Method, device and equipment for generating verbal skill for coping with client apathy response and medium

Abstract

The invention relates to the field of artificial intelligence, can be applied to the fields of finance and medical treatment, and discloses a speech generation method, device, equipment and medium for coping with customer apathy response, and the method comprises the steps: obtaining a target sales strategy matched with a customer state recognition result from a sales strategy library when it is determined that a customer is in an apathy state; a sales promotion prompt template and a guide question are obtained through the target sales strategy, the historical dialogue text and the guide semantic slot position set, and then customer interest contour information is obtained through customer feedback information corresponding to the guide question; and inputting the historical dialogue text, the customer state recognition result, the sales promotion prompt template and the customer interest contour information into a pre-trained large language model to obtain a target verbal skill output by the large language model. According to the method, the semantic understanding capability, the context sensing capability and the sales strategy knowledge of the large language model are fused, so that the verbal skill content output by the large language model is optimized, and the sales success rate is improved.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology and can be applied to the financial and medical fields. In particular, it relates to a method, apparatus, computer equipment, and storage medium for generating scripts to deal with unfriendly customer responses. Background Technology

[0002] Current large-scale model-driven intelligent dialogue systems have made some progress in automated communication. These systems can be widely applied across various industries; for example, in the financial sector, they can be used to promote insurance products, and in the medical field, they can be used to promote health checkup packages, thereby improving work efficiency and customer experience through automation and intelligent technologies. However, current intelligent dialogue systems face significant bottlenecks in process progression and intent recognition when dealing with indifferent, perfunctory, or unstructured responses (such as "um," "not interested," "not now," "we'll talk about it later," etc.). Traditional systems typically employ keyword-based intent classification methods, but these are prone to getting stuck in a loop or directly terminating the conversation when faced with indifferent responses with low information density or ambiguous semantics, making it difficult to achieve the sales capability of "turning cold into warm."

[0003] Especially in industries with strong sales attributes such as merchandise, insurance, and education and training, it is common for customers to initially show indifference or perfunctory attitudes, while their true purchase intentions are often hidden in subsequent interactions. Current mainstream practices lack sophisticated dialogue strategy design for scenarios of "weak intent and cold response," and lack the ability to structuredly mine and guide potential interests, which easily leads to high missed conversions and high false final judgments, thus reducing the overall conversion rate.

[0004] Furthermore, existing large-scale model application solutions often rely on single-round generation or template-based retries, lacking semantic modeling of the possible motivations behind cold responses, dynamic dialogue strategy planning mechanisms, and psychological response adjustment capabilities. When faced with ambiguous attitudes of "perfunctory but not explicitly rejecting," the system cannot proactively control the pace or guide customers into the stage of genuine expression. This deficiency directly affects the authenticity and credibility of intelligent dialogue systems in human-computer interaction and their ability to close deals.

[0005] Therefore, existing intelligent dialogue systems are unable to accurately identify unstructured weak responses and generate guiding dialogue when dealing with lukewarm customer responses, resulting in low sales success rates. Summary of the Invention

[0006] This invention provides a method, apparatus, computer device, and storage medium for generating dialogue for dealing with lukewarm customer responses. It aims to solve the problem that existing intelligent dialogue systems cannot accurately identify and generate guided dialogue for unstructured weak responses when dealing with lukewarm customer responses, resulting in a low sales success rate.

[0007] In a first aspect, embodiments of the present invention provide a method for generating scripts to respond to a customer's lukewarm response, comprising:

[0008] Obtain historical dialogue text and extract features from the historical dialogue text to obtain historical dialogue features;

[0009] The customer status identification result is obtained based on the historical dialogue features and the pre-trained indifferent response identification model;

[0010] If the customer status identification result indicates that the customer is in a lukewarm state, then the target sales strategy matching the customer status identification result is obtained from the preset sales strategy library, and a sales promotion prompt template is constructed based on the target sales strategy and the historical dialogue text.

[0011] Based on the target sales strategy and the preset set of guiding semantic slots, guiding questions are constructed.

