A customer sentiment steering model

Abstract

The present application relates to the technical field of data processing, in particular to a customer emotion guiding model, comprising: an emotion acquisition module, used for acquiring browsing information of a customer before communication, converting the browsing information into text information, and calculating an emotion value of the customer before communication through a preset emotion classification model; an information acquisition module, used for acquiring text information and sound information of the customer during communication; a sorting module, used for acquiring to-be-pushed information data to be transmitted to the customer, calculating a score value of each piece of information data through the emotion acquisition module, and performing descending arrangement on the score values; and a customer terminal, used for acquiring the sorted information data; if a typing speed / speaking speed is different from a usual speed by a threshold, an abnormal signal is generated, a proportion of the difference is taken as a variable, and a current emotion value of the user is updated; the customer emotion can be improved by sorting to-be-reported messages when old customers in the intellectual property industry are served.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a customer emotion guidance model. Background Art

[0002] With the rise and development of artificial intelligence, numerous big data models have emerged. These models, trained using unique language training models, can answer questions tailored to the needs of their customers. Consequently, many of these models have found applications in customer service, such as in e-commerce platforms' customer bots and chatbots in group chats. Current customer service bots can be broadly divided into two categories. The first focuses on answering customer inquiries. These systems analyze customer questions and, using various natural language processing (NLP) methods to calculate semantic similarity, identify the most similar questions in a FAQ (frequently asked questions and their corresponding answers) knowledge base and recommend corresponding answers. The other category integrates features such as product recommendations and customer service. These systems leverage machine learning and deep learning technologies to build end-to-end neural network models. These systems use the context of the conversation between the customer and the bot, as well as historical customer behavior data, to answer customer questions and make product recommendations.

[0003] The invention patent with patent number 202011459385.X discloses a customer service robot dialogue method based on reinforcement learning and its related components. The method includes: obtaining the first text information input by the user in the previous round and selecting the corresponding action strategy for reply; obtaining the second text information of the user's feedback on the action strategy, using a pre-trained emotion classification model to perform emotion classification on the second text information, obtaining the user's emotion, and calculating the total response value score according to the maximum expectation algorithm; converting the text information input by the user into a text vector and inputting it into a multi-layer neural network for reinforcement learning to obtain the probability of selecting the corresponding action strategy, optimizing the multi-layer neural network according to the total response value score and the probability of the action strategy to obtain the customer service robot response model. The customer service robot response model of the present invention based on reinforcement learning technology guides the customer service robot to learn to continuously adjust its response actions, thereby improving customer satisfaction during the conversation process and reducing user churn.

[0004] The above-mentioned technology uses reinforcement learning to continuously improve the accuracy of customer service robots' responses by performing reinforcement learning on customer service models. However, it only classifies the emotions of customers' text messages, which is not accurate enough to accurately grasp customer emotions. Moreover, customer emotions are fluid. The above-mentioned technology only responds to messages based on the customer's emotion value, that is, responds to messages corresponding to the customer's emotions based on the customer's emotional changes. It does not guide customer emotions. For example, when a customer is in a bad mood, replying with positive messages can guide their emotions in a positive direction and improve their mood. Applying the above-mentioned technology in the intellectual property industry can maintain service for regular customers and enhance their sensory experience. However, there is still a lack of targeted guidance of customer emotions. This is especially true when reporting to customers on news, such as the status of their intellectual property authorizations and other intellectual property business matters. There is a need for a technology that can sort the messages to be reported to regular customers in the intellectual property industry and improve their emotions. Summary of the Invention

[0005] The present invention provides a customer emotion guidance model, which can sort the messages to be reported when serving old customers in the intellectual property industry and improve the customer's emotion.

[0006] In order to solve the above technical problems, this application provides the following technical solutions:

[0007] A customer emotion guidance model, including:

[0008] Emotion acquisition module: used to obtain the browsing information of customers before communication, convert it into text information, and calculate the customer's emotion value before communication through a preset emotion classification model;

[0009] Information acquisition module: used to obtain text and voice information from customers during the communication process;

[0010] Sorting module: used to obtain the information data to be pushed to the customer, and calculate the score value of each piece of information data through the sentiment acquisition module, and then sort them in descending order by the score value;

[0011] Client terminal: used to obtain sorted information data;

[0012] Among them, the information acquisition module also acquires the customer's usual typing speed / speaking speed, and compares the user's typing speed / speaking speed with the usual speed during the communication process. If the typing speed / speaking speed differs from the usual speed by reaching a threshold, an abnormal signal is generated, and the user's current emotion value is updated based on the ratio of the difference as a variable; the emotion acquisition module is also used to predict the user's predicted emotion value after receiving a push information data based on the scoring data of the information data, and compare the predicted emotion value after receiving the information data with the current emotion value after receiving the information data. If the difference is within the threshold range, it is determined that the emotion acquisition module is reasonable for the emotion prediction, otherwise the difference is directed to increase or decrease the difference in the subsequent predicted values ​​of the emotion acquisition module; the sorting module sets the customer's minimum emotion value, and when sorting the data information to be pushed, it first obtains the predicted emotion value of the user when receiving the data information, and the predicted emotion value must be greater than the minimum emotion value.

