Intelligent call-out processing method and device based on artificial intelligence, and server

Abstract

The invention relates to the technical field of intelligent outbound calls, in particular to an intelligent outbound call processing method and device based on artificial intelligence and a server. Real-time voice data input by a user is converted into text data based on a voice analysis model, intention analysis is conducted on the text data based on an intention analysis model, and the intention of the user is obtained. And performing sentiment analysis on the text data based on the sentiment analysis model to obtain a sentiment analysis result of the user, generating a dialogue strategy matched with the user according to a user intention analysis result and the sentiment analysis result, and finally executing dynamic adjustment operation according to the dialogue strategy and dialogue skill data to obtain verbal skill data matched with the user. According to the method, the accurate dialogue strategy is generated for the user not only through the intention of the user but also in combination with the sentiment analysis result of the user, so that the verbal skill data to be provided for the user is adjusted, the determination accuracy and precision of the verbal skill data needed by the user are improved, and the interaction accuracy of the intelligent outbound dialogue process is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of intelligent outbound call, and in particular to an intelligent outbound call processing method and device based on artificial intelligence and a server. BACKGROUND

[0002] With the rapid progress of society, under the impetus of digital transformation, efficient communication between enterprises and users has become a key factor in improving competitiveness. The traditional manual outbound call mode has been difficult to meet the needs of enterprises for efficient communication due to high labor costs, low efficiency, poor user experience and other drawbacks.

[0003] With the rapid development of artificial intelligence technologies such as natural language processing and machine learning, intelligent outbound call management systems have emerged, which have improved the communication efficiency between enterprises and users to some extent. Currently, the specific operation of the intelligent outbound call management system is: when detecting the voice input by the user, the voice is analyzed, and then the analysis result is compared with the fixed set template to determine the matched dialogue to answer the user.

[0004] However, it is found in practice that although the existing intelligent outbound call technology introduces natural language processing technology, there are still deficiencies in user intent recognition, such as low recognition accuracy for complex semantics and dialects, and difficulty in accurately understanding the diversified needs of users, resulting in low accuracy of the matched dialogue, and thus low accuracy of the outbound call. Therefore, it is urgent to propose a new intelligent outbound call processing method to improve the determination accuracy of the corresponding dialogue of the user, thereby improving the accuracy and efficiency of the intelligent outbound call. SUMMARY

[0005] The present application provides an intelligent outbound call processing method and device based on artificial intelligence and a server, which can improve the determination accuracy and precision of the corresponding dialogue of the user, thereby improving the accuracy and efficiency of the intelligent outbound call.

[0006] To solve the above technical problems, the first aspect of the present application discloses an intelligent outbound call processing method based on artificial intelligence, which is applied to a plurality of target users. For any target user, the method comprises: When it is detected that the current condition meets the condition determined in advance that an outbound call request needs to be triggered for the target user, an outbound call request for the target user is sent to the target user, and real-time voice data input by the target user is collected according to the outbound call request; The real-time voice data of the target user is converted into text data based on a pre-trained voice analysis model, and the text data corresponding to the target user is analyzed based on a pre-trained intent analysis model to obtain an intent analysis result of the target user, which includes a target intent. performing sentiment analysis on the text data corresponding to the target user based on the pre-trained sentiment analysis model to obtain a sentiment analysis result of the target user, and generating a dialogue strategy matched with the target user according to the intent analysis result and the sentiment analysis result of the target user; performing dynamic adjustment operation on the obtained dialogue data corresponding to the target user according to the dialogue strategy corresponding to the target user to obtain target dialogue data matched with the target user.

[0007] As an optional implementation, in the first aspect of the present application, the conversion of the real-time voice data of the target user into text data based on the pre-trained voice analysis model comprises: performing voice feature recognition on the real-time voice data of the target user based on the voice feature layer of the pre-trained voice analysis model to obtain voice features of the voice data, performing long-distance voice dependency relationship capture on the voice features based on the multi-layer encoder of the voice analysis model to obtain text data of the voice data, and performing text decoding and correction on the text data of the voice data based on the decoder of the voice analysis model to obtain the text data corresponding to the target user; wherein the intent analysis on the text data corresponding to the target user based on the pre-trained intent analysis model to obtain the intent analysis result of the target user comprises: extracting context features of the text data corresponding to the target user based on the multi-layer feature extractor of the pre-trained intent analysis model, and analyzing the context features based on the classification layer of the intent analysis model to obtain the intent analysis result of the target user; wherein the sentiment analysis on the text data corresponding to the target user based on the pre-trained sentiment analysis model to obtain the sentiment analysis result of the target user comprises: analyzing all emotion words in the text data corresponding to the target user for expressing emotion based on the sequence processing layer of the pre-trained sentiment analysis model, and analyzing the all emotion words based on the sequence processing layer of the sentiment analysis model to obtain the sentiment analysis result of the target user.

[0008] As an optional implementation, in the first aspect of the present application, the dynamic adjustment operation on the obtained dialogue data corresponding to the target user according to the dialogue strategy corresponding to the target user to obtain target dialogue data matched with the target user comprises: When the target user corresponds to a dialogue strategy indicating that a voice dialogue needs to be carried out with the target user through voice, a feature extraction layer of a predetermined voice synthesis model is used to perform text feature extraction on the dialogue data corresponding to the target user, to obtain text features of the dialogue data, and a feature conversion layer of the voice synthesis model is used to convert the text features of the dialogue data into voice features; Based on the emotional analysis result of the target user and the obtained context dialogue data corresponding to the target user, an emotional demand parameter required by the target user is analyzed, and a voice synthesis layer of the voice synthesis model is used to perform voice conversion analysis on the voice features of the dialogue data and the emotional demand parameter required by the target user, to obtain target dialogue data matched with the target user.

[0009] As an optional implementation, in the first aspect of the present application, for any target user, the method further comprises: obtaining user information of the target user and historical transaction data in a past target time period, and determining potential value of the target user according to the user information and the historical transaction data of the target user; obtaining historical interaction data of the target user in the target time period, and analyzing historical response rate of the target user according to the historical interaction data; obtaining business data of the target user in the target time period, and analyzing business urgency matched with the target user according to the business data corresponding to the target user; According to the obtained current business demand of the target user, the potential value, the historical response rate and the business urgency of the target user are assigned corresponding priority weights, and the business priority level of the target user is analyzed according to the potential value, the historical response rate, the business urgency and the corresponding priority weights of the target user; According to the business priority level of the target user, an outbound call task parameter matched with the target user is generated, wherein the outbound call task parameter of the target user includes an outbound call time period, an outbound call frequency and a dialogue template of the target user; The dialogue data corresponding to the target user includes the dialogue template of the target user. If the real-time time arrives at the outbound call time period of the target user, or if the current outbound call number of the target user is less than the outbound call frequency of the target user, it is detected that the current condition meets the condition of triggering the outbound call request for the target user.

[0010] The second aspect of the embodiment of the present application discloses an intelligent outbound call processing device based on artificial intelligence, which is applied to a plurality of target users, and the device comprises: The communication module is configured to send an outbound call request for the target user to the target user when it is detected that a current condition meets a predetermined condition requiring triggering the outbound call request for the target user, and collect real-time voice data input by the target user according to the outbound call request; The analysis module is configured to convert the real-time voice data of the target user into text data based on a pre-trained voice analysis model, and perform intent analysis on the text data corresponding to the target user based on a pre-trained intent analysis model to obtain an intent analysis result of the target user, wherein the intent analysis result of the target user includes a target intent. The analysis module is further configured to perform sentiment analysis on the text data corresponding to the target user based on a pre-trained sentiment analysis model to obtain a sentiment analysis result of the target user. The generation module is configured to generate a dialogue strategy matched with the target user according to the intent analysis result and the sentiment analysis result of the target user. The adjustment module is configured to perform a dynamic adjustment operation on the obtained dialogue data corresponding to the target user according to the dialogue strategy corresponding to the target user to obtain target dialogue data matched with the target user.

[0011] As an optional implementation, in the second aspect of the present application, the specific manner in which the analysis module converts the real-time voice data of the target user into text data based on the pre-trained voice analysis model includes: The voice analysis model includes a voice feature layer, a multi-layer encoder, and a decoder. The voice feature layer is configured to perform voice feature recognition on the real-time voice data of the target user to obtain voice features of the voice data, the multi-layer encoder is configured to capture long-distance voice dependency relationships of the voice features to obtain text data of the voice data, and the decoder is configured to perform text decoding and correction on the text data of the voice data to obtain the text data corresponding to the target user. The analysis module is configured to perform intent analysis on the text data corresponding to the target user based on a pre-trained intent analysis model to obtain an intent analysis result of the target user. The specific manner in which the analysis module performs intent analysis on the text data corresponding to the target user based on the pre-trained intent analysis model includes: The intent analysis model includes a multi-layer feature extractor and a classification layer. The multi-layer feature extractor is configured to extract context features of the text data corresponding to the target user, and the classification layer is configured to analyze the context features to obtain the intent analysis result of the target user. The analysis module is configured to perform sentiment analysis on the text data corresponding to the target user based on a pre-trained sentiment analysis model to obtain a sentiment analysis result of the target user. The specific manner in which the analysis module performs sentiment analysis on the text data corresponding to the target user based on the pre-trained sentiment analysis model includes: The sequence processing layer based on the pre-trained sentiment analysis model analyzes all sentiment words in the text data corresponding to the target user for expressing sentiment, and obtains a sentiment analysis result of the target user based on the sequence processing layer of the sentiment analysis model.