[0012] Obtain customer feedback information corresponding to the guiding question, and obtain customer interest profile information based on the customer feedback information;

[0013] The historical dialogue text, the customer status recognition result, the sales promotion prompt template, and the customer interest profile information are input into a pre-trained large language model to obtain the target script output by the large language model.

[0014] Secondly, embodiments of the present invention provide a script generation apparatus for responding to a lukewarm response from a customer. The apparatus is used to execute the script generation method for responding to a lukewarm response from a customer as described in the first aspect. The apparatus includes:

[0015] The dialogue feature extraction unit is used to acquire historical dialogue text and extract features from the historical dialogue text to obtain historical dialogue features.

[0016] A customer status identification unit is used to obtain customer status identification results based on the historical dialogue features and a pre-trained indifferent response identification model;

[0017] The sales promotion prompt unit is used to retrieve a target sales strategy that matches the customer status identification result from a preset sales strategy library if the customer status identification result determines that the customer is in a lukewarm state, and to construct a sales promotion prompt template based on the target sales strategy and the historical dialogue text.

[0018] The guiding question construction unit is used to construct guiding questions based on the target sales strategy and a preset set of guiding semantic slots;

[0019] The customer intent mining unit is used to obtain customer feedback information corresponding to the guiding question, and to obtain customer interest profile information based on the customer feedback information.

[0020] The target dialogue generation unit is used to input the historical dialogue text, the customer status recognition result, the sales promotion prompt template, and the customer interest profile information into a pre-trained large language model to obtain the target dialogue output by the large language model.

[0021] Thirdly, embodiments of the present invention also provide a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-mentioned method for generating scripts to respond to a cold response from a customer.

[0022] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed by a processor, causes the processor to perform the above-described method for generating scripts to respond to a cold response from a customer.

[0023] This invention provides a method, apparatus, device, and medium for generating dialogue to address unresponsive customer responses. It relates to the field of artificial intelligence and can be applied to finance and healthcare. When a customer is identified as unresponsive, the method retrieves a target sales strategy from a sales strategy library that matches the customer's state identification result. It then uses the target sales strategy, historical dialogue text, and a set of guiding semantic slots to obtain a sales promotion prompt template and guiding questions. Furthermore, it obtains customer interest profile information through customer feedback corresponding to the guiding questions. The historical dialogue text, customer state identification result, sales promotion prompt template, and customer interest profile information are input into a pre-trained large language model to obtain the target dialogue output by the large language model. This invention optimizes the dialogue content output by the large language model by integrating its semantic understanding capabilities, context awareness capabilities, and sales strategy knowledge, thereby improving the sales success rate. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A flowchart illustrating a method for generating responses to a cold customer response, provided in an embodiment of the present invention.

[0026] Figure 2A schematic block diagram of a script generation device for dealing with unresponsive customers provided in an embodiment of the present invention;

[0027] Figure 3 This is a schematic block diagram of a computer device provided in an embodiment of the present invention;

[0028] Figure 4 This is a schematic diagram illustrating the application environment of the method for generating responses to lukewarm customer feedback provided in this embodiment of the invention. Detailed Implementation

[0029] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0030] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0031] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0032] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. Embodiments of this invention provide a method, apparatus, device, and medium for generating scripts to address lukewarm customer responses. For the method of generating scripts to address lukewarm customer responses, please refer to [link to relevant documentation]. Figure 4 , Figure 4 This is a schematic diagram illustrating the application environment of a script generation method for responding to unresponsive customers, as provided in an embodiment of the present invention. The script generation method for responding to unresponsive customers is applied in, for example... Figure 4In this application environment, the user terminal communicates with the host device via a network. The user terminal can be, but is not limited to, a smartphone, laptop, or tablet, and the host device can be, but is not limited to, a server, desktop computer, or personal computer. The host device executes a method for generating dialogue to address a lukewarm response from a customer. The host device acquires historical dialogue text from interactions with the user terminal and extracts features from the historical dialogue text to obtain historical dialogue features. Based on the historical dialogue features and a pre-trained lukewarm response recognition model, a customer status recognition result is obtained. If the customer status recognition result indicates that the customer is in a lukewarm state, a target sales strategy matching the customer status recognition result is retrieved from a preset sales strategy library, and a sales promotion prompt template is constructed based on the target sales strategy and the historical dialogue text. A guiding question is constructed based on the target sales strategy and a preset set of guiding semantic slots. Customer feedback information corresponding to the guiding question is acquired, and customer interest profile information is obtained based on the customer feedback information. The historical dialogue text, the customer status recognition result, the sales promotion prompt template, and the customer interest profile information are input into a pre-trained large language model to obtain the target dialogue output by the large language model. The method for generating responses to lukewarm customer feedback in this invention can be applied to the financial sector, for example, to promote insurance products; it can also be applied to the medical sector, for example, to promote medical products. The invention will now be described in detail through specific embodiments.

