes and emotions, thereby improving parsing speed and reducing the time required to generate customer complaint information. Therefore, it can reduce computer resource consumption and shorten processing time, improve the timeliness of detecting hardware malfunctions, and thus improve the accuracy of semantic analysis of the complete complaint call content and the user experience. Attached Figure Description

[0012] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.

[0013] Figure 1 This is a flowchart of some embodiments of a customer complaint information processing method applied to long text calls according to the present disclosure; Figure 2 This is a schematic diagram of the structure of some embodiments of a customer complaint information processing device applied to long text calls according to the present disclosure; Figure 3 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation

[0014] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0015] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.

[0016] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.

[0017] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0018] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.

[0019] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.

[0020] Figure 1 A flow 100 of some embodiments of a customer complaint information processing method for long text calls according to the present disclosure is shown. This customer complaint information processing method for long text calls includes the following steps: Step 101: In response to receiving the target call text information for the target terminal device, modify the target call text information to obtain user enhanced text information and customer service enhanced text information.

[0021] In some embodiments, in response to receiving target call text information for a target terminal device, the executing entity (e.g., a computing device) of the customer complaint information processing method applied to long text calls can modify the target call text information to obtain user-enhanced text information and customer service-enhanced text information. The target terminal device can represent an ATM. Here, the specific type of the target terminal device is not limited and can be adjusted according to actual needs. The target call text information can represent the call text of a user making a complaint by dialing a complaint hotline. The call text can represent the statements of a conversation between a user and customer service. The user-enhanced text information can represent structured text containing user call text information, preset keyword information, preset entity information, and preset emotion range information. The customer service-enhanced text information can represent structured text containing customer service call text information, the preset keyword information, the preset entity information, and the preset emotion range information. Here, the specific content of the aforementioned enhanced user text information and enhanced customer service text information is not limited. For example, enhanced user text information could be "User Call Text Information": "Hello, I want to complain!...", "Preset Keyword Information": ["Complaint"], "Preset Entity Information": ["Transaction Amount"], "Preset Emotional Range Information": 0.85, "Dialogue Position": 1. Enhanced customer service text information could be "Customer Service Call Text Information": "Sir, don't worry, I'll check your transaction records for you.", "Preset Keyword Information": ["Query"], "Preset Entity Information": ["Transaction Records"], "Preset Emotional Range Information": 0.1, "Dialogue Position": 2. The aforementioned user call text information can represent the statements of the target user in the aforementioned target call text information. The aforementioned target user can represent a user who initiates a complaint, provides feedback, or requests consultation in the aforementioned target call text information. The aforementioned customer service call text information can represent the statements of the target customer service representative in the aforementioned target call text information. The aforementioned target customer service representative can represent a user who receives feedback from the aforementioned target user, processes the aforementioned target user's complaint, or answers the aforementioned target user's consultation requests in the aforementioned call text information. The aforementioned preset keyword information can represent the core words used by the target user in their complaint within the user's call text. The specific content of these core words is not limited; for example, the core word could be "card swallowing." The aforementioned preset entity information can represent the identifier of the indirect carrier involved in the value transfer. For example, the indirect carrier could be a bank card, and its identifier could be a bank card number. The aforementioned preset emotion interval information can represent the various intervals used to divide the target user's emotion during the call. This preset emotion interval information can include a first interval, a second interval, a third interval, and a fourth interval. The first interval information can represent that the target user's emotion is calm.The first range of information described above represents an emotion score ranging from 0.0 to 0.3. The second range of information describes the target user's emotion as dissatisfaction, with an emotion score ranging from 0.3 to 0.6. The third range of information describes the target user's emotion as anger, with an emotion score ranging from 0.6 to 0.8. The fourth range of information describes the target user's emotion as rage, with an emotion score ranging from 0.8 to 1.0. It should be noted that the more dissatisfied the target user is, the higher the corresponding emotion score. The long text call describes a call text related to a complaint against the target terminal device.

[0022] In some optional implementations of certain embodiments, the aforementioned executing entity may modify the target call text information through the following steps to obtain user-enhanced text information and customer service-enhanced text information: The first step is to update the target call text information to obtain user call text information and customer service call text information. In practice, the executing entity can use regular expressions to split the target call text information to obtain the split user call text information and customer service call text information. It should be noted that the target call text information has a fixed format. Each line of text begins with "Customer Service:" or "User:" as the starting identifier and uses a semicolon or newline as the message separator. For example, the target call text information can be "Customer Service: Hello, it's a pleasure to serve you; User: Hello, I want to file a complaint...; Customer Service: Please don't worry, let me check the specific situation for you...".

[0023] The second step is to generate enhanced user text information and enhanced customer service text information based on the aforementioned user call text information and customer service call text information.

[0024] In some optional implementations of certain embodiments, the aforementioned execution entity may generate enhanced user text information and enhanced customer service text information based on the aforementioned user call text information and the aforementioned customer service call text information through the following steps: The first step is to treat each piece of user call text information as text information and perform the following recognition steps: The first sub-step involves processing the aforementioned text information according to a preset recognition method and preset keyword information to obtain keyword information corresponding to the text information. Both the aforementioned user call text information and the aforementioned customer service call text information can include various call text messages. The aforementioned call text information can represent individual statements within the aforementioned user call text information or the aforementioned customer service call text information. The aforementioned preset recognition method can be a method of recognition using regular expressions. The aforementioned keyword information can represent the vocabulary included in the aforementioned preset keyword information. In practice, the executing entity can use the aforementioned preset recognition method to process the aforementioned text information and obtain the vocabulary contained within the aforementioned preset keyword information as the keyword information corresponding to the aforementioned text information.

[0025] The second sub-step involves processing the text information using a pre-trained entity recognition model and the aforementioned preset entity information to obtain the value transfer entity information corresponding to the text information. The pre-trained entity recognition model can be any entity recognition model, such as a BERT-based model. This model takes text information as input and a label sequence as output. It may include an embedding layer, a multi-layer Transformer encoder, and an output layer. Each layer of the multi-layer Transformer encoder may include a multi-head self-attention layer, a feedforward neural network layer, two residual connections, and two normalization layers. The embedding layer takes text information as input and the corresponding vector sequence as output. The multi-layer Transformer encoder takes the corresponding vector sequence as input and a high-dimensional feature vector sequence as output. The output layer takes a high-dimensional feature vector sequence as input and a label sequence as output. The corresponding vector sequence represents the sequence formed by mapping each character in the text information to a vector and arranging them sequentially. The aforementioned high-dimensional feature vector sequence can represent the high-dimensional vector sequence output after the vector sequence is fused with contextual semantics by the Transformer encoder. The aforementioned label sequence can represent a sequence of the same length as the text information. The label at each position in the aforementioned label sequence can be used to indicate the type of BIO label to which the corresponding character belongs in the preset entity information. For example, for the text information "card number 6217", if "6217" is a bank card number entity, then the label sequence is "OOBIII" (where B corresponds to "6", the first I corresponds to "2", the second I corresponds to "1", and the third I corresponds to "7"). The aforementioned pre-trained entity recognition model can be trained in batches. The aforementioned value transfer entity information can represent the value transfer data included in the aforementioned preset entity information. Here, no specific limitation is made on the aforementioned value transfer data; for example, value transfer data can include: bank card number, account name. In practice, firstly, the aforementioned executing entity can input the text information into the aforementioned pre-trained entity recognition model to obtain the label sequence. Then, the characters corresponding to consecutive non-empty tags of the same type (i.e., tags whose tags are not "O") in the tag sequence are merged to obtain at least one continuous segment. Finally, the text information corresponding to the characters in each of the obtained continuous segments is determined as the value transfer entity information.