[0013] Basic principles and beneficial effects: The emotion acquisition module is used to preliminarily acquire the customer's emotion value before the communication, that is, through the customer's browsing history before the communication, such as web browsing history, app browsing history, etc. Since the information recommendation of some websites nowadays depends on the customer's personal preferences, that is, what the customer likes is recommended, the user's emotional state before the communication can be acquired through the customer's browsing history, and then the emotion value is obtained through the preset emotion classification model; during the communication process, the customer's communication behavior is captured, that is, for online communication, the customer's typing speed or speaking speed is collected, and for offline communication, the customer's speaking speed is collected, that is, the customer's content output speed is collected. If the speed during the communication is different from the customer's usual typing speed, the customer's emotional state can be acquired through the customer's browsing history. If the speed / speaking speed is not much different (within the threshold), the emotional value will not be changed. If the difference exceeds the threshold, the customer's current emotional value will be updated; the sorting module is used to sort the information data to be pushed to the customer. First, the emotional value of each information data to be pushed is scored through the emotion acquisition module, that is, the scoring value is obtained, and then the predicted emotional value of the customer after obtaining the pushed information data is predicted. If the predicted emotion is higher than the minimum emotional value, it can be pushed to the customer. If the predicted emotion is lower than the minimum emotional value, different information data are adapted to encourage the customer's emotional value to be higher than the minimum emotional value, that is, good news is pushed when the customer is in a low mood, so as to guide the customer's emotions in a positive direction.

[0014] This solution sorts the information data to be pushed according to the customer's tolerance (minimum emotional value), and completes the targeted guidance of customer emotions on the premise of completing the push of information data to customers.

[0015] This solution is different from the existing technology in that it pushes information data in a targeted manner according to the customer's preferences. Instead, it sorts the order of pushed information based on the known push information, ensuring the integrity of the information data push. This avoids the problem of information cocoons that exists in the existing technology, and pushes both good news and bad news to customers.

[0016] Furthermore, it also includes an emotion compensation module, which is used to extract positive content from the favorites list in the user's browsing information when the customer's predicted emotion value is lower than the minimum customer emotion value. The areas of the positive content at least include success studies, inspirational stories, and motivational stories; then search for similar information based on the positive content in the favorites list and send it to the customer terminal.

[0017] Furthermore, it also includes an incentive module, which is used to collect policy information of the customer's industry, extract policy information that meets the customer's application conditions, and display the policy information that meets the customer's application conditions before pushing information data to the customer.

[0018] Furthermore, the policy information includes policy rewards and application conditions, and the sorting module is further configured to place the policy rewards in the policy information before the application conditions.

[0019] Furthermore, when the ranking module ranks the rating values ​​of the information data to be pushed, the information data with the highest rating value is separately removed, and after the ranking of other information data to be pushed is completed, the information data with the highest rating value is ranked last and pushed.

[0020] Furthermore, the voice information includes the loudness of the customer's voice. During the communication process, the information acquisition module acquires the voice loudness of the user within the interval. If the change in the loudness of a sentence exceeds a threshold, the keyword in the content of the sentence is marked, and the threshold is changed for the predicted value of information data related to the keyword that appears in subsequent push.

[0021] Furthermore, the information acquisition module is also used to obtain the customer's facial features during communication; the emotion acquisition module also has a built-in facial emotion recognition neural network module, which is used to identify the customer's current emotion type based on the customer's facial features, and to preset scores in ascending order based on the facial features from negative emotions to positive emotions. When the customer's facial features are identified as a certain emotion type, the current preset score is extracted, and a new current emotion value is generated by weighted calculation of the preset score and the current emotion value. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a logic diagram of the first embodiment of a customer emotion guidance model. DETAILED DESCRIPTION

[0023] The following is further described in detail through specific implementation methods:

[0024] Example 1 is as shown in the attached Figure 1 As shown,

[0025] A customer emotion guidance model, including:

[0026] Emotion acquisition module: used to obtain the browsing information of customers before communication, convert it into text information, and then calculate the customer's emotion value before communication through a preset emotion classification model; the emotion classification model uses the pre-trained emotion classification model in the existing technology to determine the customer's current emotion value.