[0033] Please see Figure 1 , Figure 1 This is a flowchart illustrating a method for generating responses to a lukewarm customer reaction, as provided in an embodiment of the present invention. Figure 1 As shown, the method for generating scripts to respond to a customer's lukewarm response includes the following steps S11 to S16.

[0034] S11. Obtain historical dialogue text and extract features from the historical dialogue text to obtain historical dialogue features.

[0035] In this embodiment, the user interacts with the host device through the user terminal. The host device obtains the historical dialogue text of the interaction with the user terminal and extracts features from the historical dialogue text to obtain historical dialogue features, which provide a basis for subsequent identification of customer status.

[0036] S12. Based on the historical dialogue features and the pre-trained indifferent response recognition model, the customer status recognition result is obtained.

[0037] In this embodiment, the historical dialogue features are processed and then input into a pre-trained indifferent response recognition model. The customer status recognition result is obtained based on the output of the indifferent response recognition model. The backbone network of the indifferent response recognition model can be BERT or BiLSTM, and weakly supervised annotation training is performed on the corresponding "warming signals".

[0038] In one embodiment, step S12 includes:

[0039] The features in the historical dialogue features are concatenated to obtain historical dialogue concatenation features; wherein, the historical dialogue features include at least single-turn word count features, sentiment polarity features, utterance ambiguity features, and interaction frequency features;

[0040] The historical dialogue splicing features are input into the indifferent response recognition model to obtain the emotion status code, silence tendency and non-rejection potential score output by the indifferent response recognition model;

[0041] The emotional state code, the tendency to remain silent, and the non-rejection potential are combined to form the customer state identification result.

[0042] In this embodiment, historical dialogue features include, but are not limited to, single-turn word count features, sentiment polarity features, utterance ambiguity features, and interaction frequency features. The single-turn word count feature analyzes whether the customer's word count per turn is too low; the sentiment polarity feature analyzes whether the customer's language expresses positive, neutral, or negative emotions; the utterance ambiguity feature analyzes whether vague words such as "okay" or "just browsing" frequently appear in the customer's language; and the interaction frequency feature analyzes whether the customer remains silent for extended periods or only responds with "um" or "okay." First, the features from the historical dialogue features are concatenated to obtain concatenated historical dialogue features. Then, the concatenated historical dialogue features are input into the indifferent response recognition model for indifferent response judgment. The indifferent response recognition model outputs an emotion status code, a silence tendency, and a non-rejection potential score. The emotion status code, silence tendency, and non-rejection potential score are combined to obtain the customer status recognition result. The emotion status code indicates the customer's emotional tendency, such as neutral, indifferent, or averse emotions; the silence tendency quantifies the likelihood of the customer ceasing interaction; and the non-rejection potential score quantifies the probability that there is still a sales opportunity. The customer status is determined by the customer status recognition results. Specifically, the customer status is jointly assessed by emotion status code, silence tendency and non-rejection potential score. Compared with the traditional keyword-based intent classification method, the embodiments of the present invention assess the customer status from different dimensions and can more accurately identify the customer status.

[0043] In one embodiment, after step S12, the method further includes:

[0044] If the customer status identification result indicates that the customer is in a rejection state, the preset closing remarks and closing voice will be sent to the user terminal.