[0026] The second step is to determine the keyword information corresponding to the aforementioned user call text information as user call keyword information. Specifically, the aforementioned user call keyword information can represent the various keywords included in the aforementioned user call text information.

[0027] The third step is to identify the various value transfer entity information corresponding to the aforementioned user call text information as user call value transfer entity information. Specifically, the aforementioned user call value transfer entity information can characterize the various value transfer entity information included in the aforementioned user call text information.

[0028] The fourth step is to generate enhanced text information for users and enhanced text information for customer service based on the aforementioned user call keyword information and user call value transfer entity information.

[0029] In some optional implementations of certain embodiments, the aforementioned executing entity can generate enhanced user text information and enhanced customer service text information based on the aforementioned user call keyword information and the aforementioned user call value transfer entity information through the following steps: The first step involves generating emotion interval values ​​for each user call text in the aforementioned call text information based on a pre-trained emotion recognition model. These emotion interval values ​​are then grouped into user emotion interval value sets. The pre-trained emotion recognition model can represent an emotion analysis model; for example, it could be an LSTM-based regression model. This pre-trained emotion recognition model takes each user call text as input and outputs the corresponding emotion interval values. The model can include an input embedding layer, an LSTM layer, and an output layer. The input embedding layer takes the call text as input and outputs the corresponding word vector sequence. The LSTM layer takes the corresponding word vector sequence as input and outputs a contextual semantic vector. The output layer takes the contextual semantic vector as input and outputs the emotion interval values. The contextual semantic vector represents a fixed-dimensional vector output after the LSTM layer performs sequence modeling on the input word vector sequence. The pre-trained emotion recognition model can be trained in batches. The emotion interval value information of each call text message in the aforementioned user call text information can represent the emotion score corresponding to each statement included in the call text information. Here, the specific value of the emotion score is not limited; for example, the emotion score can be 0.3. In practice, firstly, the aforementioned executing entity can input each call text message in the aforementioned user call text information into the aforementioned pre-trained emotion recognition model to obtain each emotion interval value information. Then, the obtained emotion interval value information is determined as a user emotion interval value information group.

[0030] The second step involves generating user-enhanced text information corresponding to the aforementioned user call text information, user call keyword information, user call value transfer entity information, user emotion interval value information group, and user call location information. The user call location information represents the order in which each piece of call text appears in the user call text information. The specific numerical value of this order is not limited; for example, the order could be the first. In practice, the executing entity can use a structured integration method to integrate the aforementioned user call text information, user call keyword information, user call value transfer entity information, user emotion interval value information group, and user call location information according to a structured format of "text-keyword-entity-emotion-dialogue location," obtaining the integrated result as the user-enhanced text information.

[0031] Optionally, after step 101 above, the executing entity may also perform the following steps: The first step involves treating each piece of customer service call text as text information and performing the aforementioned recognition steps to obtain the corresponding keyword information (customer service call keyword information) and value transfer entity information (customer service call value transfer entity information). The customer service call keyword information represents the keywords included in the customer service call text information. The customer service call value transfer entity information represents the various value transfer entities included in the customer service call text information.

[0032] The second step involves determining a preset threshold as the customer service emotion interval value for each message in the aforementioned customer service call text information, thus creating a customer service emotion interval value group. The preset threshold can represent a pre-defined score value. The preset threshold can be 0.1. The customer service emotion interval value information can represent the emotion score corresponding to each statement included in the aforementioned customer service call text information. The customer service emotion interval value information can also be 0.1.

[0033] The third step involves generating enhanced customer service text information based on the aforementioned customer service call text information, customer service call keyword information, customer service call value transfer entity information, customer service emotion interval value information group, and customer service call location information. The customer service call location information represents the order in which each piece of call text appears in the customer service call text information. The specific numerical value of this order is not limited here; for example, the order could be 2. In practice, the executing entity can integrate the aforementioned customer service call text information, customer service call keyword information, customer service call value transfer entity information, customer service emotion interval value information group, and customer service call location information using a structured integration method, following a structured format of "text-keyword-entity-emotion-dialogue location," to obtain the integrated result as the enhanced customer service text information.

[0034] Step 102: Based on the user call location information and customer service call location information of the corresponding target call text information, the user enhanced text information and customer service enhanced text information are fused to obtain fused call text information.

[0035] In some embodiments, the executing entity can perform a fusion process on the user-enhanced text information and the customer service-enhanced text information based on the user call location information and the customer service call location information corresponding to the target call text information, to obtain fused call text information. The fused call text information can represent the call text obtained after arranging and fusing the user-enhanced text information and the customer service-enhanced text information according to the user call location information and the customer service call location information. In practice, firstly, the executing entity can sort the user-enhanced text information and the customer service-enhanced text information in ascending order based on the user call location information and the customer service call location information. Then, the ascending-ordered user-enhanced text information and customer service-enhanced text information are determined as the fused call text information.

[0036] Step 103: Generate an emotional dialogue block information set based on the fused call text information, preset constraint information, and preset dialogue selection conditions.

[0037] In some embodiments, the executing entity may generate an emotion dialogue block information set based on the fused call text information, preset constraint information, and preset dialogue selection conditions. The preset constraint information may characterize a function constructed based on a linear weighted sum method for determining the comprehensive score of the boundaries of the aforementioned topic dialogue blocks.

[0038] In some optional implementations of certain embodiments, the aforementioned executing entity can generate a set of emotion dialogue block information based on the aforementioned fused call text information, preset constraint information, and preset dialogue selection conditions through the following steps: The first step is to generate a set of topic dialogue block information based on the aforementioned merged call text information and preset constraint information. The topic dialogue block information in this set can represent various call text information related to the same customer complaint topic, segmented from the merged call text information according to the preset constraint information.