[0027] The information acquisition module is used to acquire text and audio information from customers during the communication process. The information acquisition module is also used to acquire the customer's facial features during the communication process. The emotion acquisition module also includes a built-in facial emotion recognition neural network module, which identifies the customer's current emotion type based on their facial features and assigns ascending preset scores to facial features from negative to positive emotions. When the customer's facial features are identified as belonging to a certain emotion type, the current preset score is extracted and weighted by the preset score and the current emotion value to generate a new current emotion value. The information acquisition module is built into the customer terminal, which is used to acquire sorted information data. For example, in online communication (or offline integrated self-service kiosk), the customer service terminal is the mobile phone used by the customer to communicate. The mobile phone's camera, microphone, and other functions are used to acquire real-time facial features and audio information from the customer. The audio information includes the customer's voice loudness. During the communication process, the information acquisition module acquires the voice loudness within the user interval. If the voice loudness change in a sentence exceeds a threshold, the sentence is marked as a keyword, and the threshold is adjusted for the predicted value of information data related to the keyword in subsequent push notifications. That is, during the communication process, the customer may suddenly become emotionally excited, and the content of this sentence may be the point that excited the customer. By extracting and marking the keywords in this sentence, and subsequently pushing the keyword-related content, the predicted emotion value after the message data is pushed is corrected. If the emotion value when the user obtains the keyword increases, the predicted emotion value is quantitatively increased, otherwise it is reduced. That is, the increase / decrease threshold is 5% of the lowest emotion value.

[0028] Sorting module: used to obtain the information data to be pushed to the customer, and calculate the score value of each information data through the emotion acquisition module, and then arrange them in descending order according to the value; when the sorting module arranges the score values ​​of the information data to be pushed, the information data with the highest score value is separately removed, and after the other information data to be pushed are sorted, the information data with the highest score value is arranged to be pushed last.

[0029] Among them, the information acquisition module also acquires the customer's usual typing speed / speaking speed, and compares the user's typing speed / speaking speed with the usual speed during the communication process. If the typing speed / speaking speed differs from the usual speed by reaching a threshold, an abnormal signal is generated, and the user's current emotion value is updated based on the ratio of the difference as a variable; the emotion acquisition module is also used to predict the user's predicted emotion value after receiving a push information data based on the scoring data of the information data, and compare the predicted emotion value after receiving the information data with the current emotion value after receiving the information data. If the difference is within the threshold range, it is determined that the emotion acquisition module is reasonable for the emotion prediction, otherwise the difference is directed to increase or decrease the difference in the subsequent predicted values ​​of the emotion acquisition module; the sorting module sets the customer's minimum emotion value, and when sorting the data information to be pushed, it first obtains the predicted emotion value of the user when receiving the data information, and the predicted emotion value must be greater than the minimum emotion value.

[0030] It also includes an emotion compensation module, which is used to extract positive content from the favorites list in the user's browsing information when the customer's predicted emotion value is lower than the minimum customer emotion value. The areas of the positive content at least include success studies, inspirational stories, and motivational stories; then search for similar information based on the positive content in the favorites list and send it to the customer terminal.

[0031] The system also includes an incentive module, which collects policy information related to the client's industry and extracts policy information that meets the client's application requirements. This information is then displayed before the information data is pushed to the client. The policy information includes policy rewards and application requirements, and the ranking module is further configured to prioritize the policy rewards before the application requirements. This provides good news before the message is pushed to boost the client's initial sentiment, facilitating their subsequent acceptance of the information data to be pushed.

[0032] Example 2

[0033] The client terminal is provided with a brightness adjustment module, which is used to adjust the screen brightness of the client terminal. When the emotion acquisition module detects that the user's current emotion value is lower than the minimum emotion value, the camera acquires the user's facial information, the data processing module recognizes the customer's face, extracts the user's eye image, and determines the user's line of sight by the direction of the eyeballs. If the line of sight deviates from the client terminal, a brightness reduction signal is sent, and the brightness adjustment module adjusts the screen brightness to the minimum brightness. When the user's line of sight moves back to the screen, the screen brightness is adjusted to the first mode. The screen brightness in the first mode is the screen brightness in the power saving mode. When the user does not perform the next operation, the screen brightness remains in the brightness in the first mode. When the camera captures the user's hand approaching the screen, the screen brightness is increased to normal brightness.