[0012] As an optional implementation, in the second aspect of the application, the adjustment module performs a dynamic adjustment operation on the obtained dialogue data corresponding to the target user according to the dialogue strategy corresponding to the target user, and the specific manner of obtaining target dialogue data matched with the target user comprises: When the dialogue strategy corresponding to the target user indicates that voice dialogue needs to be performed with the target user through voice, a text feature extraction is performed on the dialogue data corresponding to the target user based on a feature extraction layer of a pre-determined voice synthesis model, to obtain a text feature of the dialogue data, and a text feature of the dialogue data is converted into a voice feature based on a feature conversion layer of the voice synthesis model; Based on the sentiment analysis result of the target user and the obtained context dialogue data corresponding to the target user, a sentiment demand parameter required by the target user is analyzed, and a voice conversion analysis is performed on the voice feature corresponding to the dialogue data and the sentiment demand parameter required by the target user based on a voice synthesis layer of the voice synthesis model, to obtain target dialogue data matched with the target user.

[0013] The third aspect of the application discloses an intelligent outbound service side, the service side comprises: a memory storing executable program codes; a processor coupled with the memory; The processor calls the executable program codes stored in the memory to execute part or all steps of any one of the intelligent outbound processing methods based on artificial intelligence disclosed in the first aspect of the application.

[0014] The fourth aspect of the application discloses a computer storage medium storing computer instructions, which are called to execute part or all steps of any one of the intelligent outbound processing methods based on artificial intelligence disclosed in the first aspect of the application.

[0015] Compared with the prior art, the embodiments of the application have the following beneficial effects: In the embodiment of the present application, the user is sent an outbound call request, real-time voice data input by the user is collected according to the outbound call request, the real-time voice data of the user is converted into text data based on a voice analysis model, the converted text data is analyzed for intent based on an intent analysis model to obtain the intent of the user, the converted text data is analyzed for sentiment based on a sentiment analysis model to obtain the sentiment analysis result of the user, a dialog strategy matched with the user is generated according to the analyzed intent analysis result and sentiment analysis result of the user, and finally the dialog strategy corresponding to the user is used to perform a dynamic adjustment operation on the obtained dialog data corresponding to the user to obtain dialog data matched with the user. That is, not only the intent of the user, but also the sentiment analysis result of the user, is used to generate an accurate dialog strategy for the user to perform an adjustment operation on the dialog data to be provided to the user, thereby improving the determination accuracy and precision of the dialog data required by the user, and improving the interaction accuracy, efficiency and personalization of the intelligent outbound call dialog process, thereby enhancing the participation and satisfaction of the user, and further improving the accuracy, smoothness and efficiency of the intelligent outbound call, while reducing the labor cost. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0017] Figure 1 is a flow diagram of an intelligent outbound call processing method based on artificial intelligence disclosed by the embodiment of the present application; Figure 2 is a flow diagram of another intelligent outbound call processing method based on artificial intelligence disclosed by the embodiment of the present application; Figure 3 is a structural diagram of an intelligent outbound call processing device based on artificial intelligence disclosed by the embodiment of the present application; Figure 4 is a structural diagram of another intelligent outbound call processing device based on artificial intelligence disclosed by the embodiment of the present application; Figure 5 is a structural diagram of an intelligent outbound call service end disclosed by the embodiment of the present application. DETAILED DESCRIPTION

[0018] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0019] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, apparatus, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0020] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of the invention. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0021] This invention discloses an intelligent outbound call processing method, device, and server based on artificial intelligence. It sends an outbound call request to a user and collects real-time voice data input by the user based on the request. The real-time voice data is converted into text data based on a voice analysis model. Intent analysis is then performed on the converted text data based on an intent analysis model to obtain the user's intent. Finally, sentiment analysis is performed on the converted text data based on a sentiment analysis model to obtain the user's sentiment analysis results. Based on the analyzed user intent and sentiment analysis results, a dialogue strategy matching the user is generated. Finally, dynamic adjustments are performed on the acquired user-specific script data according to the corresponding dialogue strategy to obtain script data matching the user. In other words, it not only uses the user's intent but also combines the user's sentiment analysis results to generate an accurate dialogue strategy for the user, adjusting the script data prepared for the user. This improves the accuracy and precision of determining the required script data, thereby improving the interactive accuracy, efficiency, and personalization of the intelligent outbound call dialogue process, enhancing user participation and satisfaction, and ultimately improving the accuracy, smoothness, and efficiency of intelligent outbound calls while reducing labor costs. Detailed explanations follow.

[0022] Example 1 Please see Figure 1 , Figure 1 This is a flowchart illustrating an artificial intelligence-based intelligent outbound call processing method disclosed in an embodiment of the present invention. This method can be applied to any scenario requiring intelligent outbound call processing, such as item rental, item purchase, or service provision scenarios. Furthermore, each scenario may include, but is not limited to, sub-scenarios such as welfare activity promotion, overdue reminders, and satisfaction surveys. Each scenario is equipped with a corresponding device. This device is applied to an intelligent outbound call server. The intelligent outbound call server includes, but is not limited to, one of the following: intelligent outbound call device, intelligent outbound call system (local or cloud system), and intelligent outbound call server (local or cloud server). The intelligent outbound call device includes, but is not limited to, devices capable of intelligent outbound calls such as mobile phones, computers, and smart bracelets. Figure 1 As shown, the method may include the following operations: 101. When it is detected that the current conditions meet the pre-determined conditions for triggering an outbound call request for the target user, an outbound call request for the target user is sent to the target user, and real-time voice data input by the target user is collected according to the outbound call request.

[0023] In this embodiment of the invention, optionally, the target user can be any user who needs to perform intelligent outbound call interaction. Further optionally, when the real-time time reaches the target user's intelligent outbound call time, or when it is detected that an outbound call request needs to be triggered for the target user, or when other conditions that meet the corresponding intelligent outbound call requirements are met, it indicates that an outbound call request needs to be sent to the target user. Further optionally, intelligent outbound call interaction can be performed with multiple target users simultaneously.

[0024] 102. Based on a pre-trained speech analysis model, the real-time speech data of the target user is converted into text data. Then, based on a pre-trained intent analysis model, intent analysis is performed on the corresponding text data of the target user to obtain the intent analysis result of the target user. This intent analysis result includes the target intent. Optionally, the intent analysis result of the target user may also include the intensity of the target intent and / or key information in the text data, such as "camera" and "rental fee," for use in subsequent script data adjustments, which helps improve the accuracy of script data adjustments.

[0025] 103. Based on the pre-trained sentiment analysis model, perform sentiment analysis on the text data corresponding to the target user to obtain the sentiment analysis results of the target user, and generate a dialogue strategy that matches the target user based on the intent analysis results and sentiment analysis results of the target user.

[0026] 104. Based on the dialogue strategy corresponding to the target user, perform dynamic adjustment operations on the obtained dialogue data corresponding to the target user to obtain target dialogue data that matches the target user.

[0027] In this embodiment of the invention, optionally, different target users and different target intentions correspond to different script data. Furthermore, this script data can be pre-set or dynamically determined based on the aforementioned outbound call interaction process.

[0028] It is evident that implementation Figure 1 The described method involves sending an outbound call request to the user, collecting real-time voice data input by the user based on the request, converting the real-time voice data into text data based on a voice analysis model, performing intent analysis on the converted text data based on an intent analysis model to obtain the user's intent, and performing sentiment analysis on the converted text data based on a sentiment analysis model to obtain the user's sentiment analysis results. Based on the analyzed user intent and sentiment analysis results, a dialogue strategy matching the user is generated. Finally, according to the user's corresponding dialogue strategy, the acquired user-specific script data is dynamically adjusted to obtain script data matching the user. In other words, it not only generates accurate dialogue strategies for the user based on their intent but also combines the user's sentiment analysis results to adjust the script data prepared for the user. This improves the accuracy and precision of determining the script data required by the user, thereby improving the interactive accuracy, efficiency, and personalization of the intelligent outbound call dialogue process, enhancing user participation and satisfaction, and ultimately improving the accuracy, smoothness, and efficiency of intelligent outbound calls while reducing labor costs.

[0029] In this embodiment of the invention, optionally, converting the target user's real-time voice data into text data based on a pre-trained voice analysis model includes: The speech feature layer based on the pre-trained speech analysis model performs speech feature recognition on the real-time speech data of the target user to obtain the speech features of the speech data (such as Mel spectrum). Then, the multi-layer encoder based on the speech analysis model performs long-distance speech dependency capture on the speech features of the speech data to obtain the text data of the speech data. Finally, the decoder based on the speech analysis model performs text decoding and correction on the text data of the speech data to obtain the text data corresponding to the target user.