[0039] The second step involves generating an emotion dialogue block information set based on the aforementioned set of topic dialogue block information and preset dialogue selection criteria. The preset dialogue selection criteria can be used to determine whether multiple consecutive adjacent topic dialogue blocks can be merged into one emotion dialogue block. For example, the preset dialogue selection criteria may include, but are not limited to: the length of each of two adjacent topic dialogue block information is less than a first threshold, and the average value of the target emotion interval value of each statement in each of the two adjacent topic dialogue block information is less than a second threshold. The length of each topic dialogue block can represent the number of statements contained in the topic dialogue block information. The first threshold can be 4. The second threshold can be 0.3. In practice, the executing entity can merge multiple adjacent topic dialogue block information that meet the preset dialogue selection criteria in the aforementioned topic dialogue block information set into one emotion dialogue block, obtaining each emotion dialogue block as an emotion dialogue block information set.

[0040] In addressing the aforementioned technical issues using technical solutions, and considering the application scenario—processing complex and lengthy complaint texts from users urgently needing to withdraw cash when ATMs swallow cards, cash, or fail to process withdrawals—the following technical problem often arises: Directly analyzing these complex and lengthy complaint texts results in excessively long and semantically complex input content for the large language model, leading to prolonged processing time and lower accuracy in generating customer complaint information. This can also result in the ATM not being remotely shut down in a timely manner, potentially causing hardware damage due to malfunction. Given the following requirements for this application scenario: adaptability to long complaint text processing, adaptability to complex complaint text processing, and adaptability to short processing timeframes, we have decided to adopt the following solution: In some optional implementations of certain embodiments, the aforementioned execution entity can generate a set of topic dialogue block information based on the aforementioned fused call text information and preset constraint information through the following steps: The first step is to determine the current topic dialogue block information as the preset topic dialogue block information that meets the preset initial selection criteria from the preset topic dialogue block information set. Here, the aforementioned merged call text information includes various merged dialogue texts. Each merged dialogue text can represent a statement included in the aforementioned merged call text information. The preset topic dialogue block information in the aforementioned preset topic dialogue block information set can represent pre-defined topic dialogue blocks. The aforementioned topic dialogue block can represent a set of consecutive statements surrounding the same topic. Here, the specific type of the aforementioned topic is not limited; for example, the topic could be "ATM machine swallowing cash." The aforementioned preset initial selection criteria can be that the topic dialogue block information is the first topic dialogue block in the aforementioned preset topic dialogue block information set. It should be noted that the aforementioned preset topic dialogue block information set can be an initialized set containing multiple empty topic dialogue block information.

[0041] The second step is to add the merged dialogue texts that meet the preset text selection criteria from the aforementioned merged call text information to the aforementioned current topic dialogue block information. The preset text selection criteria can be that the merged dialogue texts are the first and second items in the aforementioned merged call text information.

[0042] The third step is to identify the current topic dialogue block information as the target topic dialogue block information.

[0043] The fourth step is to generate a set of topic dialogue block information based on the above-mentioned fused call text information, preset constraint information and target topic dialogue block information.

[0044] In some optional implementations of certain embodiments, the aforementioned execution entity may generate a topic dialogue block information set based on the aforementioned fused call text information, preset constraint information, and target topic dialogue block information through the following steps: The first step, based on the target topic dialogue block information, is to perform the following segmentation steps for the fusion dialogue texts that meet the selection order conditions in the above-mentioned fusion call text information: The first sub-step involves, in response to determining that the aforementioned fused dialogue text does not meet the aforementioned preset text selection criteria, identifying the fused dialogue text within the target topic dialogue block information that meets the preset dialogue text selection criteria as the target fused dialogue text. The aforementioned selection order criterion can be selecting the fused dialogue text to be divided at the current moment. The aforementioned preset dialogue text selection criteria can be the fused dialogue text with the highest order in the aforementioned target topic dialogue block information.

[0045] The second sub-step generates keyword result information, entity result information, interval value result information, and semantic similarity information based on the fused dialogue text and the target fused dialogue text. The keyword result information represents a score indicating whether the target keyword information in the fused dialogue text and the target fused dialogue text are identical. This keyword result information can be represented as 1 or 0. The target keyword information can represent either the user call keyword information or the customer service call keyword information. The entity result information represents a score indicating whether there is an intersection between the target value transfer entity information in the fused dialogue text and the target fused dialogue text. This entity result information can be represented as 1 or 0. The target value transfer entity information can include both the user call value transfer entity information and the customer service call value transfer entity information. The interval value result information represents a score indicating whether the target emotion interval value information between the fused dialogue text and the target fused dialogue text meets a preset emotion comparison threshold. The target emotion interval value information can include both the user emotion interval value information group and the customer service emotion interval value information group. The preset emotion comparison threshold can be 0.4. The interval value result information can be represented as 1 or 0. The aforementioned semantic similarity information represents the score corresponding to whether the semantic similarity between the fused dialogue text and the target fused dialogue text meets a preset semantic comparison threshold. The preset semantic comparison threshold can be 0.5. The semantic similarity information can represent 1 or 0. In practice, firstly, the executing entity can use a string matching algorithm to determine whether the target keyword information between the fused dialogue text and the target fused dialogue text is the same. If the target keyword information is determined to be the same, the keyword result information is set to 0. If the target keyword information is determined to be different, the keyword result information is set to 1. Then, based on a regular expression matching algorithm, it is determined whether there is an intersection between the target value transfer entity information between the fused dialogue text and the target fused dialogue text. If there is an intersection, the entity result information is set to 0. If there is no intersection, the entity result information is set to 1. Next, the difference between the emotion score corresponding to the aforementioned fused dialogue text and the emotion score corresponding to the aforementioned target fused dialogue text is calculated as the target difference. In response to determining that the aforementioned target difference is greater than the aforementioned preset emotion comparison threshold, the aforementioned interval value result information is determined to be 1. In response to determining that the aforementioned target difference is less than or equal to the aforementioned preset emotion comparison threshold, the aforementioned interval value result information is determined to be 0.Then, using the Sentence-BERT model, embedding processing is performed on the fused dialogue text and the target fused dialogue text to obtain two vectors corresponding to the fused dialogue text and the target fused dialogue text, respectively. Next, the cosine similarity between the two vectors is calculated. If the cosine similarity is greater than the preset semantic comparison threshold, the semantic similarity information is set to 1. If the cosine similarity is less than or equal to the preset semantic comparison threshold, the semantic similarity information is set to 0.

[0046] The third sub-step involves generating boundary value information based on the aforementioned keyword result information, entity result information, interval value result information, preset constraint information, and semantic similarity information. The boundary value information represents the score used to determine whether the fused dialogue text belongs to the target topic dialogue block. In practice, the executing entity can input the aforementioned keyword result information, entity result information, interval value result information, and semantic similarity information into the preset constraint information to obtain the boundary value information corresponding to the fused dialogue text. As an example, the preset constraint information can be: .