[0045] In this embodiment, if the customer's status recognition result meets the preset rejection conditions, such as "emotional status code is disgust (i.e., negative emotion), silence tendency is greater than 0.6 and non-rejection potential score is less than 0.4", it can be determined that the customer is in a rejection state. Then, the preset closing dialogue and closing voice are sent to the user terminal, thereby ending the conversation with the customer. The rejection conditions can be set according to actual usage and are not specifically limited here.

[0046] S13. If the customer is determined to be in a lukewarm state based on the customer status identification result, a target sales strategy matching the customer status identification result is obtained from the preset sales strategy library, and a sales promotion prompt template is constructed based on the target sales strategy and the historical dialogue text.

[0047] In this embodiment, if the customer's status recognition result meets a preset indifference condition, such as "emotional status code is cold (i.e., indifferent emotion) and non-rejection potential score is greater than 0.4", it can be determined that the customer is in an indifferent state. Then, a target sales strategy matching the customer status recognition result is obtained from the sales strategy library, and a sales promotion prompt template is constructed based on the target sales strategy and historical dialogue text. The sales strategy library contains several sales strategies, such as: a slow-paced strategy: first express understanding and respect to reduce the customer's defensiveness; a scenario implantation strategy: guide the customer to make associations through life scenarios (e.g., "This flavor is often chosen for elders..."); a question-driven strategy: replace open-ended questions with multiple-choice questions (e.g., "Do you prefer strong aroma or soy sauce aroma?"); a case guidance strategy: cite relevant cases to create a bandwagon effect (e.g., "Other customers have similar thoughts..."); and a focus shift strategy: shift from discussing product and price to non-core issues such as service, packaging, and gifts. Based on the customer status identification results, the corresponding sales strategy in the sales strategy library is dynamically selected as the target sales strategy. The best sales strategy is accurately provided based on the customer's real-time status. By integrating sales psychology strategies, dynamic progress can be achieved, reducing the risk of conversation breakdown and preventing customers from being lost completely due to deteriorating emotions.

[0048] In one embodiment, the step of constructing a sales promotion prompt template based on the target sales strategy and the historical dialogue text includes:

[0049] Obtain the preset strategy template corresponding to the target sales strategy;

[0050] Obtain the current dialogue summary based on the historical dialogue text;

[0051] The sales promotion prompt template is constructed based on the preset strategy template and the current dialogue summary.

[0052] In this embodiment, each sales strategy in the sales strategy library has a corresponding preset strategy template. The preset strategy template corresponding to the target sales strategy is obtained, and the current dialogue summary is extracted from the historical dialogue text. Then, a sales promotion prompt template is constructed based on the preset strategy template and the current dialogue summary. For example, if the target sales strategy is a question-based promotion strategy, the preset strategy template corresponding to the question-based promotion strategy is "Dialogue summary display: [Current dialogue summary]; Please try the following methods to promote the sale: Use the question-based promotion strategy to guide the customer to select [relevant key information determined based on the current dialogue summary] in a multiple-choice format." If the current dialogue summary is "The current customer is showing indifference but has not explicitly refused, and has inquired about the product type," then the constructed sales promotion prompt template is "Dialogue summary display: The current customer is showing indifference but has not explicitly refused, and has inquired about the product type; Please try the following methods to promote the sale: Use the question-based promotion strategy to guide the customer to select the product type in a multiple-choice format."

[0053] S14. Construct guiding questions based on the target sales strategy and the preset guiding semantic slot set.

[0054] In this embodiment, multiple guiding semantic slots are preset from dimensions such as scenario, purpose, target audience, and habits. For example, [purpose slot], [time slot], [flavor slot], etc., can be set, which has hierarchy and scalability. The preset multiple guiding semantic slots are combined into a guiding semantic slot set. In order to address communication barriers when customers do not express themselves clearly, guiding questions are dynamically generated according to the target sales strategy and the preset guiding semantic slot set, and the guiding questions are sent to the user terminal. The guiding questions are used to uncover the customer's hidden intentions. The tone of the guiding questions can be introduced to create a guiding style, so that the customer will be willing to respond with low pressure.

