Information processing method and device, storage medium and program product

Abstract

The embodiment of the invention provides an information processing method and device, a storage medium and a program product. In the embodiment of the invention, the call service process is triggered based on the browsing duration of the browsing service page of the target user, and the service demand of the user is determined step by step by collecting the multi-stage dialogue content, so that the reply is accurately made and the candidate service information is pushed, the user demand is better met, and the user experience is improved. And the second service information in which the user is interested is further screened, so that the accuracy and conversion rate of service recommendation are improved. The AI dialogue model simulates the customer service staff to actively call the user, and establishes the audio and video channel for communication, so that potential requirements of the user can be responded in time, and more active and more personalized service experience is provided for the user. After the fact that the user agrees to communicate with the target merchant is determined, the merchant is automatically called through the AI model, and the audio and video channel is established, so that rapid connection between the user and the merchant is realized, communication links are reduced, and communication efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of computer processing technologies, and in particular, to an information processing method, device, storage medium, and program product. Background Art

[0002] In order to provide services to users more conveniently, many websites and application (App) clients of websites, etc., all provide users with online customer service. Users can consult questions through the online customer service, which is very convenient to use.

[0003] In the prior art, the server side of the online customer service can be monitored by human customer service. After receiving the questions sent by users through the online customer service, the human customer service answers them online. However, with the popularization of the use of the Internet, more and more users choose to consult questions through the online customer service, resulting in a large working pressure on the human customer service and being unable to meet the soaring online question consultations. Subsequently, an online intelligent customer service gradually emerged. The intelligent customer service can simulate human customer service to communicate with users, return the answers to the questions asked by users, and thus complete the online consultations of users. The online consultation process is very intelligent and very convenient for users to use. Further, the intelligent customer service can also analyze the needs of users according to the conversation content with users, and recommend appropriate service information to users according to the needs.

[0004] However, at present, the answers returned by the intelligent customer service to users are not highly adaptable to the questions asked by users, the recommendation of service information is not accurate enough, and the intelligent customer service does not provide subsequent follow-up services to users in a timely manner after recommending service information, resulting in low communication efficiency with users, affecting the user experience, and leading to a low conversion rate of services. Summary of the Invention

[0005] Multiple aspects of this application provide an information processing method, device, storage medium, and program product, which are used to provide personalized services to users, improve the accuracy of the answers of the intelligent customer service, improve the accuracy of the recommendation of service information, and provide subsequent follow-up services to users in a timely manner after recommending service information, improve the communication efficiency between the intelligent customer service and users, and improve the user experience and the conversion rate of service information.

[0006] An embodiment of the present application provides an information processing method, including: in response to an operation of a target user browsing a service page of a target application through a first terminal, obtaining the browsing duration of the target user for the service page; in the case where the browsing duration exceeds a first time threshold, calling an AI dialogue model based on artificial intelligence to simulate a customer service staff to call the first terminal, and an advertisement space for displaying first service information is included in the service page; in response to an answering instruction of the first terminal, establishing a first audio-video channel between the first terminal and a server; collecting first conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel, and determining target service requirement information of the target user according to the first conversation content; determining candidate service information that meets the target service requirement information from the service information provided by the server, and pushing the candidate service information to the first terminal for the target user to view the candidate service information during the process of communicating with the customer service staff simulated by the AI dialogue model; collecting second conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel for the candidate service information, and determining second service information that the target user is interested in from the candidate service information according to the second conversation content; collecting third conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel for the second service information, and in the case where it is determined according to the third conversation content that the target user agrees to communicate with a target merchant providing the second service information, calling the AI dialogue model to simulate the customer service staff to call a second terminal of the target merchant, and establishing a second audio-video channel between the first terminal and the second terminal through the server for the target user and the target merchant to conduct service communication based on the second audio-video channel.

[0007] An embodiment of the present application further provides an electronic device, including: a memory and a processor; the memory is used for storing a computer program; the processor is coupled with the memory and is used for executing the computer program to implement the steps in the above methods.