[0047] in, It can characterize the boundary value information corresponding to the above-mentioned fused dialogue text. This can represent the information of the above keyword results. The weighting coefficients can represent the keyword results information mentioned above. It can characterize the above entity result information. The weighting coefficients can represent the weighting coefficients of the above entity results information. This can represent the information of the above interval value results. The weighting coefficients can represent the information of the above interval value results. It can represent the semantic similarity information mentioned above. The weight coefficients can represent the semantic similarity information mentioned above.

[0048] The fourth sub-step involves adding the fused dialogue text to the target topic dialogue block information in response to determining that the boundary value information is less than or equal to a preset value, thereby obtaining the updated target topic dialogue block information. The preset value can be 0.6.

[0049] The second step is to determine, in response to the determination that there is no fusion dialogue text in the above fusion call text information that satisfies the above selection order conditions, to determine the obtained updated target topic dialogue block information as a topic dialogue block information set.

[0050] The above-described technical solution, as an inventive point of this disclosure, solves technical problem two: "The generation of customer complaint information is time-consuming and inaccurate, resulting in the failure to remotely shut down the ATM device in a timely manner, leading to hardware damage to the ATM device due to malfunction." The reasons for the long generation time and low accuracy of customer complaint information are as follows: Directly analyzing the complex and long complaint text results in excessively long input content and complex semantic information for the large language model, leading to long generation time and low accuracy of customer complaint information, resulting in the failure to remotely shut down the ATM device in a timely manner, leading to hardware damage to the ATM device due to malfunction. Solving these factors can shorten the generation time of customer complaint information and improve its accuracy. To achieve this effect, the customer complaint information processing method of this disclosure for long text calls first adds the fused dialogue text that meets the preset text selection conditions from the fused call text to the current topic dialogue block information, and then determines the current topic dialogue block information as the target topic dialogue block information. Subsequently, based on the aforementioned fused dialogue text and the aforementioned target fused dialogue text, keyword result information, entity result information, interval value result information, and semantic similarity information between the fused dialogue text and the aforementioned target fused dialogue text can be obtained. Based on the aforementioned keyword result information, entity result information, interval value result information, preset constraint information, and semantic similarity information, boundary value information can be obtained. Therefore, the aforementioned target topic dialogue blocks can be divided according to the aforementioned boundary value information, resulting in a set of topic dialogue block information with consistent themes. Subsequently, customer complaint information is generated based on the aforementioned topic dialogue block information set, which can shorten the time required for customer complaint information generation and improve accuracy. Therefore, it can reduce the situation where the aforementioned ATM equipment cannot be remotely shut down in a timely manner, and can reduce the situation where the hardware of the ATM equipment is damaged due to malfunction.

[0051] Optionally, after step 103 above, the executing entity may also perform the following steps: The first step involves, in response to determining that the boundary value information is greater than the preset value, adding the fused dialogue text to the preset topic dialogue block information set that meets the preset order condition. The preset order condition can represent a preset topic dialogue block in the preset topic dialogue block set that is located after the current topic dialogue block in the current topic dialogue block order and whose difference from the current topic dialogue block order is 1.

[0052] The second step is to determine that there is a fusion dialogue text in the above-mentioned fusion call text information that meets the above-mentioned selection order conditions, and then take the topic dialogue block information in the above-mentioned preset topic dialogue block information set that meets the above-mentioned preset order conditions as the target topic dialogue block information, and then perform the above-mentioned division steps again.

[0053] Step 104: Generate customer complaint reason information and warning level information based on the emotional dialogue block information set, the preset customer complaint prompt word information set, and the preset warning level information set.

[0054] In some embodiments, the executing entity can generate customer complaint reason information and warning level information based on the aforementioned emotional dialogue block information set, preset customer complaint prompt word information set, and preset warning level information set. The preset customer complaint prompt word information in the preset customer complaint prompt word information set can represent statements indicating the reason for the target user's complaint. The specific content of these statements is not limited; for example, the statement could be "ATM machine is lagging." The preset warning level information in the preset warning level information set can represent the name of the warning level classification. The specific content of these preset warning level information is not limited; for example, the preset warning level information could be "service dissatisfaction."

[0055] In addressing the aforementioned technical issues using technical solutions, and considering the application scenario—processing simple complaint texts regarding user lag or slow response times when using ATMs—the following technical problem often arises: When generating customer complaint information from the emotional dialogue blocks of these simple complaint texts using a unified, fixed large language model, resource allocation for processing simple complaint texts becomes mismatched. Using complex generation methods for customer complaint information results in high computational resource consumption and long processing times, leading to slower timeliness in handling ATM software faults. This sustained high load on the software layer can accelerate the aging of ATM hardware components and shorten the lifespan of the ATM hardware. Therefore, this application scenario requires the following characteristics: adaptability to simple complaint text processing, low-cost processing, and reasonable allocation of processing resources. We have decided to adopt the following solution: In some optional implementations of certain embodiments, the aforementioned execution entity can generate customer complaint reason information and warning level information by following the steps described above: based on the aforementioned emotional dialogue block information set, preset customer complaint prompt word information set, and preset warning level information set. The first step involves dividing the aforementioned emotional dialogue block information set according to a preset division method, resulting in dialogue block complexity division groups. These groups can represent simple, complex, or extremely complex dialogue block information sets. The preset division method can be based on the number of statements in each emotional dialogue block information set. Each simple dialogue block information set in the simple dialogue block information set can be characterized by having fewer statements than a first preset round threshold. This first preset round threshold can be 10. Each complex dialogue block information set in the complex dialogue block information set can be characterized by having more than or equal to the first preset round threshold and less than a second preset round threshold. This second preset round threshold can be 20. Each extremely complex dialogue block information set in the extremely complex dialogue block information set can be characterized by having more than or equal to the second preset round threshold. In practice, firstly, in response to determining that the number of statements in the aforementioned emotional dialogue block information is less than a first preset round threshold, the executing entity can classify the aforementioned emotional dialogue block information as simple dialogue block information. Then, in response to determining that the number of statements in the aforementioned emotional dialogue block information is greater than or equal to the first preset round threshold and less than a second preset round threshold, the executing entity can classify the aforementioned emotional dialogue block information as complex dialogue block information. Finally, in response to determining that the number of statements in the aforementioned emotional dialogue block information is greater than or equal to the second preset round threshold, the executing entity can classify the aforementioned emotional dialogue block information as extremely complex dialogue block information. It should be noted that the aforementioned set of simple dialogue block information, the aforementioned set of complex dialogue block information, or the aforementioned set of extremely complex dialogue block information are not empty.

[0056] The second step involves generating a value flow entity graph based on the aforementioned set of preset customer complaint prompts and the aforementioned set of preset warning levels. This value flow entity graph can represent a structured knowledge graph related to value flow. In practice, the executing entity can use the preset customer complaint prompts from the preset customer complaint prompts set as entity nodes and the preset warning levels from the preset warning levels set as attribute values ​​for these entity nodes. Through the mapping between entity nodes and attributes, the value flow entity graph can be constructed. For example, the value flow entity graph could include "Entity node: ATM machine swallowing cash," with the corresponding attribute value being "Preset warning level information: Service dissatisfaction."