[0055] In one embodiment, step S14 includes:

[0056] Obtain the target guidance semantic slot that matches the target sales strategy from the guidance semantic slot set;

[0057] The guiding question is constructed based on the preset product knowledge graph and the target guiding semantic slot, and then sent to the user terminal.

[0058] In this embodiment, the guidance semantic slot set contains multiple guidance semantic slots, and each sales strategy is matched with a corresponding guidance semantic slot. Therefore, the target guidance semantic slot matching the target sales strategy can be obtained from the guidance semantic slot set. Then, relevant information from the product knowledge graph is used to fill the target guidance semantic slot, thereby constructing the guidance question, which is then sent to the user terminal. For example, based on the product knowledge graph and the [use slot], the guidance question "Are you drinking this yourself or as a gift?" is generated; based on the product knowledge graph and the [time slot], the guidance question "On what occasions do you usually drink alcohol?" is generated; based on the product knowledge graph and the [flavor slot], the guidance question "Do you have any preferences for taste?" is generated, etc. When the customer is in a lukewarm state, the hidden intentions of the customer are further explored through guidance questions.

[0059] S15. Obtain customer feedback information corresponding to the guiding question, and obtain customer interest profile information based on the customer feedback information.

[0060] In this embodiment, customer feedback information sent by the user terminal corresponding to the guiding question is obtained, that is, the customer's feedback information to the guiding question is obtained. Then, customer interest profile information is obtained based on the customer feedback information. The customer interest profile information can express the customer's hidden intentions and needs. The customer interest profile information is used to guide the large language model to generate dialogue that can drive the activity and depth of the conversation.

[0061] In one embodiment, obtaining customer interest profile information based on the customer feedback information includes:

[0062] Obtain keyword information from the customer feedback information;

[0063] The customer interest profile information is obtained by matching the keyword information with the preset customer demand map.

[0064] In this embodiment, a customer demand graph is constructed based on a five-dimensional semantic graph of "scenario-purpose-person-time-value proposition". Keyword information from customer feedback is obtained, and the customer interest profile is established by matching the keyword information with the customer demand graph. For example, if the customer feedback is "Hmm, maybe it's for a friend.", the keyword information is "for a friend", and the customer interest profile established by matching the customer demand graph is ["gift", "mid-range price", "holiday"].

[0065] S16. Input the historical dialogue text, the customer status recognition result, the sales promotion prompt template, and the customer interest profile information into the pre-trained large language model to obtain the target script output by the large language model.

[0066] In this embodiment, historical dialogue text, customer status recognition results, sales promotion prompt templates, and customer interest profile information are jointly input into a pre-trained large language model. This integrates the semantic understanding, context awareness, and sales strategy knowledge of the large language model, enabling the target script output by the model to maintain naturalness while possessing guiding power, logical clarity, and emotional regulation capabilities. By dynamically adjusting the script based on historical dialogue text, customer status recognition results, sales promotion prompt templates, and customer interest profile information, the model accurately matches customer needs and intentions and triggers targeted script output. This avoids repeatedly provoking customers with a single script line, effectively improving the discovery rate and conversion rate of potential customer intentions, and enhancing call effectiveness and sales success rate. Simultaneously, the customer experience is more natural and more readily accepted.

[0067] In one embodiment, after step S16, the method further includes:

[0068] Generate corresponding target speech based on the target speech, and send the target speech and the target speech to the user terminal.

[0069] In this embodiment, the target speech is converted into the corresponding target speech, and the target speech and target speech are sent to the user terminal at the same time. The user terminal then displays and plays the final target speech to the user.