[0008] An embodiment of the present application further provides a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the processor is caused to implement the steps in the above methods.

[0009] An embodiment of the present application further provides a computer program product, and the computer program product includes computer programs / instructions, and when the computer programs / instructions are executed by a processor, the processor is caused to be able to implement the steps in the above methods.

[0010] In the embodiment of the present application, when the browsing duration of the target user on the target page exceeds the set duration, the AI dialogue model will be called to simulate a customer service staff to call the first terminal, and affect the answer instruction of the first terminal to establish a first audio-video channel with the first terminal, so that the AI dialogue model can communicate with the target user to enable the AI dialogue model to understand the needs of the target user in detail. That is, it can trigger the call service process based on the browsing duration of the target user, can respond to the potential needs of the target user in a timely manner, and provide more proactive and personalized services for users. After that, the AI dialogue model and the target user conduct progressive multi-stage conversations through the first audio-video channel, and gradually determine the service needs of the target user by collecting the conversation content. The more accurate the service needs of the target user are gradually obtained, the more accurate the reply information and service information recommendation of the AI dialogue model will be. Specifically, according to the content of the first-stage conversation, candidate service information that meets the target service need information of the target user is determined, and according to the content of the second-stage conversation related to the candidate service information, the target service information that the target user is interested in is determined. And when it is determined according to the content of the third-stage conversation related to the target service information that the target user agrees to establish communication with the target merchant providing the target service information, the AI dialogue model simulates a customer service staff to call the second terminal of the target merchant to establish a second audio-video channel between the first terminal and the second terminal, realizing the rapid connection between the target user and the target merchant, facilitating further communication between the target merchant and the target user, and reducing the communication links and improving the communication efficiency. Through staged conversations, the reply of the AI dialogue model is more targeted, improving the accuracy and reply efficiency of the AI dialogue model's reply; by determining the target service information that is most suitable for the target user's interests through multi-stage conversation content, and promoting further communication between the target user and the target merchant through double calling, the accuracy of service information recommendation, user experience, and conversion rate of target service information can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation of the present application. In the drawings:

[0012] Figure 1 It is a schematic flowchart of an information processing method provided by an exemplary embodiment of the present application;

[0013] Figure 2a It is a schematic flowchart of another information processing method provided by another exemplary embodiment of the present application;

[0014] Figure 2b It is a schematic diagram of storing question-and-answer pairs in a knowledge base provided by an exemplary embodiment of the present application;

[0015] Figure 2c Schematic diagram of model prompt words provided by an exemplary embodiment of the present application;

[0016] Figure 3 Schematic diagram of the structure of an electronic device provided by an exemplary embodiment of the present application. Detailed implementation manners

[0017] To make the objectives, technical solutions, and advantages of the present application clearer, the technical solutions of the present application will be clearly and completely described below in conjunction with specific embodiments of the present application and the corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without creative efforts shall fall within the protection scope of the present application.

[0018] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data for analysis, stored data, displayed data, etc.) involved in the present application are all information and data authorized by the user or fully authorized by all parties. Moreover, the collection, use, and processing of relevant data need to comply with relevant laws, regulations, and standards of relevant countries and regions, and corresponding operation entrances are provided for users to choose to authorize or reject.