[0030] In this embodiment of the invention, optionally, intent analysis is performed on the text data corresponding to the target user based on a pre-trained intent analysis model to obtain the intent analysis result of the target user, including: A multi-layer feature extractor based on a pre-trained intent analysis model extracts contextual features of text data corresponding to the target user, and the classification layer of the intent analysis model analyzes the contextual features of the text data to obtain the intent analysis results of the target user.

[0031] In this embodiment of the invention, optionally, sentiment analysis is performed on the text data corresponding to the target user based on a pre-trained sentiment analysis model to obtain the sentiment analysis results for the target user, including: Based on the sequence processing layer of the pre-trained sentiment analysis model, all sentiment words used to express emotions in the text data corresponding to the target user are analyzed, and the sentiment analysis results of the target user are obtained.

[0032] In this embodiment of the invention, optionally, after obtaining real-time speech data, data preprocessing, such as noise reduction and frame segmentation, is performed. Further optionally, the speech analysis model is pre-trained on a base model (such as a recurrent neural network (RNN) or Transformer architecture) based on corresponding sample data. The sample data includes labeled speech data and corresponding text data. During training, the model parameters are optimized using a backpropagation algorithm to minimize the difference between the predicted sample text and the actual text. After training, the accuracy, robustness, and other performance characteristics of the speech analysis model are evaluated using validation data, and the hyperparameters of the speech analysis model (such as learning rate and batch size) are adjusted to improve the training accuracy of the speech analysis model, thereby improving the accuracy and reliability of the speech recognition and text analysis.

[0033] In this embodiment of the invention, optionally, text data is input into an intent analysis model for semantic and intent analysis to obtain the target user's intent from the text data. For example, if the text data corresponding to the user is "How much is the rental fee for this camera?", the intent of the text is analyzed by the classification layer of the intent analysis model as "Querying the rental fee for the camera". The intent analysis model is pre-trained on a base model (such as BERT or GPT) based on corresponding sample data. This sample data includes, but is not limited to, user consultation records and product manual data. Before training the base model, the pre-trained layers are frozen, and the base model is trained using the sample data. The parameters of multiple layers of the model (such as the output layer and specific layers) are adjusted to adapt it to the intent recognition and semantic understanding tasks of the rental industry. After training, the model's accuracy and robustness are also verified based on validation data.

[0034] In this embodiment of the invention, optionally, sentiment analysis is performed on the target user's text data using the sequence processing layer of a sentiment analysis model (such as an LSTM model) to obtain the corresponding sentiment analysis results. For example, if a user says, "This camera is too expensive, I'm not satisfied," the sequence processing extracts the sentiment features from the text, identifies sentiment keywords such as "too expensive" and "not satisfied," and outputs the user's sentiment analysis result as "negative." It should be noted that the sentiment in this invention includes the target user's feelings and emotions.

[0035] As can be seen, the embodiments of the present invention can also perform corresponding dimension data analysis on real-time voice data input by users based on voice analysis model, intent analysis model and sentiment analysis model in sequence, accurately analyze user intent and user current emotional state, which is conducive to further improving the accuracy and reliability of user dialogue strategy generation, thereby further improving the accuracy and reliability of user speech data determination; and through data analysis of different dimensions in sequence, even when facing complex semantics and / or dialects, it can accurately identify their intent and emotional state, that is, it can better cope with diverse user groups.

[0036] In this embodiment of the invention, optionally, a dynamic adjustment operation is performed on the obtained dialogue data corresponding to the target user according to the dialogue strategy corresponding to the target user, to obtain target dialogue data matching the target user, including: When the dialogue strategy corresponding to the target user is used to indicate that a voice dialogue needs to be conducted with the target user, the text features of the speech data to be provided to the target user are extracted based on the feature extraction layer of the pre-determined speech synthesis model (such as WaveNet or Tacotron), and the text features of the speech data are obtained. Then, the text features of the speech data are converted into speech features (such as Mel spectrum) based on the feature conversion layer of the speech synthesis model. Based on the sentiment analysis results of the target user and the obtained contextual dialogue data corresponding to the target user, the emotional needs parameters required by the target user are analyzed. Then, based on the speech synthesis layer of the speech synthesis model, the speech features corresponding to the dialogue data and the emotional needs parameters required by the target user are used to perform speech conversion analysis to obtain target dialogue data that matches the target user.

[0037] In this embodiment of the invention, the optional dialogue strategy for the target user includes interaction via text and / or voice. The voice interaction includes fully automated voice interaction or semi-voice interaction, that is, a combination of voice and human interaction.

[0038] In this embodiment of the invention, optionally, during the speech synthesis process, the speech synthesis model generates emotional demand parameters based on the context of the interaction with the target user and / or the context and sentiment analysis results in the target user's current text data. These emotional demand parameters include one or more of the following: intonation demand parameters, speech rate demand parameters, tone demand parameters, and timbre demand parameters. For example, if the user is in a negative mood, the system will respond with a gentler tone.

[0039] As can be seen, the embodiments of the present invention can also perform text feature analysis on the speech data to be output to the user based on the speech synthesis model, convert it into speech features, and add emotional needs that match the user to generate accurate speech data. This makes the generated speech data more natural and more in line with the user's emotions, which is conducive to further improving user participation and satisfaction, thereby further improving the smoothness and accuracy of outbound call interaction.

[0040] In an optional embodiment, the method may further include the following steps: For any target user, obtain the target user's user information and historical transaction data within the past target time period (e.g., 30 days), and determine the potential value of the target user based on the target user's user information and historical transaction data; For any target user, obtain the target user's historical interaction data within the target time period, and analyze the target user's historical response rate based on the historical interaction data; For any target user, obtain the target user's business data within the target time period. The business data corresponding to the target user includes the target user's contract data, historical customer service record data, and business scenario characteristic data. Based on the business data corresponding to the target user, analyze the business urgency that matches the target user. For any target user, based on the target user's current business needs, assign corresponding priority weights to the target user's potential value, historical response rate, and business urgency, and analyze the target user's business priority level based on the target user's potential value, historical response rate, business urgency, and corresponding priority weights. For any target user, outbound call task parameters matching the target user are generated based on the target user's business priority level. The target user's outbound call task parameters include the target user's outbound call time period, outbound call frequency, and call script template.

[0041] In this optional embodiment, the script data corresponding to the target user may include the script template of the target user; if the real-time moment reaches the outbound call period of the target user, or if the current outbound call count of the target user is less than the outbound call frequency of the target user, it indicates that the current conditions are met to trigger an outbound call request for the target user.

[0042] In this optional embodiment, for any target user, data can be collected from at least one of the following methods: a user management system, online forms, telephone records, etc. Optionally, the target user data includes the target user's user information, historical transaction data within a past target time period, and historical interaction data. The target user's user information includes basic information, purchasing power level data, consumption habit data, and user level. The target user's purchasing power level data indicates their purchase activity for goods and / or services within a past target time period; more purchases and / or larger purchase amounts indicate stronger purchasing power. The target user's basic information includes their identifier (e.g., name, identity) and contact information (e.g., telephone number, address). The target user's user level indicates their transaction level. The target user's consumption habit data includes purchase preference data and service demand data. The target user's historical transaction data includes historical transaction amounts, historical transaction frequency, and historical transaction object data, where the historical transaction object data includes type and / or identifier. Historical transaction objects include tangible goods and / or intangible services. Historical interaction data of target users includes at least one of the following: historical telephone interaction data, historical SMS interaction data, historical email interaction data, and historical activity participation data. The more historical interaction data available, the more accurate the analysis of historical response rates will be.

[0043] In this optional embodiment, further optionally, after collecting the target user's data, the data can be cleaned first, such as by deduplication, error correction (e.g., transaction object type), and missing data filling (e.g., address), to improve data accuracy, thereby improving the accuracy of intent recognition, and ultimately improving the accuracy and efficiency of intelligent outbound calling. Further optionally, the cleaned data can be stored in a relational database for subsequent querying and analysis. Further optionally, indexes can be created for key fields in the target user's data, such as historical transaction data, to further improve data query efficiency.

[0044] In this optional embodiment, the sum of all priority weights may be equal to 1, and may be preset. Current business needs include, but are not limited to, one or more of the following: welfare activity promotion needs, overdue reminder needs, satisfaction survey needs, etc. The priority weights may be set differently for different current business needs. For example, for welfare activity promotions, the priority weight corresponding to historical response rate is higher; for overdue reminders, the priority weight corresponding to business urgency is higher, without limitation.

[0045] As can be seen, this optional embodiment analyzes the user's potential value, historical response rate, and business urgency in sequence, and sets corresponding priority weights according to the user's current business needs, so as to accurately analyze the corresponding business priorities, thereby improving the accuracy of generating outbound call task parameters. This, in turn, helps to improve the accuracy of intelligent outbound call task allocation, and further helps to improve the accuracy and reliability of intelligent outbound call task allocation, so as to adapt to different business needs and market changes, and to improve the overall resource utilization efficiency and utilization rate.