[0070] This invention discloses a method for generating dialogue for dealing with unresponsive customers. When a customer is identified as unresponsive, the method retrieves a target sales strategy from a sales strategy library that matches the customer's state identification result. It then uses the target sales strategy, historical dialogue text, and a set of guiding semantic slots to obtain a sales promotion prompt template and guiding questions. Furthermore, it obtains customer interest profile information through customer feedback corresponding to the guiding questions. The historical dialogue text, customer state identification result, sales promotion prompt template, and customer interest profile information are input into a pre-trained large language model to obtain the target dialogue output by the large language model. This invention optimizes the dialogue content output by integrating the semantic understanding and context awareness capabilities of the large language model with sales strategy knowledge. This results in target dialogue output by the large language model that maintains naturalness while possessing guiding power, logical clarity, and emotional regulation capabilities, thereby improving the sales success rate.

[0071] This invention also provides a script generation device for responding to lukewarm customer responses. This device is used to execute any embodiment of the aforementioned script generation method for responding to lukewarm customer responses. Specifically, please refer to... Figure 2 , Figure 2This is a schematic block diagram of a conversation script generation device for dealing with indifferent customer responses according to an embodiment of the present invention. The device includes a dialogue feature extraction unit 11, a customer status recognition unit 12, a sales promotion prompting unit 13, a guiding question construction unit 14, a customer intent mining unit 15, and a target conversation script generation unit 16. Detailed descriptions of each functional unit are as follows:

[0072] The dialogue feature extraction unit 11 is used to acquire historical dialogue text and extract features from the historical dialogue text to obtain historical dialogue features.

[0073] The customer status identification unit 12 is used to obtain the customer status identification result based on the historical dialogue features and the pre-trained indifferent response identification model;

[0074] The sales promotion prompt unit 13 is used to retrieve a target sales strategy that matches the customer status identification result from a preset sales strategy library if the customer status identification result determines that the customer is in a lukewarm state, and to construct a sales promotion prompt template based on the target sales strategy and the historical dialogue text.

[0075] The guiding question construction unit 14 is used to construct guiding questions based on the target sales strategy and a preset set of guiding semantic slots;

[0076] The customer intent mining unit 15 is used to obtain customer feedback information corresponding to the guiding question, and to obtain customer interest profile information based on the customer feedback information.

[0077] The target dialogue generation unit 16 is used to input the historical dialogue text, the customer status recognition result, the sales promotion prompt template and the customer interest profile information into a pre-trained large language model to obtain the target dialogue output by the large language model.

[0078] In one embodiment, the customer status identification unit 12 is specifically used for:

[0079] The features in the historical dialogue features are concatenated to obtain historical dialogue concatenation features; wherein, the historical dialogue features include at least single-turn word count features, sentiment polarity features, utterance ambiguity features, and interaction frequency features;

[0080] The historical dialogue splicing features are input into the indifferent response recognition model to obtain the emotion status code, silence tendency and non-rejection potential score output by the indifferent response recognition model;

[0081] The emotional state code, the tendency to remain silent, and the non-rejection potential are combined to form the customer state identification result.

[0082] In one embodiment, when the sales promotion prompt unit 13 performs the step of constructing a sales promotion prompt template based on the target sales strategy and the historical dialogue text, it is specifically used for:

[0083] Obtain the preset strategy template corresponding to the target sales strategy;

[0084] Obtain the current dialogue summary based on the historical dialogue text;

[0085] The sales promotion prompt template is constructed based on the preset strategy template and the current dialogue summary.

[0086] In one embodiment, the guiding question construction unit 14 is specifically used for:

[0087] Obtain the target guidance semantic slot that matches the target sales strategy from the guidance semantic slot set;

[0088] The guiding question is constructed based on the preset product knowledge graph and the target guiding semantic slot, and then sent to the user terminal.

[0089] In one embodiment, when performing the step of obtaining customer interest profile information based on the customer feedback information, the customer intent mining unit 15 is specifically used for:

[0090] Obtain keyword information from the customer feedback information;

[0091] The customer interest profile information is obtained by matching the keyword information with the preset customer demand map.

[0092] In one embodiment, the speech generation device for dealing with a cold response from a customer provided by the present invention further includes a target speech generation unit, which is used to generate a corresponding target speech based on the target speech and send the target speech and the target speech to a user terminal.