[0057] The third step involves generating a template information set based on the aforementioned value flow entity graph information. This template information set can include a first template information set and a second template information set. The first template information in the first template information set can represent fixed-format templates corresponding to different relationship types. Here, the specific content of the relationship types and the first template information is not limited; for example, the relationship type could be "leading to," and the first template information could be "{subject} performs {behavior} through {method}, thereby leading to {result}." The second template information in the second template information set can represent fixed-format templates corresponding to the aforementioned preset warning level information. Here, the specific content of the second template information is not limited; for example, the second template information could be "{risk level} refers to the customer's {emotional state} and {core demand} when expressing themselves. Typical manifestations include {specific scenario examples}. Such situations usually imply {potential impact / follow-up actions}." In practice, firstly, the executing entity can use the first mapping relationship to match the corresponding first template text for each relationship type in the aforementioned value flow entity graph information, thus obtaining the first template information set. Then, through the second mapping relationship, the corresponding second template text is matched for each preset warning level information included in the above-mentioned value flow entity graph information, resulting in a second template information set. Finally, the above-mentioned first template information set and the above-mentioned second template information set are determined as the template information set. The above-mentioned first mapping relationship can represent the correspondence between each relation type in the above-mentioned value flow entity graph information and the above-mentioned first template text. The above-mentioned second mapping relationship can represent the correspondence between each preset warning level information included in the above-mentioned value flow entity graph information and the above-mentioned second template text. For example, the first mapping relationship can be "relationship type: leads to" corresponding to "template text: {subject} leads to {result}", and the second mapping relationship can be "preset warning level information: service dissatisfaction" corresponding to "template text: {service dissatisfaction} refers to a customer being emotionally agitated and demanding an immediate solution to the ATM problem (such as "money was swallowed, handle it immediately")".

[0058] The fourth step involves generating a set of entity association path information based on the aforementioned value flow entity graph information. The entity association path information in this set represents the causal relationships corresponding to the aforementioned value flow entity graph information. The specific content of this entity association path information is not limited here; for example, it could be "ATM machine swallows cash - leading to - customer financial loss." In practice, the executing entity can use a breadth-first search algorithm to perform knowledge extraction processing on the aforementioned value flow entity graph information to obtain the set of entity association path information.

[0059] Fifth, based on the aforementioned entity association path information set and template information set, generate a semantically extended text information set. This semantically extended text information set may include a first semantically extended text information set and a second semantically extended text information set. The first semantically extended text information in the first semantically extended text information set can represent the semantic interpretation of each preset customer complaint prompt word in the aforementioned preset customer complaint prompt word information set. The second semantically extended text information in the second semantically extended text information set can represent the semantic interpretation of each preset warning level information in the aforementioned preset warning level information set. Here, the specific content of the aforementioned first and second semantically extended text information is not limited. For example, the first semantically extended text information could be that when a customer deposits money at an ATM, the machine displays a successful transaction but the funds are not actually credited to the account, or the machine malfunctions and swallows the customer's cash without printing a receipt, causing the customer's funds to be temporarily unavailable, leading to customer anxiety and complaints (assuming the preset customer complaint prompt is "ATM swallowed cash"). The second semantically extended text information could be that the customer is emotionally agitated and explicitly demands that the bank immediately resolve the ATM swallowed cash problem and recover the swallowed funds, such as "My money has been swallowed, deal with it immediately, or I will call the police" (assuming the preset warning level is "service dissatisfaction"). In practice, the aforementioned executing entity can use a template-based text generation method to fuse the aforementioned entity association path information set and the aforementioned template information set to obtain the semantically extended text information set.

[0060] Step 6: Based on the semantically extended text information mentioned above, invoke the first, second, or third customer complaint warning information generation model to perform semantic analysis on the above dialogue block complexity division group, obtaining customer complaint cause information and warning level information. The first customer complaint warning information generation model can represent Qwen2.5-0.5B-Instruct. This model takes the above simple dialogue block information set as input and outputs the first type of customer complaint cause information and the first type of warning level division information. The training method for this model can be batch training. The first type of customer complaint cause information represents the customer complaint cause corresponding to the above simple dialogue block information set. The customer complaint cause can represent the text corresponding to the complaint cause of the target user. Here, the specific content of the text is not limited; for example, the text could be "ATM swallowing cash." The first type of warning level division information represents the warning level corresponding to the above simple dialogue block information set. The warning level can represent the first warning level, second warning level, third warning level, fourth warning level, or fifth warning level. The first warning level can represent that the warning value of the target call text is less than a first preset threshold. The second warning level can represent that the warning value of the target call text is greater than or equal to the first preset threshold and less than the second preset threshold. The third warning level can represent that the warning value of the target call text is greater than or equal to the second preset threshold and less than the third preset threshold. The fourth warning level can represent that the warning value of the target call text is greater than or equal to the third preset threshold and less than the fourth preset threshold. The fifth warning level can represent that the warning value of the target call text is greater than or equal to the fourth preset threshold. The first preset threshold can be 2.0. The second preset threshold can be 3.0. The third preset threshold can be 4.0. The fourth preset threshold can be 4.5. The warning value can represent the score corresponding to the warning level of the target call text information. The second customer complaint warning information generation model can represent Qwen2.5-7B-Instruct. The second customer complaint warning information generation model can take the complex dialogue block information set as input and the second type of customer complaint reason information and the second type of warning level classification information as output. The second type of customer complaint reason information described above can characterize the customer complaint reason corresponding to the aforementioned complex dialogue block information set. The second type of early warning level classification information described above can characterize the early warning classification level corresponding to the aforementioned complex dialogue block information set. The third customer complaint early warning information generation model described above can characterize Qwen2.5-14B-Instruct. The third customer complaint early warning information generation model described above can take the aforementioned extremely complex dialogue block information set as input and take the third type of customer complaint reason information and the third type of early warning level classification information as output.The aforementioned third type of customer complaint cause information can characterize the customer complaint cause corresponding to the aforementioned extremely complex dialogue block information set. The aforementioned third type of warning level classification information can characterize the warning classification level corresponding to the aforementioned extremely complex dialogue block information set. In practice, in response to detecting that the aforementioned dialogue block complexity classification group is a simple dialogue block information set, the aforementioned simple dialogue block information set is input into the aforementioned first customer complaint warning information generation model to obtain the first type of customer complaint cause information and the first type of warning level classification information. In response to detecting that the aforementioned dialogue block complexity classification group is a complex dialogue block information set, the aforementioned complex dialogue block information set is input into the aforementioned second customer complaint warning information generation model to obtain the second type of customer complaint cause information and the second type of warning level classification information. In response to detecting that the aforementioned dialogue block complexity classification group is an extremely complex dialogue block information set, the aforementioned extremely complex dialogue block information set is input into the aforementioned third customer complaint warning information generation model to obtain the third type of customer complaint cause information and the third type of warning level classification information. Then, the aforementioned first type of customer complaint cause information, the aforementioned second type of customer complaint cause information, and the aforementioned third type of customer complaint cause information are determined as customer complaint cause information. Subsequently, the aforementioned first-category warning level classification information, the aforementioned second-category warning level classification information, and the aforementioned third-category warning level classification information are determined as warning level information.