[0046] In this optional embodiment, the potential value of the target user can be determined based on the target user's user information and historical transaction data, including: Based on the historical transaction amount and number of transactions of the target users, analyze the historical transaction frequency and historical average consumption amount of the target users within the target time period; Based on the historical transaction count and transaction data of the target users, analyze the repurchase rate of the target users within the target time period, and analyze the loyalty of the target users based on the repurchase rate and user level of the target users. Based on the target user's purchasing power and consumption habits data, the target user's historical transaction amount and number of transactions, predict the target user's lifecycle, and based on the target user's lifecycle and the target user's historical transaction amount, predict the target user's future lifecycle value. Analyze the potential value of target users based on their lifetime value, historical average spending, historical transaction frequency, and loyalty.

[0047] In this optional embodiment, for each transaction within the target time period, its type and / or identifier are recorded. By calculating the number of times the same type and / or identifier appears within the target time period and comparing it with the historical transaction count, the repurchase rate of the target user for that type and / or identifier is obtained. The target user's corresponding repurchase rate includes the historical repurchase rate for each type and / or identifier within the target time period.

[0048] In this optional embodiment, the loyalty of the target user is positively correlated with the repurchase rate and the user level of the target user. That is, the higher the repurchase rate and / or the higher the user level, the higher the loyalty.

[0049] In this optional embodiment, the target user's lifetime value is positively correlated with the target user's purchasing power and consumption habits, as well as the target user's historical transaction amount and number of transactions. Potential value is also positively correlated with the target user's lifetime value, historical average spending, historical transaction frequency, and loyalty.

[0050] As can be seen, this optional embodiment improves the accuracy and reliability of analyzing users' potential value by sequentially analyzing users' historical transaction frequency, historical average spending amount, repurchase rate, loyalty, and lifetime value, and then comprehensively analyzing these data. This further improves the accuracy and reliability of determining users' business priorities, and consequently improves the accuracy and reliability of intelligent outbound call task allocation, so as to adapt to different business needs and market changes, and is conducive to improving overall resource utilization efficiency and utilization rate.

[0051] In this optional embodiment, optionally, analyzing the target user's historical response rate based on historical interaction data includes: Analyze the response rate of each sub-data point in the historical interaction data, and analyze the historical response rate of the target user based on the response rates of all sub-data points.

[0052] In this optional embodiment, the historical response rate of the target user can be obtained by taking the average response rate of all sub-data, or by removing the average response rate after removing the response rate with large differences (such as the response rate with the smallest value), or by directly determining the maximum response rate, or by setting a corresponding weight for each sub-data and calculating the product of the weight of each sub-data and its response rate, and by calculating the sum of the products of all sub-data.

[0053] As can be seen, this optional embodiment improves the accuracy of the analysis of the user's historical response rate by analyzing the response rate of each sub-data such as historical telephone interaction data, historical SMS interaction data, historical email interaction data, and historical activity participation data separately, and then analyzing the overall response rate of all sub-data.

[0054] In this optional embodiment, optionally, the urgency of the business matching the target user is analyzed based on the business data corresponding to the target user, including: Based on the target user's contract data, analyze the contract amount and contract expiration time, and based on the target user's historical customer service record data, analyze the severity of the target user's business problems; Based on the business scenario characteristics data of the target users, analyze the urgency of the target users' service needs, and analyze the business urgency matching the target users based on the contract amount and contract expiration time, the severity of business problems and the urgency of service needs and the corresponding urgency weight.

[0055] In this optional embodiment, the historical customer service record data may include historical consultation data and / or historical complaint data, wherein the historical consultation data includes the type and content of historical consultations, and the historical complaint data includes the type and content of historical complaints.

[0056] In this optional embodiment, all urgency weights may be predetermined or dynamically determined based on the business type.

[0057] As can be seen, this optional embodiment analyzes the business urgency by sequentially analyzing contract data, the severity of business issues, the urgency of service needs, and the corresponding urgency weights. This improves the accuracy and reliability of business urgency analysis, thereby helping to further improve the accuracy of user task priority analysis and, consequently, the accuracy of intelligent outbound call task allocation.

[0058] In another alternative embodiment, the method may further include the following steps: Based on the target users' historical interaction data, historical transaction data, business data, and current business needs, analyze the business objectives that match the target users; among these, business objectives include, but are not limited to, one or more of the following: improving the conversion rate of high-value users, reducing the overdue bad debt rate, and controlling the cost of low-value users. Collect user data of target users, including the type of target users, which can be individual or organizational. When the target user type includes organizational, the user data also includes the size of the target user. The larger the size, the greater the priority weight of the potential value. Collect the historical number of outbound calls triggered against the target user within the target time period, the target number of successful outbound calls, the target number of successful conversions from outbound calls, and the time required for each successful conversion. Based on the business objectives matching the target users, user data, and the target users' repurchase rate, lifetime value, and historical outbound call count within the target time period, the target number of successful outbound calls, the target number of successful conversions, the duration required for each successful conversion, the contract amount, and the contract expiration time, the previously allocated priority weights are adjusted to obtain the adjusted priority weights. This includes analyzing the target user's business priority level based on their potential value, historical response rate, business urgency, and corresponding priority weights, including: Analyze the target user's business priority level based on the target user's potential value, historical response rate, business urgency, and adjusted priority weights.

[0059] For example, if a user's annual rent is 150,000 yuan (high historical spending), they are in a mature stage (stable repurchase rate), they are a listed company (significant industry influence), and their historical outbound call data indicates they are "considering long-term cooperation" (good future prospects), then their potential value takes precedence over historical conversion rate and business urgency. Similarly, if a user's rent is overdue (1 day remaining) and there is a 60% risk of bad debt (high risk), then business urgency takes precedence over potential value and historical response rate. Furthermore, if a user has a 90% success rate on 9 outbound calls out of the past 10 (90% success rate), 70% of successful calls result in conversion (high conversion rate), and the average conversion time is 12 hours (high timeliness), then historical response rate takes precedence over potential value and business urgency.

[0060] As can be seen, this optional embodiment first analyzes the user's matched business objectives, and then combines the business objectives with the user's attribute data, lifetime value, and repurchase rate to dynamically adjust and optimize the priority weights corresponding to potential value, business urgency, and historical response rate. This makes the priority weights more consistent with the actual situation such as business objectives, improves the accuracy of priority weight analysis, and thus helps to further improve the accuracy of user business priority analysis. In turn, it helps to further improve the accuracy and reliability of intelligent outbound call task resource allocation.

[0061] In this optional embodiment, optionally, based on business objectives matching the target user, user data, and the target user's repurchase rate, lifetime value, and historical outbound call count within the target time period, the target number of successful outbound calls, the target number of successful conversions, the duration required for each successful conversion, the contract amount, and the contract expiration date, an adjustment operation is performed on the aforementioned allocated priority weights to obtain adjusted priority weights, including: Based on the target user's user data and the target user's repurchase rate and lifetime value within the target time period, the previously allocated priority weights are adjusted to obtain the adjusted first priority weight. Based on the historical number of outbound calls and the target number of successful outbound calls, the outbound call success rate is analyzed. Based on the target number of outbound calls and the target number of successful conversions, the outbound call conversion rate is analyzed. The average outbound call conversion time required for each successful conversion is also calculated. Based on user data, outbound call success rate, outbound call conversion rate and average outbound call conversion time, the previously allocated priority weights are adjusted to obtain the adjusted second priority weights. Based on user data, contract amount and contract expiration time, the previously allocated priority weights are adjusted to obtain the adjusted third priority weights. Based on the business objectives that match the target users, the first priority weight, the second priority weight, and the third priority weight are adjusted to obtain the adjusted priority weights.

[0062] As can be seen, this optional embodiment can further improve the accuracy and reliability of priority weight optimization by adjusting and optimizing the aforementioned allocated priority weights from different dimensions, and then optimizing the adjusted and optimized priority weights again based on business objectives. This is conducive to further improving the accuracy and reliability of the analysis of users' business priority levels.

[0063] In yet another optional embodiment, the method may further include the following steps: Collect outbound call interaction data of target users within a preset time period (e.g., within 30 days). The outbound call interaction data corresponding to the target users includes the target users' feedback content, sentiment tags, text data, and dialogue content. Cluster the feedback content of all target users to obtain all feedback categories corresponding to all target users, and analyze the sentiment distribution of each feedback category based on the sentiment tags of all target users; Based on the sentiment distribution of each feedback category, the satisfaction level of that feedback category is analyzed, and a correlation analysis is performed on all feedback categories based on the dialogue content of all target users to obtain the attention of different types of target users to each feedback category. Based on the satisfaction level of each feedback category, the attention level of all feedback categories, and the text data and dialogue content of all target users, the outbound call response rate and response conversion rate are analyzed within a preset time period. The outbound call response rate is used to represent the ratio of the number of all target users to the number of intelligent outbound calls requested within the preset time period, and the response conversion rate is used to represent the ratio of the number of all target users who successfully transacted within the preset time period to the number of all target users who successfully received a response. Based on outbound call response rate, response conversion rate, satisfaction level of all feedback categories, sentiment distribution of all feedback categories, and corresponding attention levels, visualized outbound call data is generated for a preset time period. This visualized data is then displayed on a monitor.