[0093] In one embodiment, the speech generation device for dealing with a cold response from a customer provided by the present invention further includes a rejection state processing unit, which is used to send a preset ending speech and ending voice to the user terminal if it is determined from the customer state identification result that the customer is in a rejection state.

[0094] This invention discloses a speech generation device for dealing with lukewarm customer responses. This device is used to execute any embodiment of the aforementioned speech generation method for dealing with lukewarm customer responses. When a customer is determined to be in a lukewarm state, it retrieves a target sales strategy matching the customer state identification result from a sales strategy library. It then obtains a sales promotion prompt template and a guiding question using the target sales strategy, historical dialogue text, and a set of guiding semantic slots. Furthermore, it obtains customer interest profile information through customer feedback information corresponding to the guiding question. The historical dialogue text, customer state identification result, sales promotion prompt template, and customer interest profile information are input into a pre-trained large language model to obtain the target speech output by the large language model. This invention optimizes the speech content output by the large language model by integrating its semantic understanding ability, context awareness ability, and sales strategy knowledge. This ensures that the target speech output by the large language model maintains naturalness while possessing guiding power, logical clarity, and emotional regulation capabilities, thereby improving the sales success rate.

[0095] The above-mentioned method for generating scripts to respond to unfriendly customer responses can be implemented as a computer program, which can be used in various ways, such as... Figure 3 It runs on the computer device shown.

[0096] Please see Figure 3 , Figure 3 This is a schematic block diagram of a computer device provided in an embodiment of the present invention. The computer device 500 includes a processor 502, a memory, and a network interface 505 connected via a device bus 501, wherein the memory may include a storage medium 503 and internal memory 504.

[0097] The storage medium 503 may store the operating device 5031 and the computer program 5032. When the computer program 5032 is executed, it causes the processor 502 to execute a method for generating scripts to respond to unfriendly customer responses.

[0098] The processor 502 provides computing and control capabilities to support the operation of the entire computer device 500.

[0099] The internal memory 504 provides an environment for the operation of the computer program 5032 in the storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute a method for generating scripts to deal with a cold response from a customer.

[0100] This network interface 505 is used for network communication, such as providing data transmission. Those skilled in the art will understand that... Figure 3The structure shown is merely a block diagram of a portion of the structure related to the present invention and does not constitute a limitation on the computer device 500 to which the present invention is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0101] The processor 502 is used to run the computer program 5032 stored in the memory to implement the method for generating scripts to deal with indifferent customer responses disclosed in the embodiments of the present invention.

[0102] Those skilled in the art will understand that Figure 3 The embodiments of the computer device shown do not constitute a limitation on the specific configuration of the computer device. In other embodiments, the computer device may include more or fewer components than illustrated, or combine certain components, or have different component arrangements. For example, in some embodiments, the computer device may include only memory and a processor. In such embodiments, the structure and function of the memory and processor are different from those shown. Figure 3 The embodiments shown are consistent and will not be described again here.

[0103] It should be understood that, in this embodiment of the invention, the processor 502 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0104] In another embodiment of the invention, a computer-readable storage medium is provided. This computer-readable storage medium may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores a computer program, wherein when executed by a processor, the computer program implements the script generation method for responding to unresponsive customers disclosed in embodiments of the invention.

[0105] Those skilled in the art will readily understand that, for the sake of convenience and brevity, the specific working processes of the devices and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this invention.

[0106] In the embodiments provided by this invention, it should be understood that the disclosed devices, systems, and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Units with the same function may be grouped into one unit. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, systems, or units, or it may be an electrical, mechanical, or other form of connection.

[0107] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0108] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0109] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of this invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a backend server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), magnetic disks, or optical disks.

[0110] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for generating scripts to respond to indifferent customer responses, characterized in that, include: Obtain historical dialogue text and extract features from the historical dialogue text to obtain historical dialogue features; The customer status identification result is obtained based on the historical dialogue features and the pre-trained indifferent response identification model; If the customer status identification result indicates that the customer is in a lukewarm state, then the target sales strategy matching the customer status identification result is obtained from the preset sales strategy library, and a sales promotion prompt template is constructed based on the target sales strategy and the historical dialogue text. Based on the target sales strategy and the preset set of guiding semantic slots, guiding questions are constructed. Obtain customer feedback information corresponding to the guiding question, and obtain customer interest profile information based on the customer feedback information; The historical dialogue text, the customer status recognition result, the sales promotion prompt template, and the customer interest profile information are input into a pre-trained large language model to obtain the target script output by the large language model.