[0019] Continuing from the background art, aiming at the current technical problems that the communication efficiency between intelligent customer service and users is not high, the answers returned to users are not highly adaptable to the questions asked by users, the recommended service information is not accurate enough, and no follow-up service is provided to users in a timely manner after recommending service information, resulting in a low conversion rate. In the embodiments of the present application, when the browsing duration of the target user on the target page through the first terminal exceeds the set duration, the AI dialogue model will be called to simulate a customer service staff to call the first terminal, and affect the answer instruction of the first terminal to establish a first audio-video channel between the first terminal, so that the AI dialogue model can communicate with the target user to enable the AI dialogue model to understand the needs of the target user in detail. That is, it can trigger the call service process based on the browsing duration of the target user, respond to the potential needs of the target user in a timely manner, and provide more proactive and personalized services to users. After that, the AI dialogue model and the target user have a progressive multi-stage dialogue through the first audio-video channel, and gradually determine the service needs of the target user by collecting the dialogue content. The more accurate the service needs of the target user are gradually obtained, the more accurate the reply information and service information recommended by the AI dialogue model will be. Specifically, according to the first-stage dialogue content, candidate service information that meets the target service needs of the target user is determined. According to the second-stage dialogue content related to the candidate service information, the target service information that the target user is interested in is determined. And when it is determined according to the third-stage dialogue content related to the target service information that the target user agrees to establish communication with the target merchant providing the target service information, the AI dialogue model simulates a customer service staff to call the second terminal of the target merchant to establish a second audio-video channel between the first terminal and the second terminal, realizing the rapid connection between the target user and the target merchant, facilitating further communication between the target merchant and the target user, and reducing the communication links and improving the communication efficiency. Through staged dialogue, the reply of the AI dialogue model is more targeted, improving the accuracy and reply efficiency of the AI dialogue model; by determining the target service information that is most adaptable to the interests of the target user through multi-stage dialogue content and promoting further communication between the target user and the target merchant through double calling, the accuracy of service information recommendation, user experience and conversion rate of target service information can be improved.

[0020] In summary, the call service process is triggered based on the browsing duration of the service page by the target user, and the service needs of the user are gradually determined by collecting multi-stage conversation content, so as to accurately make a reply and push candidate service information, better meet the user's needs, further screen out the second service information that the user is interested in, and improve the accuracy and conversion rate of service recommendation. By using the AI dialogue model to simulate the customer service staff to actively call the user and establish an audio-video channel for communication, the potential needs of the user can be responded to in a timely manner, providing a more proactive and personalized service experience for the user. After determining that the user agrees to communicate with the target merchant, the AI model automatically calls the merchant and establishes an audio-video channel, realizing a fast connection between the user and the merchant, reducing the communication links and improving the communication efficiency. Using the AI dialogue model to simulate the customer service staff reduces the dependence on manual customer service, reduces the labor cost of the enterprise, and at the same time ensures the continuity and stability of the service. Enhance user stickiness: Through proactive service and accurate recommendation, it can better meet the user's needs, improve the user's satisfaction and loyalty to the platform, and then enhance user stickiness and promote the long-term development of the platform.

[0021] The above solutions provided by the embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0022] Figure 1 It is a schematic flowchart of an information processing method provided by an exemplary embodiment of the present application. Figure 2a It is a schematic flowchart of another information processing method provided by another exemplary embodiment of the present application. As Figure 1 and Figure 2a shown, this method is applied to the server, and this method includes:

[0023] 101. In response to the operation of the target user browsing the service page of the target application through the first terminal, obtain the browsing duration of the target user for the service page;

[0024] 102. When the browsing duration exceeds the first time threshold, call the AI dialogue model based on artificial intelligence to simulate the customer service staff to call the first terminal. The service page includes an advertisement position where the first service information is displayed;

[0025] 103. In response to the answering instruction of the first terminal, establish a first audio-video channel between the first terminal and the server;

[0026] 104. Collect the first conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel, and determine the service requirement information of the target user according to the first conversation content;

[0027] 105. Determine the first candidate service information that meets the service demand information from the service information provided by the server, and push the first candidate service information to the first terminal for the target user to view the candidate service information during the communication with the customer service staff simulated by the AI dialogue model;

[0028] 106. Collect the second conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel regarding the first candidate service information, and determine the second service information that the target user is interested in from the first candidate service information according to the second conversation content;

[0029] 107. Collect the third conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel regarding the second service information, and when it is determined according to the third conversation content that the target user agrees to communicate with the target merchant providing the second service information, call the second terminal of the target merchant by the customer service staff simulated by the AI dialogue model, and establish a second audio-video channel between the first terminal and the second terminal through the server for the target user and the target merchant to conduct service communication based on the second audio-video channel.