[0064] In this optional embodiment, the emotion tags may be included in the aforementioned emotion analysis results. The emotion tags may contain multiple sub-tags representing different emotional states, such as very satisfied, satisfied, neutral, dissatisfied, and very dissatisfied. The more detailed the emotion tags, the higher the accuracy of the speech synthesis.

[0065] In this optional embodiment, after obtaining the outbound call interaction data, the feedback data is first converted into vector form, and then cluster analysis is performed on the vector-form feedback data to obtain all feedback categories. Vector transformation can be performed using any method capable of vector conversion, such as word frequency-inverse document frequency, and cluster analysis can be performed using any method capable of cluster analysis, such as K-means. Further optionally, all feedback categories include, but are not limited to, one or more of price type, service quality type, and product function type. The service quality type includes the intelligent interaction service quality type corresponding to the generated outbound call interaction data. Further, it also includes logistics and delivery service quality type and / or after-sales service quality type.

[0066] In this optional embodiment, association analysis can be performed in any manner, such as the Apriori algorithm. For example, by analyzing users' rental periods, the types of products purchased, and feedback content, it was found that 95% of users with rental periods exceeding 9 months reported rental discounts. Therefore, this association rule reveals the high level of attention long-term rental users pay to rental discounts, which can provide an important basis for subsequent marketing strategy adjustments.

[0067] In this optional embodiment, the outbound call visualization data may include outbound call visualization charts and reports, wherein the visualization charts include, but are not limited to, one or more of the following: bar charts, line charts, pie charts, etc. For example, a dashboard may be used to display the monthly trends of answer rates, conversion rates, satisfaction levels for each feedback category, and the real-time distribution of sentiment across different feedback categories.

[0068] As can be seen, this optional embodiment, after obtaining a large amount of outbound call interaction data from multiple users, can quickly and accurately identify the main focus areas of user feedback for different feedback categories by performing cluster analysis and sentiment analysis on all intelligent outbound call data. Furthermore, by conducting correlation analysis on different feedback categories, it can accurately understand the relationships between them, providing a precise and reliable basis for subsequent strategy adjustments and improving data utilization. Statistical analysis of key data such as satisfaction rate, outbound call response rate, and response conversion rate obtained from the aforementioned analysis can quantitatively evaluate the operational effectiveness of the intelligent outbound call system while providing intuitive and accurate visual data. This allows for a quick and intuitive understanding of the business operation status and user feedback, enabling more precise intelligent outbound call task allocation and more scientific and accurate decision-making.

[0069] In yet another optional embodiment, the method may further include the following steps: Based on the outbound call visualization data, the intelligent outbound call status of the system is monitored to obtain the system's resource usage monitoring data and the system's intelligent outbound call task data; Based on the system's resource usage monitoring data and the system's intelligent outbound call task data, determine whether the system's current conditions meet the pre-determined resource optimization conditions. When it is determined that the resource optimization conditions are met, a new target computing node is added to the system based on the system's resource usage monitoring data and the system's intelligent outbound call task data. Based on the system's resource usage monitoring data, the system's intelligent outbound call task data, outbound call visualization data, and target computing nodes, perform resource optimization operations on the system. The resource optimization operation includes at least one of the following: computing resource adjustment operation, outbound call strategy adjustment operation, script template adjustment operation, and system parameter setting operation.

[0070] In this optional embodiment, the system's resource usage monitoring data may include, but is not limited to, one or more of CPU monitoring data, memory monitoring data, and disk I / O monitoring data. The system's intelligent outbound call task data may include, but is not limited to, one or more of the following: the number of outbound call tasks, outbound call task completion data, outbound call task execution time, and outbound call task response time. Optionally, monitoring tools such as Prometheus and Grafana may be used. Prometheus monitors the intelligent outbound call status for efficient monitoring. Furthermore, charting tools such as Grafana may be used to display the monitored system resource usage data and the system's intelligent outbound call task data in intuitive chart form, allowing monitoring personnel to understand the system's operating status in real time. For example, the monitoring system can display the current CPU and memory usage rates, as well as the execution progress of outbound call tasks. If the execution time of a task exceeds a preset threshold, the monitoring system will issue an alarm to alert monitoring personnel.

[0071] In this optional embodiment, optionally, when CPU monitoring data indicates that CPU utilization and / or memory utilization are greater than or equal to the corresponding preset utilization, and / or the number of outbound call tasks is greater than or equal to the preset number of tasks, and / or the outbound call task response time is greater than or equal to the preset response time threshold, or other conditions are met, it indicates that the current conditions of the system meet the predetermined resource optimization conditions. Specifically, for computing resource adjustment operations: new computing nodes can be added through a flexible, scalable microservice architecture, and intelligent outbound call tasks can be evenly distributed across these nodes. Furthermore, for any given computing node, the number of intelligent outbound call tasks processed simultaneously is controlled to be less than or equal to the preset processing number to reduce system overload, thereby improving the processing efficiency and accuracy of intelligent outbound call tasks. For outbound call strategy adjustment operations / script template adjustment operations: by analyzing outbound call visualization data, feedback categories with user satisfaction levels less than or equal to the preset satisfaction threshold are obtained. Based on the type, satisfaction level, and sentiment analysis of the feedback category, an outbound call strategy adjustment strategy / script template adjustment strategy for that feedback category is generated. Then, the corresponding adjustment operation is executed according to the corresponding strategy, such as modifying the script template. Specifically, regarding system parameter settings: when the system's intelligent outbound call task data indicates that the execution duration of an outbound call task is greater than or equal to a preset execution duration threshold, a pre-trained machine learning model (such as a CNN model) analyzes the system's resource usage monitoring data and intelligent outbound call task data to obtain corresponding system parameter optimization strategies. Based on these strategies, system parameters are optimized to enhance the machine learning model's ability to learn parameter adjustments through interaction with the system, thereby achieving better performance. For example, the machine learning model dynamically adjusts the priority of intelligent outbound call task allocation and optimizes resource allocation strategies based on historical system resource usage monitoring data and intelligent outbound call task data, as well as current real-time system resource usage monitoring data and intelligent outbound call task data.

[0072] The following example illustrates this: If outbound call visualization data shows that in the price feedback category, the proportion of users dissatisfied with the rental price is greater than or equal to a preset proportion, the sentiment label is "negative," and the customer conversion rate for this feedback category is less than or equal to a preset conversion rate, and monitoring tools show that CPU and memory usage are high and task execution time is long when processing this type of user feedback, then based on this feedback and monitoring data, the system automatically adds new computing nodes through an elastically scalable microservice architecture and evenly distributes tasks across the nodes through an automatic load balancing module. Simultaneously, a concurrency control algorithm limits the number of tasks processed simultaneously, ensuring the system does not overload and maintains stable operation when handling high-concurrency user feedback. Next, based on a reinforcement learning model, the system generates and selects corresponding adjustment strategies (such as adjusting resource allocation, optimizing task scheduling strategies, and modifying script templates) based on the current state (e.g., resource usage, task execution progress, user feedback satisfaction, conversion rate, and other key indicators). For example, for price-sensitive users, the system adjusts the outbound call strategy, adding more information about rental discounts and optimizing script templates to improve user conversion rates and satisfaction. Meanwhile, the system optimizes resource allocation strategies, reduces task waiting time, and improves overall performance.

[0073] As can be seen, this optional embodiment monitors the system's intelligent outbound call status and analyzes the system's resource usage monitoring data, intelligent outbound call task data, and the aforementioned analyzed outbound call visualization data together to accurately analyze various resource adjustment strategies. This allows for precise resource optimization and adjustment operations, such as resource adjustment, outbound call strategy adjustment, script template adjustment, and system parameter setting, to further improve the accuracy, timeliness, and stability of intelligent outbound call task execution. This ensures efficient, stable, and accurate operation even during peak business periods, further enhancing the user's intelligent outbound call experience.

[0074] Example 2 Please see Figure 2 , Figure 2This is a flowchart illustrating another AI-based intelligent outbound call processing method disclosed in an embodiment of the present invention. This method can be applied to any scenario requiring intelligent outbound call processing, such as item rental, item purchase, or service provision scenarios. Furthermore, each scenario may include, but is not limited to, sub-scenarios such as welfare activity promotion, overdue reminders, and satisfaction surveys. Each scenario is equipped with a corresponding device. This device is applied to an intelligent outbound call server. The intelligent outbound call server includes, but is not limited to, one of the following: intelligent outbound call device, intelligent outbound call system (local or cloud system), and intelligent outbound call server (local or cloud server). The intelligent outbound call device includes, but is not limited to, devices capable of intelligent outbound calls such as mobile phones, computers, and smart bracelets. Figure 2 As shown, the method may include the following operations: 201. Collect outbound call interaction data of target users within a preset time period. The outbound call interaction data corresponding to the target users includes the target users' feedback content, sentiment tags, text data, and dialogue content.

[0075] 202. Perform clustering operations on the feedback content of all target users to obtain all feedback categories corresponding to all target users, and analyze the sentiment distribution of each feedback category based on the sentiment tags of all target users.