[0061] The above-described technical solution, as an inventive point of this disclosure, solves technical problem three: "The excessive computational resources consumed and the long processing time result in low timeliness of handling ATM device software faults." The reasons for the excessive computational resources consumed and the long processing time, leading to low timeliness of handling ATM device software faults, are as follows: When generating customer complaint information from the emotional dialogue block information set of the aforementioned simple complaint text using a unified and fixed large language model, a mismatch occurs in the processing resources for the simple complaint text. The use of complex generation methods to generate customer complaint information results in excessive computational resources consumed and long processing time, leading to low timeliness of handling ATM device software faults. This continuous high load at the software level can accelerate the aging of ATM device hardware components and shorten the lifespan of the ATM device hardware. Solving these factors can reduce computational resources consumed and shorten processing time, thereby improving the timeliness of handling ATM device software faults. To achieve this effect, the customer complaint information processing method disclosed herein for long text calls first divides the aforementioned emotional dialogue block information set according to the preset division method, thereby obtaining a simple dialogue block information set, a complex dialogue block information set, and an extremely complex dialogue block information set. Then, based on the aforementioned preset customer complaint prompt word information and preset semantic information set, the aforementioned first customer complaint warning information generation model, the aforementioned second customer complaint warning information generation model, and the aforementioned third customer complaint warning information generation model are used to process the simple dialogue block information set, the complex dialogue block information set, and the extremely complex dialogue block information set, respectively, to obtain first-type customer complaint reason information and first-type warning level information corresponding to the simple dialogue block information set, second-type customer complaint reason information and second-type warning level information corresponding to the complex dialogue block information set, and third-type customer complaint reason information and third-type warning level information corresponding to the extremely complex dialogue block information set. This reduces computational resources and shortens processing time, thereby improving the timeliness of handling ATM device software faults.

[0062] Step 105: Generate customer complaint information for the corresponding target terminal device based on the customer complaint reason information and warning level information.

[0063] In some embodiments, the executing entity may generate customer complaint information corresponding to the target terminal device based on the customer complaint reason information and the warning level information. The customer complaint information may represent the customer complaint reason information and warning level information corresponding to the target terminal device.

[0064] In some optional implementations of certain embodiments, the aforementioned executing entity may generate customer complaint information corresponding to the aforementioned target terminal device through the following steps: based on the aforementioned customer complaint reason information and the aforementioned warning level information: The first step is to fuse the various customer complaint reason texts included in the aforementioned customer complaint reason information to obtain the fused customer complaint reason information corresponding to the aforementioned target call text information. This fused customer complaint reason information can represent the various customer complaint reason texts obtained after fusion processing. In practice, the executing entity can fuse the various customer complaint reason texts included in the aforementioned customer complaint reason information by deduplication and concatenation to obtain the fused customer complaint reason information corresponding to the aforementioned target call text information.

[0065] The second step involves fusing the various warning levels included in the aforementioned warning level information according to the preset function information, thereby obtaining the warning level fusion information corresponding to the target call text information. This warning level fusion information characterizes the warning level corresponding to the target call text information. In practice, the executing entity can input the various warning levels included in the aforementioned warning level information into the preset function information to obtain the warning level fusion information corresponding to the target call text information. As an example, the preset function information can be: .

[0066] in, It can characterize the obtained early warning level fusion information. The weighting coefficients that characterize the various warning levels can be as follows: For a warning level designated as Level 1, the weighting coefficient can be 0.1; for a warning level designated as Level 2, the weighting coefficient can be 0.2; for a warning level designated as Level 3, the weighting coefficient can be 0.4; for a warning level designated as Level 4, the weighting coefficient can be 0.7; and for a warning level designated as Level 5, the weighting coefficient can be 1. It can represent the number of times the above-mentioned warning levels occur. It can represent the various warning levels mentioned above.

[0067] The third step is to determine the above-mentioned customer complaint reason information and the above-mentioned early warning level information as customer complaint information.

[0068] Optionally, after step 105 above, the executing entity may also perform the following steps: The first step involves generating proportional information corresponding to the aforementioned customer complaint reason fusion information and preset time interval information. The preset time interval information can be the quarter in which the received target call text information is located. The proportional information represents the proportion of any one type of customer complaint reason among all customer complaint reasons. In practice, firstly, the executing entity can count the total number of all customer complaint reasons within the preset time interval information. Then, for each customer complaint reason in the aforementioned customer complaint reason fusion information, the executing entity can determine the number of customer complaint reasons identical to the aforementioned customer complaint reason among all the aforementioned customer complaint reasons as a first quantity. Next, the ratio between the first quantity and the total number of all the aforementioned customer complaint reasons is determined as the ratio corresponding to the aforementioned customer complaint reason. Finally, each of the obtained ratios is determined as proportional information.

[0069] The second step involves displaying a warning message corresponding to the aforementioned customer complaint cause fusion information, based on the aforementioned ratio information and preset threshold conditions. The preset threshold condition can be that the ratio of the customer complaint causes in the aforementioned customer complaint cause fusion information is greater than a preset ratio threshold. The preset ratio threshold can be 30%. The warning message can represent an output prompt statement, for example, "Please prioritize handling complaints related to 'ATM machine swallowing cash'." In practice, in response to the determination that there is a ratio in the aforementioned customer complaint cause fusion information that meets the aforementioned preset threshold conditions, the executing entity can control the alarm display to show the aforementioned warning message. The alarm display can represent the display screen of the communication connection. It should be noted that the aforementioned communication connection can include, but is not limited to, 3G / 4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other currently known or future developed communication methods.

[0070] In some optional implementations of certain embodiments, the aforementioned executing entity may determine the processing method corresponding to the aforementioned customer complaint information based on the aforementioned warning level fusion information through the following steps: The first step, based on the above-mentioned integrated warning level information, is to perform the following steps: The first sub-step involves determining the aforementioned warning level fusion information as the first warning level, and then setting the first preset processing method as the processing method corresponding to the aforementioned customer complaint information. The first preset processing method can be to process the complaint within one week.

[0071] The second sub-step involves determining the aforementioned warning level fusion information as a second warning level, and then defining the second preset processing method as the processing method corresponding to the aforementioned customer complaint information. The second preset processing method can be implemented under a first preset time constraint. The first preset time constraint can be less than or equal to 48 hours.