2. The method for generating scripts to respond to indifferent customer responses according to claim 1, characterized in that, The customer status identification result obtained by the customer-based historical dialogue features and the pre-trained indifferent response identification model includes: The features in the historical dialogue features are concatenated to obtain historical dialogue concatenation features; wherein, the historical dialogue features include at least single-turn word count features, sentiment polarity features, utterance ambiguity features, and interaction frequency features; The historical dialogue splicing features are input into the indifferent response recognition model to obtain the emotion status code, silence tendency and non-rejection potential score output by the indifferent response recognition model; The emotional state code, the tendency to remain silent, and the non-rejection potential are combined to form the customer state identification result.

3. The method for generating scripts to respond to indifferent customer responses according to claim 1, characterized in that, The sales promotion prompt template constructed based on the target sales strategy and the historical dialogue text includes: Obtain the preset strategy template corresponding to the target sales strategy; Obtain the current dialogue summary based on the historical dialogue text; The sales promotion prompt template is constructed based on the preset strategy template and the current dialogue summary.

4. The method for generating scripts to respond to indifferent customer responses according to claim 1, characterized in that, The step of constructing guiding questions based on the target sales strategy and a preset set of guiding semantic slots includes: Obtain the target guidance semantic slot that matches the target sales strategy from the guidance semantic slot set; The guiding question is constructed based on the preset product knowledge graph and the target guiding semantic slot, and then sent to the user terminal.

5. The method for generating scripts to respond to indifferent customer responses according to claim 1, characterized in that, The step of obtaining customer interest profile information based on the customer feedback information includes: Obtain keyword information from the customer feedback information; The customer interest profile information is obtained by matching the keyword information with the preset customer demand map.

6. The method for generating scripts to respond to indifferent customer responses according to claim 1, characterized in that, After the step of inputting the historical dialogue text, the customer status recognition result, the sales promotion prompt template, and the customer interest profile information into the pre-trained large language model to obtain the target speech output by the large language model, the method further includes: Generate corresponding target speech based on the target speech, and send the target speech and the target speech to the user terminal.

7. The method for generating scripts to respond to indifferent customer responses according to claim 1, characterized in that, After the step of obtaining the customer status identification result based on the historical dialogue features and the pre-trained indifferent response identification model, the method further includes: If the customer status identification result indicates that the customer is in a rejection state, the preset closing remarks and closing voice will be sent to the user terminal.

8. A script generation device for responding to indifferent customer responses, characterized in that, The apparatus is used to perform the script generation method for responding to a lukewarm customer response as described in any one of claims 1-7, the apparatus comprising: The dialogue feature extraction unit is used to acquire historical dialogue text and extract features from the historical dialogue text to obtain historical dialogue features. A customer status identification unit is used to obtain customer status identification results based on the historical dialogue features and a pre-trained indifferent response identification model; The sales promotion prompt unit is used to retrieve a target sales strategy that matches the customer status identification result from a preset sales strategy library if the customer status identification result determines that the customer is in a lukewarm state, and to construct a sales promotion prompt template based on the target sales strategy and the historical dialogue text. The guiding question construction unit is used to construct guiding questions based on the target sales strategy and a preset set of guiding semantic slots; The customer intent mining unit is used to obtain customer feedback information corresponding to the guiding question, and to obtain customer interest profile information based on the customer feedback information. The target dialogue generation unit is used to input the historical dialogue text, the customer status recognition result, the sales promotion prompt template, and the customer interest profile information into a pre-trained large language model to obtain the target dialogue output by the large language model.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the script generation method for responding to a cold response from a customer as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, causes the processor to perform the script generation method for responding to a lukewarm customer response as described in any one of claims 1 to 7.

Images