[0076] 203. Based on the sentiment distribution of each feedback category, analyze the satisfaction level of that feedback category, and conduct a correlation analysis on all feedback categories based on the dialogue content of all target users to obtain the attention of different types of target users to each feedback category.

[0077] 204. Based on the satisfaction level of each feedback category, the attention level corresponding to all feedback categories, and the text data and dialogue content of all target users, analyze the outbound call response rate and response conversion rate within the preset time period.

[0078] In this embodiment of the invention, the outbound call response rate is used to represent the ratio of the number of all target users to the number of outbound intelligent calls requested within a preset time period, and the response conversion rate is used to represent the ratio of the number of all target users who successfully transacted within a preset time period to the number of all target users who successfully received a response.

[0079] 205. Generate outbound call visualization data within a preset time period based on outbound call response rate, response conversion rate, satisfaction level of all feedback categories, emotional distribution of all feedback categories, and corresponding attention level.

[0080] It should be noted that for other related descriptions of steps 201-205, please refer to the description of the relevant content in Embodiment 1, which will not be repeated here.

[0081] It is evident that implementation Figure 3 The described method, after obtaining outbound call interaction data from multiple users, performs cluster analysis and sentiment analysis on all intelligent outbound call data. This allows for the rapid and accurate identification of the main focus areas of user feedback across different feedback categories. Furthermore, by conducting correlation analysis on different feedback categories, the method can accurately determine the relationships between them, providing a precise and reliable basis for subsequent strategy adjustments. Statistical analysis of key data such as satisfaction rate, outbound call response rate, and response conversion rate obtained from the aforementioned analysis enables a quantitative evaluation of the intelligent outbound call system's operational effectiveness while providing intuitive and accurate visualization data. This allows for a quick and intuitive understanding of the business operation status and user feedback.

[0082] Example 3 Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of an artificial intelligence-based intelligent outbound call processing device disclosed in an embodiment of the present invention. This device can be applied to any scenario requiring intelligent outbound call processing, such as item rental, item purchase, and service provision scenarios. Furthermore, each scenario may include, but is not limited to, sub-scenarios such as welfare activity promotion, overdue reminders, and satisfaction surveys. The device is applied to an intelligent outbound call server. The intelligent outbound call server includes, but is not limited to, one of the following: intelligent outbound call device, intelligent outbound call system (local system or cloud system), and intelligent outbound call server (local server or cloud server). The intelligent outbound call device includes, but is not limited to, devices capable of intelligent outbound calls such as mobile phones, computers, and smart bracelets. Figure 3 As shown, the device includes: The communication module 301 is used to send an outbound call request to the target user when it detects that the current conditions meet the predetermined conditions for triggering an outbound call request to the target user, and to collect real-time voice data input by the target user according to the outbound call request. The analysis module 302 is used to convert the real-time voice data of the target user into text data based on a pre-trained voice analysis model, and to perform intent analysis on the text data corresponding to the target user based on a pre-trained intent analysis model, so as to obtain the intent analysis result of the target user, which includes the target intent. Analysis module 302 is also used to perform sentiment analysis on text data corresponding to the target user based on a pre-trained sentiment analysis model, and obtain the sentiment analysis results of the target user; The generation module 303 is used to generate a dialogue strategy that matches the target user based on the intent analysis results and sentiment analysis results of the target user. The adjustment module 304 is used to perform dynamic adjustment operations on the obtained dialogue data corresponding to the target user according to the dialogue strategy corresponding to the target user, so as to obtain target dialogue data that matches the target user.

[0083] It is evident that implementation Figure 3 The described device sends outbound call requests to users and, based on these requests, collects real-time voice data input by the users. It then converts this voice data into text data using a voice analysis model, performs intent analysis on the converted text data using an intent analysis model to determine the user's intent, and performs sentiment analysis on the converted text data using a sentiment analysis model to obtain the user's sentiment analysis results. Based on the analyzed user intent and sentiment analysis results, it generates a dialogue strategy matching the user. Finally, it dynamically adjusts the acquired user-specific dialogue data according to the corresponding dialogue strategy to obtain dialogue data that matches the user. In other words, it not only generates accurate dialogue strategies for users based on their intent but also combines this with the user's sentiment analysis results, adjusting the dialogue data prepared for the user. This improves the accuracy and precision of determining the dialogue data required by the user, thereby enhancing the interactive accuracy, efficiency, and personalization of the intelligent outbound call dialogue process. It increases user participation and satisfaction, ultimately improving the accuracy, smoothness, and efficiency of intelligent outbound calls while reducing labor costs.

[0084] In this embodiment of the invention, optionally, the specific method by which the analysis module 302 converts the target user's real-time voice data into text data based on a pre-trained voice analysis model includes: The speech feature layer based on the pre-trained speech analysis model performs speech feature recognition on the real-time speech data of the target user to obtain the speech features of the speech data. The multi-layer encoder based on the speech analysis model performs long-distance speech dependency capture on the speech features of the speech data to obtain the text data of the speech data. The decoder based on the speech analysis model performs text decoding and correction on the text data of the speech data to obtain the text data corresponding to the target user. Specifically, the analysis module 302 performs intent analysis on the text data corresponding to the target user based on a pre-trained intent analysis model, and obtains the intent analysis results of the target user in the following ways: A multi-layer feature extractor based on a pre-trained intent analysis model extracts contextual features of text data corresponding to the target user, and analyzes the contextual features of text data based on the classification layer of the intent analysis model to obtain the intent analysis results of the target user. Specifically, the analysis module 302 performs sentiment analysis on the text data corresponding to the target user based on a pre-trained sentiment analysis model, and obtains the sentiment analysis results for the target user in the following ways: Based on the sequence processing layer of the pre-trained sentiment analysis model, all sentiment words used to express emotions in the text data corresponding to the target user are analyzed, and the sentiment analysis results of the target user are obtained.

[0085] As can be seen, the embodiments of the present invention can also perform corresponding dimension data analysis on real-time voice data input by users based on voice analysis model, intent analysis model and sentiment analysis model in sequence, accurately analyze user intent and user current emotional state, which is conducive to further improving the accuracy and reliability of user dialogue strategy generation, thereby further improving the accuracy and reliability of user speech data determination; and through data analysis of different dimensions in sequence, even when facing complex semantics and / or dialects, it can accurately identify their intent and emotional state, that is, it can better cope with diverse user groups.

[0086] In this embodiment of the invention, optionally, the adjustment module 304 performs a dynamic adjustment operation on the obtained dialogue data corresponding to the target user according to the dialogue strategy corresponding to the target user, and obtains the specific method of target dialogue data matching the target user, including: When the dialogue strategy corresponding to the target user is used to indicate that a voice dialogue needs to be conducted with the target user, the text features of the speech data to be provided to the target user are extracted based on the feature extraction layer of the pre-determined speech synthesis model, and the text features of the speech data are obtained. Then, the text features of the speech data are converted into speech features based on the feature conversion layer of the speech synthesis model. Based on the sentiment analysis results of the target user and the obtained contextual dialogue data corresponding to the target user, the emotional needs parameters required by the target user are analyzed. Then, based on the speech synthesis layer of the speech synthesis model, the speech features corresponding to the dialogue data and the emotional needs parameters required by the target user are used to perform speech conversion analysis to obtain target dialogue data that matches the target user.

[0087] As can be seen, the embodiments of the present invention can also perform text feature analysis on the speech data to be output to the user based on the speech synthesis model, convert it into speech features, and add emotional needs that match the user to generate accurate speech data. This makes the generated speech data more natural and more in line with the user's emotions, which is conducive to further improving user participation and satisfaction, thereby further improving the smoothness and accuracy of outbound call interaction.

[0088] In an optional embodiment, Figure 4 This is a schematic diagram of another intelligent outbound call processing device based on artificial intelligence disclosed in an embodiment of the present invention, as shown below. Figure 4 As shown, the device may further include: The acquisition module 305 is used to acquire, for any target user, user information and historical transaction data within the past target time period; The determination module 306 is used to determine the potential value of the target user based on the target user's user information and historical transaction data. The acquisition module 305 is also used to acquire, for any target user, the historical interaction data of the target user within a target time period; Analysis module 302 is also used to analyze the historical response rate of the target user based on historical interaction data; The acquisition module 305 is also used to acquire business data of any target user within a target time period. The business data corresponding to the target user includes the target user's contract data, historical customer service record data, and business scenario feature data. Analysis module 302 is also used to analyze the urgency of the business matching the target user based on the business data corresponding to the target user; The allocation module 307 is used to assign corresponding priority weights to any target user based on the target user's current business needs, including the target user's potential value, historical response rate, and business urgency. Analysis module 302 is also used to analyze the business priority level of the target user based on the target user's potential value, historical response rate, business urgency and corresponding priority weight; The generation module 303 is also used to generate outbound call task parameters matching the target user based on the target user's business priority level for any target user. The outbound call task parameters of the target user include the target user's outbound call time period, outbound call frequency, and script template. The script data corresponding to the target user includes the target user's script template. If the real-time moment reaches the target user's outbound call time period, or if the target user's current outbound call count is less than the target user's outbound call frequency, it indicates that the current conditions are met to trigger an outbound call request for the target user.