[0072] The third sub-step involves determining the aforementioned warning level fusion information as the third warning level, and then defining the third preset processing method as the processing method corresponding to the aforementioned customer complaint information. This third preset processing method can be implemented under a second preset time constraint. The second preset time constraint can be less than or equal to 24 hours.

[0073] The fourth sub-step involves determining the aforementioned warning level fusion information as the fourth warning level, and then determining the fourth preset processing method as the processing method corresponding to the aforementioned customer complaint information. Specifically, the fourth preset processing method can be processing under a third preset time constraint, which can be less than or equal to 12 hours.

[0074] The fifth sub-step involves determining the aforementioned warning level fusion information as the fifth warning level, and then setting the fifth preset processing method as the processing method corresponding to the aforementioned customer complaint information. This fifth preset processing method can be used for processing at the current moment.

[0075] Step 106: In response to determining that the warning level information of the target terminal device representing the customer complaint information is the fifth warning level, shut down the target terminal device.

[0076] In some embodiments, in response to determining that the warning level information of the aforementioned customer complaint information characterizing the target terminal device is the fifth warning level, the aforementioned executing entity may shut down the aforementioned target terminal device. In practice, the aforementioned executing entity may send a shutdown command to the aforementioned target terminal device to shut it down. It should be noted that the connection between the aforementioned executing entity and the aforementioned target terminal device can be a communication connection. It should be pointed out that the aforementioned communication connection may include, but is not limited to, 3G / 4G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultrawideband) connection, and other communication methods that are currently known or will be developed in the future.

[0077] The above-described embodiments of this disclosure have the following beneficial effects: A customer complaint information processing method applied to long text calls, according to some embodiments of this disclosure, can reduce computer resource consumption and shorten processing time, improve the timeliness of detecting hardware device malfunctions, and thus improve the accuracy of semantic analysis of the complete complaint call content and the user experience. The reasons for the large computer resource consumption, long processing time, poor timeliness of detecting hardware device malfunctions, low accuracy of semantic analysis of the complete complaint call content, and poor user experience are as follows: The maximum input length of a large model is highly correlated with computer resources. When processing long call texts, the computer resource consumption is excessive and the processing time is long, resulting in poor timeliness of detecting hardware device malfunctions, which may lead to hardware device damage due to failure to remotely shut down the hardware device in a timely manner. Furthermore, randomly truncating excessively long complaint call texts can cause missing contextual semantics, leading to low accuracy of semantic analysis of the complete complaint call content, resulting in inaccurate information about the hardware device's status, causing incorrect shutdown of the hardware device and affecting the user experience, or failure to shut down the device in a timely manner due to malfunction, leading to device damage. Based on this, some embodiments of the present disclosure of a customer complaint information processing method for long text calls firstly, in response to receiving target call text information for a target terminal device, modifying the target call text information to obtain user-enhanced text information and customer service-enhanced text information. This allows for enhancement processing of the target call text information, resulting in enhanced user-enhanced call text and enhanced customer service call text. Secondly, based on the user call location information and customer service call location information corresponding to the target call text information, fusing the user-enhanced text information and the customer service-enhanced text information to obtain fused call text information. This yields the fused call text. Then, based on the fused call text information, preset constraint information, and preset dialogue selection conditions, an emotional dialogue block information set is generated. This yields each emotional dialogue block. Next, based on the emotional dialogue block information set, a preset customer complaint prompt word information set, and a preset warning level information set, customer complaint reason information and warning level information are generated. This yields the customer complaint reason and warning level for each emotional dialogue block in the emotional dialogue block set. Finally, based on the customer complaint reason information and the warning level information, customer complaint information corresponding to the target terminal device is generated. Thus, the customer complaint information from the aforementioned call text can be obtained. Finally, in response to determining that the aforementioned customer complaint information represents the warning level information of the aforementioned target terminal device as the fifth warning level, the aforementioned target terminal device is shut down. This is because the process does not involve simply and crudely truncating the complaint call text and then using a large model to analyze and process the split complaint call text. Instead, the aforementioned call text information undergoes enhanced fusion processing to obtain fused call text information.Then, the aforementioned fused call text information is segmented to obtain a set of segmented emotional dialogue blocks. Finally, based on the aforementioned emotional dialogue block information set, the preset customer complaint prompt word information set, and the preset warning level information set, customer complaint information corresponding to the aforementioned call text information is obtained. Furthermore, the aforementioned target call text information can be segmented to obtain emotional dialogue block information sets for each emotional dialogue block with similar themes and emotions. This allows for segment-by-segment parsing of each emotional dialogue block with similar themes and emotions, thereby improving parsing speed and reducing the time required to generate customer complaint information. Therefore, it can reduce computer resource consumption and shorten processing time, improve the timeliness of detecting hardware malfunctions, and thus improve the accuracy of semantic analysis of the complete complaint call content and the user experience.

[0078] Further reference Figure 2 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of a customer complaint information processing method applied to long text calls. These device embodiments are similar to... Figure 1 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.

[0079] like Figure 2 As shown, some embodiments of the customer complaint information processing device 200 for long text calls include: a modification unit 201, a fusion unit 202, a first generation unit 203, a second generation unit 204, a third generation unit 205, and a closing unit 206. The system includes the following components: Modification unit 201, configured to modify target call text information received for a target terminal device to obtain enhanced user text information and enhanced customer service text information; Fusion unit 202, configured to fuse enhanced user text information and enhanced customer service text information based on user call location information and customer service call location information corresponding to the target call text information to obtain fused call text information; First generation unit 203, configured to generate an emotional dialogue block information set based on the fused call text information, preset constraint information, and preset dialogue selection conditions; Second generation unit 204, configured to generate customer complaint reason information and warning level information based on the emotional dialogue block information set, preset customer complaint prompt word information set, and preset warning level information set; Third generation unit 205, configured to generate customer complaint information corresponding to the target terminal device based on the customer complaint reason information and the warning level information; and Closing unit 206, configured to close the target terminal device upon determining that the warning level information of the target terminal device represented by the customer complaint information is the fifth warning level.

[0080] It is understandable that the units described in the customer complaint information processing device 200 used for long text calls are related to the reference. Figure 1 The steps in the described method correspond to each other. Therefore, the operations, features, and beneficial effects described above for the method are also applicable to the customer complaint information processing device 200 and its constituent units used in long text calls, and will not be repeated here.

[0081] The following is for reference. Figure 3 It shows a schematic diagram of the structure of an electronic device 300 (e.g., a computing device) suitable for implementing some embodiments of the present disclosure. Figure 3 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.