[0034] This solution uses a brightness adjustment module in conjunction with an emotion acquisition module to adjust screen brightness. When a user is feeling depressed (the current emotion level is below the minimum), they need to calm themselves down and block out external information, avoiding new information. Consequently, their gaze wanders, losing focus on the content they're currently browsing and looking elsewhere. When this gaze wanders, coupled with the user's low mood, it determines that they need to adjust themselves and lowers the screen brightness. This saves power consumption and reduces the risk of secondary emotional impact. Self-adjustment typically involves self-persuasion and selective forgetting. This means that after a period of calming down, the user looks at the screen again. If they haven't calmed down yet and still see a message that calms them down, their emotion level is lowered again. By lowering the screen brightness, the user isn't able to directly and quickly see the message that calms them down, preventing a sudden emotional shock and facilitating emotional self-calming and adjustment. Once the user has calmed down and has sufficient mental preparation, they will resume operating the terminal to acquire information. When the user's intention to operate is detected, the screen brightness is adjusted back to normal.

[0035] The above are only embodiments of the present invention. The invention is not limited to the fields involved in this implementation case. Common knowledge such as the known specific structures and characteristics in the scheme is not described in detail here. Ordinary technicians in the relevant field are aware of all common technical knowledge in the technical field to which the invention belongs before the application date or priority date, can obtain all existing technologies in the field, and have the ability to apply conventional experimental means before that date. Ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the inspiration given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several variations and improvements can be made, which should also be regarded as the scope of protection of the present invention. These will not affect the effect of the implementation of the present invention and the practicality of the patent. The scope of protection required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. A customer emotion guidance model, characterized in that: include: Emotion acquisition module: used to obtain the browsing information of customers before communication, convert it into text information, and calculate the customer's emotion value before communication through a preset emotion classification model; Information acquisition module: used to obtain text and voice information from customers during the communication process; Sorting module: used to obtain the information data to be pushed to the customer, and calculate the score value of each piece of information data through the sentiment acquisition module, and then sort them in descending order by the score value; Client terminal: used to obtain sorted information data; Among them, the information acquisition module also acquires the customer's usual typing speed / speaking speed, and compares the user's typing speed / speaking speed with the usual speed during the communication process. If the typing speed / speaking speed differs from the usual speed by reaching a threshold, an abnormal signal is generated, and the user's current emotion value is updated based on the ratio of the difference as a variable; the emotion acquisition module is also used to predict the user's predicted emotion value after receiving a push information data based on the scoring data of the information data, and compare the predicted emotion value after receiving the information data with the current emotion value after receiving the information data. If the difference is within the threshold range, it is determined that the emotion acquisition module is reasonable for the emotion prediction, otherwise the difference is directed to increase or decrease the difference in the subsequent predicted values ​​of the emotion acquisition module; the sorting module sets the customer's minimum emotion value, and when sorting the data information to be pushed, it first obtains the predicted emotion value of the user when receiving the data information, and the predicted emotion value must be greater than the minimum emotion value.

2. A customer emotion guidance model according to claim 1, characterized in that: It also includes an emotion compensation module, which is used to extract positive content from the favorites list in the user's browsing information when the customer's predicted emotion value is lower than the minimum customer emotion value. The areas of the positive content at least include success studies, inspirational stories, and motivational stories; then search for similar information based on the positive content in the favorites list and send it to the customer terminal.

3. A customer emotion guidance model according to claim 1, characterized in that: It also includes an incentive module, which is used to collect policy information of the customer's industry, extract policy information that meets the customer's application conditions, and display the policy information that meets the customer's application conditions before pushing information data to the customer.

4. A customer emotion guidance model according to claim 3, characterized in that: The policy information includes policy rewards and application conditions, and the sorting module is further used to place the policy rewards in the policy information before the application conditions.

5. The customer emotion guidance model according to claim 1, characterized in that: When the ranking module ranks the score values ​​of the information data to be pushed, the information data with the highest score value is separately removed, and after the ranking of other information data to be pushed is completed, the information data with the highest score value is ranked and pushed last.

6. The customer emotion guidance model according to claim 1, characterized in that: The voice information includes the loudness of the customer's voice. During the communication process, the information acquisition module acquires the voice loudness of the user within the interval. If the change in the loudness of a sentence exceeds a threshold, the keyword in the content of the sentence is marked, and the threshold is changed for the predicted value of information data related to the keyword in subsequent push notifications.

7. The customer emotion guidance model according to claim 1, characterized in that: The information acquisition module is also used to obtain the customer's facial features during communication; the emotion acquisition module also has a built-in facial emotion recognition neural network module, which is used to identify the customer's current emotion type based on the customer's facial features, and to preset scores in ascending order based on the facial features from negative emotions to positive emotions. When the customer's facial features are identified as a certain emotion type, the current preset score is extracted, and a new current emotion value is generated by weighted calculation of the preset score and the current emotion value.

Images