[0089] It is evident that implementation Figure 4 The described device improves the accuracy and reliability of analyzing a user's potential value by sequentially analyzing the user's historical transaction frequency, historical average spending amount, repurchase rate, loyalty, and lifetime value, and then comprehensively analyzing this data. This further improves the accuracy and reliability of determining the user's business priorities, and consequently improves the accuracy and reliability of intelligent outbound call task allocation, in order to adapt to different business needs and market changes, and to improve the overall resource utilization efficiency and rate.

[0090] In this optional embodiment, the target user's user information includes the target user's purchasing power level data, consumption habit data, and user level. The target user's historical transaction data includes the target user's historical transaction amount, historical transaction frequency, and historical transaction object data, wherein the historical transaction object data includes type and / or identifier. Figure 4 As shown, the determining module 306 determines the potential value of the target user based on the target user's user information and historical transaction data in the following specific ways: Based on the historical transaction amount and number of transactions of the target users, analyze the historical transaction frequency and historical average consumption amount of the target users within the target time period; Based on the historical transaction count and transaction data of the target users, analyze the repurchase rate of the target users within the target time period, and analyze the loyalty of the target users based on the repurchase rate and user level of the target users. Based on the target user's purchasing power and consumption habits data, the target user's historical transaction amount and number of transactions, predict the target user's lifecycle, and based on the target user's lifecycle and the target user's historical transaction amount, predict the target user's future lifecycle value. Analyze the potential value of target users based on their lifetime value, historical average spending, historical transaction frequency, and loyalty.

[0091] In this optional embodiment, the target user's purchasing power level data is used to represent the target user's purchases of goods and / or services in the past target time period. For example, the more purchases made and / or the larger the purchase amount, the stronger the purchasing power.

[0092] It is evident that implementation Figure 4 The described device improves the accuracy and reliability of analyzing a user's potential value by sequentially analyzing the user's historical transaction frequency, historical average spending amount, repurchase rate, loyalty, and lifetime value, and then comprehensively analyzing this data. This further improves the accuracy and reliability of determining the user's business priorities, and consequently improves the accuracy and reliability of intelligent outbound call task allocation, in order to adapt to different business needs and market changes, and to improve the overall resource utilization efficiency and rate.

[0093] In this optional embodiment, such as Figure 4 As shown, the target user's historical interaction data includes at least one of the target user's historical telephone interaction data, historical SMS interaction data, historical email interaction data, and historical activity participation data; The analysis module 302 analyzes the target user's historical response rate based on historical interaction data in the following specific ways: Analyze the response rate of each sub-data point in the historical interaction data, and analyze the historical response rate of the target user based on the response rates of all sub-data points.

[0094] It is evident that implementation Figure 4 The described device improves the accuracy of historical response rate analysis by analyzing the response rates of sub-data such as historical telephone interaction data, historical SMS interaction data, historical email interaction data, and historical activity participation data separately, and then analyzing the overall response rate of all sub-data.

[0095] Based on the business data corresponding to the target users, analyze the urgency of the business matching the target users, including: Based on the target user's contract data, analyze the contract amount and contract expiration time, and based on the target user's historical customer service record data, analyze the severity of the target user's business problems; Based on the business scenario characteristics data of the target users, analyze the urgency of the target users' service needs, and analyze the business urgency matching the target users based on the contract amount and contract expiration time, the severity of business problems and the urgency of service needs and the corresponding urgency weight.

[0096] It is evident that implementation Figure 4 The described device analyzes the urgency of business by sequentially examining contract data, the severity of business issues, the urgency of service needs, and their corresponding urgency weights. This improves the accuracy and reliability of the urgency analysis, thereby enhancing the accuracy of user task priority analysis and ultimately improving the accuracy of intelligent outbound call task allocation.

[0097] In another alternative embodiment, such as Figure 4 As shown, the device may further include: The data collection module 308 is used to collect outbound call interaction data of the target user within a preset time period. The outbound call interaction data corresponding to the target user includes the target user's feedback content, sentiment tags, text data, and dialogue content. Clustering module 309 is used to perform clustering operations on the feedback content of all target users to obtain all feedback categories corresponding to all target users; The analysis module 302 is also used to analyze the sentiment distribution of each feedback category based on the sentiment tags of all target users; The analysis module 302 is also used to analyze the satisfaction of each feedback category based on the sentiment distribution of each feedback category, and to perform correlation analysis on all feedback categories based on the dialogue content of all target users to obtain the attention of different types of target users to each feedback category. The analysis module 302 is also used to analyze the outbound call response rate and response conversion rate within a preset time period based on the satisfaction of each feedback category, the attention status corresponding to all feedback categories, and the text data and dialogue content of all target users. The outbound call response rate is used to represent the ratio of the number of all target users to the number of intelligent outbound requests within the preset time period, and the response conversion rate is used to represent the ratio of the number of all target users who successfully transacted within the preset time period to the number of all target users who successfully received a response. The generation module 303 is also used to generate outbound call visualization data within a preset time period based on outbound call response rate, response conversion rate, satisfaction of all feedback categories, emotional distribution of all feedback categories and corresponding attention.

[0098] It is evident that implementation Figure 4 After obtaining a large amount of outbound call interaction data from multiple users, the described device can quickly and accurately identify the main focus areas of user feedback in different feedback categories by performing cluster analysis and sentiment analysis on all intelligent outbound call data. By performing correlation analysis on different feedback categories, it can accurately know the correlation between different feedback categories, providing a precise and reliable basis for subsequent strategy adjustments and improving data utilization. Furthermore, by statistically analyzing key data such as satisfaction, outbound call response rate, and response conversion rate obtained from the aforementioned analysis, it can quantitatively evaluate the operating effect of the intelligent outbound call system and obtain intuitive and accurate visualization data. This allows for a quick and intuitive understanding of the business operation status and user feedback, thereby enabling more accurate intelligent outbound call task allocation and more scientific and precise decision-making.

[0099] In yet another alternative embodiment, such as Figure 5 As shown, the device may further include: The monitoring module 310 is used to monitor the intelligent outbound call status of the system based on the outbound call visualization data, and obtain the system's resource usage monitoring data and the system's intelligent outbound call task data. The judgment module 311 is used to determine whether the current conditions of the system meet the predetermined resource optimization conditions based on the system's resource usage monitoring data and the system's intelligent outbound call task data. The configuration module 312 is used to add new target computing nodes to the system based on the system's resource usage monitoring data and the system's intelligent outbound call task data when it is determined that the resource optimization conditions are met. Optimization module 313 is used to perform resource optimization operations on the system based on the system's resource usage monitoring data, the system's intelligent outbound call task data, outbound call visualization data, and target computing nodes; The resource optimization operation includes at least one of the following: computing resource adjustment operation, outbound call strategy adjustment operation, script template adjustment operation, and system parameter setting operation.

[0100] As can be seen, this optional embodiment monitors the system's intelligent outbound call status and analyzes the system's resource usage monitoring data, intelligent outbound call task data, and the aforementioned analyzed outbound call visualization data together to accurately analyze various resource adjustment strategies. This allows for precise resource optimization and adjustment operations, such as resource adjustment, outbound call strategy adjustment, script template adjustment, and system parameter setting, to further improve the accuracy, timeliness, and stability of intelligent outbound call task execution. This ensures efficient, stable, and accurate operation even during peak business periods, further enhancing the user's intelligent outbound call experience.

[0101] Example 4 Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of an intelligent outbound call server disclosed in an embodiment of the present invention. This server can be applied to any scenario requiring intelligent outbound call processing, such as item rental, item purchase, and service provision scenarios. Furthermore, each scenario may include, but is not limited to, sub-scenarios such as welfare activity promotion, overdue reminders, and satisfaction surveys. The server includes, but is not limited to, one of the following: intelligent outbound call device, intelligent outbound call system (local system or cloud system), and intelligent outbound call server (local server or cloud server). The intelligent outbound call device includes, but is not limited to, devices capable of intelligent outbound calls such as mobile phones, computers, and smart bracelets. ​ As shown, the server may include: Memory 401 storing executable program code; Processor 402 coupled to memory 401; Furthermore, it may also include an input interface 403 and an output interface 404 coupled to the processor 402; The processor 402 calls the executable program code stored in the memory 401 to execute the steps in the AI-based intelligent outbound call processing method described in Embodiment 1 or Embodiment 2.

[0102] Example 5 This invention discloses a computer storage medium storing computer instructions. When these computer instructions are invoked, they are used to execute the steps in the AI-based intelligent outbound call processing method described in Embodiment 1 or Embodiment 2.

[0103] Example 6 This invention discloses a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to perform the steps in the AI-based intelligent outbound call processing method described in Embodiment 1 or Embodiment 2.