[0082] like Figure 3 As shown, the electronic device 300 may include a processing unit 301 (e.g., a central processing unit, a graphics processor, etc.), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 302 or a program loaded from a storage device 308 into a random access memory (RAM) 303. The RAM 303 also stores various programs and data required for the operation of the electronic device 300. The processing unit 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to the bus 304.

[0083] Typically, the following devices can be connected to I / O interface 305: input devices 306 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 307 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; storage devices 308 including, for example, magnetic tapes, hard disks, etc.; and communication devices 309. Communication device 309 allows electronic device 300 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 3 An electronic device 300 with various devices is shown; however, it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed alternatively. Figure 3 Each box shown can represent a device or multiple devices as needed.

[0084] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via communication device 309, or installed from storage device 308, or installed from ROM 302. When the computer program is executed by processing device 301, it performs the functions defined in the methods of some embodiments of this disclosure.

[0085] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.

[0086] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.

[0087] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently without being assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs, which, when executed by the electronic device, cause the electronic device to: in response to receiving target call text information for a target terminal device, modify the target call text information to obtain user-enhanced text information and customer service-enhanced text information; fuse the user-enhanced text information and the customer service-enhanced text information according to the user call location information and customer service call location information corresponding to the target call text information to obtain fused call text information; generate an emotional dialogue block information set according to the fused call text information, preset constraint information, and preset dialogue selection conditions; generate customer complaint reason information and warning level information according to the emotional dialogue block information set, a preset customer complaint prompt word information set, and a preset warning level information set; generate customer complaint information corresponding to the aforementioned target terminal device according to the customer complaint reason information and the aforementioned warning level information; and, in response to determining that the warning level information of the aforementioned customer complaint information represents the aforementioned target terminal device as the fifth warning level, shut down the aforementioned target terminal device.

[0088] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, and C++, and 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, as a standalone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0089] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0090] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor, and for example, can be described as: a modification unit, a fusion unit, a first generation unit, a second generation unit, a third generation unit, and a shutdown unit. The names of these units do not necessarily limit the unit itself; for example, an update unit can also be described as a unit that "updates the target call text information received for a target terminal device to obtain user call text information and customer service call text information."

[0091] The functions described above in this document can be performed at least in part by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SoCs), complex programmable logic devices (CPLDs), and so on.

[0092] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.

Claims

1. A method for processing customer complaint information in long text calls, comprising: In response to receiving target call text information for a target terminal device, the target call text information is modified to obtain user-enhanced text information and customer service-enhanced text information; Based on the user call location information and customer service call location information corresponding to the target call text information, the user enhanced text information and the customer service enhanced text information are fused to obtain fused call text information. Based on the fused call text information, preset constraint information, and preset dialogue selection conditions, an emotion dialogue block information set is generated; Based on the set of emotional dialogue block information, the set of preset customer complaint prompt words information, and the set of preset warning level information, generate customer complaint reason information and warning level information; Based on the customer complaint reason information and the warning level information, generate customer complaint information corresponding to the target terminal device; In response to determining that the customer complaint information indicates that the warning level of the target terminal device is the fifth warning level, the target terminal device is shut down.

2. The method according to claim 1, wherein, The step of generating customer complaint information corresponding to the target terminal device based on the customer complaint reason information and the warning level information includes: The various customer complaint reason texts included in the customer complaint reason information are fused together to obtain customer complaint reason fusion information corresponding to the target call text information; Based on preset function information, the various warning classification levels included in the warning level information are fused to obtain the warning level fusion information corresponding to the target call text information; The customer complaint information is determined by combining the fused information of the cause of the complaint and the fused information of the warning level.

3. The method according to claim 1, wherein, Both the user call text information and the customer service call text information include various call text information. as well as The modification process of the target call text information to obtain enhanced user text information and enhanced customer service text information includes: The target call text information is updated to obtain user call text information and customer service call text information; Based on the user call text information and the customer service call text information, generate enhanced user text information and enhanced customer service text information.

4. The method according to claim 3, wherein, The step of generating enhanced user text information and enhanced customer service text information based on the user call text information and the customer service call text information includes: Each piece of call text information in the user's call text information is treated as text information, and the following recognition steps are performed: The text information is processed according to a preset recognition method and preset keyword information to obtain keyword information corresponding to the text information; Based on the pre-trained entity recognition model and the preset entity information, the text information is processed to obtain the value transfer entity information corresponding to the text information; Each keyword information corresponding to the obtained user call text information is determined as user call keyword information; Each value transfer entity information corresponding to the obtained user call text information is determined as user call value transfer entity information; Based on the user call keyword information and the user call value transfer entity information, user enhanced text information and customer service enhanced text information are generated.

5. The method according to claim 4, wherein, The step of generating enhanced user text information and enhanced customer service text information based on the user call keyword information and the user call value transfer entity information includes: Based on the pre-trained emotion recognition model, emotion interval value information corresponding to each call text information in the user's call text information is generated, and each emotion interval value information is used as a user emotion interval value information group. Based on the user call text information, the user call keyword information, the user call value transfer entity information, the user emotion interval value information group, and the user call location information, user enhanced text information corresponding to the user call text information is generated.

6. The method according to claim 5, wherein, The method further includes: Each call text in the customer service call text information is taken as text information, and the recognition step is performed to obtain each keyword information corresponding to the customer service call text information as customer service call keyword information and each value transfer entity information as customer service call value transfer entity information. The preset threshold is determined as the customer service emotion interval value information of each call text information in the customer service call text information, and each customer service emotion interval value information is obtained as a customer service emotion interval value information group. Based on the customer service call text information, the customer service call keyword information, the customer service call value transfer entity information, the customer service emotion interval value information group, and the customer service call location information, customer service enhanced text information corresponding to the customer service call text information is generated.

7. A customer complaint information processing device for long text calls, comprising: The modification unit is configured to modify the target call text information in response to receiving target call text information for a target terminal device, thereby obtaining user-enhanced text information and customer service-enhanced text information. The fusion unit is configured to perform fusion processing on the user enhanced text information and the customer service enhanced text information based on the user call location information and customer service call location information corresponding to the target call text information, to obtain fused call text information. The first generation unit is configured to generate an emotion dialogue block information set based on the fused call text information, preset constraint information and preset dialogue selection conditions; The second generation unit is configured to generate customer complaint reason information and warning level information based on the set of emotional dialogue block information, the set of preset customer complaint prompt words information, and the set of preset warning level information. The third generation unit is configured to generate customer complaint information corresponding to the target terminal device based on the customer complaint reason information and the warning level information; The shutdown unit is configured to shut down the target terminal device in response to determining that the customer complaint information characterizes the warning level information of the target terminal device as the fifth warning level.

8. An electronic device, comprising: One or more processors; A storage device on which one or more programs are stored; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1 to 6.

9. A computer-readable medium having a computer program stored thereon, wherein, When the computer program is executed by a processor, it implements the method as described in any one of claims 1 to 6.

Images