[0104] The device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules; that is, they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0105] Through the detailed description of the above embodiments, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, including read-only memory (ROM), random access memory (RAM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), one-time programmable read-only memory (OTPROM), electrically-Erasable Programmable Read-Only Memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0106] Finally, it should be noted that the intelligent outbound call processing method, device, and server based on artificial intelligence disclosed in the embodiments of the present invention are merely preferred embodiments of the present invention and are only used to illustrate the technical solutions of the present invention, not to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent outbound call processing method based on artificial intelligence, characterized in that, The method is applied to multiple target users, wherein, for any one of the target users, the method includes: When it is detected that the current conditions meet the predetermined conditions for triggering an outbound call request for the target user, an outbound call request for the target user is sent to the target user, and real-time voice data input by the target user is collected according to the outbound call request; Based on a pre-trained speech analysis model, the real-time speech data of the target user is converted into text data, and based on a pre-trained intent analysis model, intent analysis is performed on the text data corresponding to the target user to obtain the intent analysis result of the target user, which includes the target intent. Based on a pre-trained sentiment analysis model, sentiment analysis is performed on the text data corresponding to the target user to obtain the sentiment analysis result of the target user. Based on the intent analysis result and the sentiment analysis result of the target user, a dialogue strategy matching the target user is generated. Based on the dialogue strategy corresponding to the target user, a dynamic adjustment operation is performed on the obtained dialogue data corresponding to the target user to obtain target dialogue data that matches the target user.

2. The intelligent outbound call processing method based on artificial intelligence according to claim 1, characterized in that, The process of converting the target user's real-time speech data into text data based on the pre-trained speech analysis model includes: The speech feature layer of the pre-trained speech analysis model is used to identify the speech features of the real-time speech data of the target user to obtain the speech features of the speech data. The multi-layer encoder of the speech analysis model is used to capture the long-distance speech dependency of the speech features to obtain the text data of the speech data. The decoder of the speech analysis model is used to decode and correct the text data of the speech data to obtain the text data corresponding to the target user. The step of performing intent analysis on the text data corresponding to the target user based on the pre-trained intent analysis model to obtain the intent analysis result of the target user includes: The multi-layer feature extractor based on the pre-trained intent analysis model extracts the contextual features of the text data corresponding to the target user, and analyzes the contextual features based on the classification layer of the intent analysis model to obtain the intent analysis result of the target user. The step of performing sentiment analysis on the text data corresponding to the target user based on a pre-trained sentiment analysis model to obtain the sentiment analysis results for the target user includes: Based on the sequence processing layer of the pre-trained sentiment analysis model, all sentiment words used to express emotions in the text data corresponding to the target user are analyzed, and the sentiment analysis results of the target user are obtained by analyzing all sentiment words based on the sequence processing layer of the sentiment analysis model.

3. The intelligent outbound call processing method based on artificial intelligence according to claim 1 or 2, characterized in that, The step of dynamically adjusting the obtained dialogue data corresponding to the target user based on the dialogue strategy corresponding to the target user to obtain target dialogue data matching the target user includes: When the dialogue strategy corresponding to the target user is used to indicate that a voice dialogue needs to be conducted with the target user, the text features of the speech data to be provided to the target user are extracted based on the feature extraction layer of the pre-determined speech synthesis model to obtain the text features of the speech data, and the text features of the speech data are converted into speech features based on the feature conversion layer of the speech synthesis model. Based on the sentiment analysis results of the target user and the obtained contextual dialogue data corresponding to the target user, the emotional needs parameters required by the target user are analyzed. Based on the speech synthesis layer of the speech synthesis model, speech conversion analysis is performed on the speech features corresponding to the speech data and the emotional needs parameters required by the target user to obtain target speech data that matches the target user.

4. The intelligent outbound call processing method based on artificial intelligence according to claim 1 or 2, characterized in that, For any target user, the method further includes: Obtain the user information and historical transaction data of the target user within the past target time period, and determine the potential value of the target user based on the user information and historical transaction data. Obtain the target user's historical interaction data within the target time period, and analyze the target user's historical response rate based on the historical interaction data; Obtain the business data of the target user within the target time period, and analyze the business urgency matching the target user based on the business data corresponding to the target user; Based on the current business needs of the target user, assign corresponding priority weights to the potential value, historical response rate, and business urgency of the target user, and analyze the business priority level of the target user based on the potential value, historical response rate, business urgency, and corresponding priority weights. Based on the target user's business priority level, outbound call task parameters matching the target user are generated, wherein the target user's outbound call task parameters include the target user's outbound call time period, outbound call frequency, and call script template; The target user's corresponding script data includes the target user's script template; If the target user's outbound call time period is reached in real time, or if the target user's current outbound call count is less than the target user's outbound call frequency, it indicates that the current conditions meet the conditions for triggering an outbound call request for the target user.

5. The intelligent outbound call processing method based on artificial intelligence according to claim 4, characterized in that, The target user's user information includes the target user's purchasing power level data, consumption habit data, and user level; the target user's historical transaction data includes the target user's historical transaction amount, historical transaction frequency, and historical transaction object data. The step of determining the potential value of the target user based on the target user's user information and historical transaction data includes: Based on the historical transaction amount and number of transactions corresponding to the target user, analyze the historical transaction frequency and historical average consumption amount of the target user within the target time period; Based on the historical transaction count and historical transaction data of the target user, analyze the repurchase rate of the target user within the target time period, and analyze the loyalty of the target user based on the repurchase rate and user level of the target user; Based on the target user's purchasing power level data and consumption habit data, the target user's historical transaction amount and historical transaction frequency, the target user's life cycle is predicted, and based on the target user's life cycle and the target user's historical transaction amount, the target user's future life cycle value is predicted. The potential value of the target users is analyzed based on their lifetime value, historical average spending, historical transaction frequency, and loyalty.

6. The intelligent outbound call processing method based on artificial intelligence according to claim 4, characterized in that, The step of analyzing the target user's historical response rate based on the historical interaction data includes: Based on each sub-data in the historical interaction data, analyze the response rate of that sub-data, and based on the response rates of all the sub-data, analyze the historical response rate of the target user; The step of analyzing the urgency of the business matching the target user based on the business data corresponding to the target user includes: Based on the target user's contract data, analyze the contract amount and contract expiration time, and based on the target user's historical customer service record data, analyze the severity of the target user's business problems; Based on the business scenario characteristic data of the target user, analyze the urgency of the target user's service needs, and based on the target user's contract amount and contract expiration time, the severity of business problems, the urgency of service needs and the corresponding urgency weight, analyze the business urgency matching the target user.

7. The intelligent outbound call processing method based on artificial intelligence according to any one of claims 1, 2, 5 and 6, characterized in that, The method further includes: Collect outbound call interaction data of the target user within a preset time period, wherein the outbound call interaction data corresponding to the target user includes the target user's feedback content, sentiment tags, text data and dialogue content; Clustering is performed on the feedback content of all the target users to obtain all feedback categories corresponding to all the target users, and the sentiment distribution of each feedback category is analyzed based on the sentiment tags of all the target users; Based on the sentiment distribution of each feedback category, the satisfaction level of that feedback category is analyzed, and a correlation analysis is performed on all feedback categories based on the dialogue content of all target users to obtain the attention of different types of target users to each feedback category. Based on the satisfaction level, attention level, and text data and dialogue content of all target users for each feedback category, analyze the outbound call response rate and response conversion rate within the preset time period; Based on the outbound call response rate, the response conversion rate, the satisfaction level of all feedback categories, the emotional distribution of all feedback categories, and the level of attention, visualized outbound call data for the preset time period is generated.

8. The intelligent outbound call processing method based on artificial intelligence according to claim 7, characterized in that, The method further includes: Based on the outbound call visualization data, the intelligent outbound call status of the system is monitored to obtain the resource usage monitoring data and the intelligent outbound call task data of the system. Based on the resource usage monitoring data and the intelligent outbound call task data of the system, determine whether the current conditions of the system meet the predetermined resource optimization conditions. When it is determined that the resource optimization conditions are met, a new target computing node is added to the system based on the system's resource usage monitoring data and the system's intelligent outbound call task data. Based on the system's resource usage monitoring data, the system's intelligent outbound call task data, the outbound call visualization data, and the target computing node, resource optimization operations are performed on the system.

9. An intelligent outbound call processing device based on artificial intelligence, characterized in that, The device includes: The communication module is used to send an outbound call request to the target user when it detects that the current conditions meet the predetermined conditions that require triggering an outbound call request to the target user, and to collect real-time voice data input by the target user according to the outbound call request; The analysis module is used to convert the real-time voice data of the target user into text data based on a pre-trained voice analysis model, and to perform intent analysis on the text data corresponding to the target user based on a pre-trained intent analysis model, so as to obtain the intent analysis result of the target user, wherein the intent analysis result of the target user includes the target intent. The analysis module is also used to perform sentiment analysis on the text data corresponding to the target user based on a pre-trained sentiment analysis model, and obtain the sentiment analysis result of the target user. The generation module is used to generate a dialogue strategy that matches the target user based on the intent analysis results and sentiment analysis results of the target user. The adjustment module is used to perform dynamic adjustment operations on the obtained dialogue data corresponding to the target user according to the dialogue strategy corresponding to the target user, so as to obtain target dialogue data that matches the target user.

10. An intelligent outbound call server, characterized in that, The server includes: Memory containing executable program code; A processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the intelligent outbound call processing method based on artificial intelligence as described in any one of claims 1-8.

Images