[0030] In this embodiment, the implementation form of the first terminal of the target user is not limited. For example, the first terminal may be a smart handheld device, such as a smart phone, a tablet computer, a notebook computer or a desktop computer, etc.; for another example, the first terminal may also be a smart wearable device, such as a smart watch, a smart bracelet, etc.; for still another example, the first terminal may also be various smart home appliances with a display screen, such as a smart TV, a smart large screen or a smart robot, etc. In addition, various application programs may be installed on the first terminal. The application program may be an independently running APP, or a small program that depends on the independently running APP, or a web page, which is not limited herein. The application program may be, for example, a service application, a recruitment application or a recruitment app.

[0031] Similarly, in the embodiment of the present application, the implementation form of the second terminal device of the target merchant providing the target service information is not limited. The specific implementation form can refer to the relevant description of the first terminal and will not be elaborated herein. It should be noted that the first terminal and the second terminal may be the same type of terminal or different types of terminals, which is not limited in this embodiment.

[0032] The implementation form of the server in this embodiment is also not limited. The server may be, for example, a physical server, a single cloud server or a cloud server array.

[0033] In the embodiment of the present application, the target application installed on the first terminal includes a service page, and the service page includes an advertisement slot displaying first service information. The advertisement slot of the first service information can be one or more. In practical applications, the advertisement slot of the first service information can be a brief description of the corresponding service. In response to a trigger operation on any advertisement slot, a corresponding details page can be displayed, and the details page can include a detailed introduction to the first service information, and the detailed introduction can be presented in the form of an article or a post. Among them, the article or post is a text form used by the target application to display merchant service information, similar to an advertisement. Through a carefully designed title and content, it conveys the unique service content and advantages provided by the merchant, aiming to attract users' attention and convey relevant information. Taking the domestic service field as an example, the service information includes types such as domestic service, repair, second-hand, and recycling.

[0034] In the embodiment of the present application, in response to the operation that the target user browses the service page of the target application through the first terminal, the browsing duration of the target user on the service page can be obtained, and when the browsing duration exceeds the first time threshold, an AI dialogue model based on artificial intelligence can be called to simulate a customer service staff to call the first terminal. This embodiment does not limit the first time threshold, and the first time threshold can be, for example, several minutes, dozens of minutes, and so on. It can be understood that the longer the browsing duration of the target user on the service page, such as when the browsing duration exceeds the first time threshold, it indicates that the target user is a user with a demand for service information and is a potential user for placing an order. Then, the target user can be actively communicated with to understand the actual needs of the user, so as to recommend service information adapted to the user's needs according to the actual needs and improve the conversion rate of the service information.

[0035] In the embodiment of the present application, in response to the answer instruction of the first terminal, a first audio-video channel is established between the first terminal and the server, and the AI dialogue model can communicate with the target user through the first audio-video channel. The dialogue method can be a voice call, a video call, etc. The AI dialogue model communicates with the target user through the first audio-video channel, which can understand the actual needs of the user, so as to recommend service information adapted to the user's needs according to the actual needs and improve the conversion rate of the service information.

[0036] Among them, the AI dialogue model can be a large language model (LLMs) that supports multi-modal capabilities, with cross-modal and cross-language deep semantic understanding and generation capabilities. In the process of generating dialogue content with the target user, natural language processing can be used to search in the information library and / or generate content creation based on semantic understanding and other at least one method to generate question information for asking the target user and reply information for replying to the target user. The large language models include, but are not limited to, models of the GLM (Generalized Linear Model) series, models of the Qwen series, etc.

[0037] In the embodiment of the present application, the first conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-visual channel can be collected, and according to the first conversation content, the service demand information of the target user can be determined. The service demand information can include: the initial location information of the target user and the initial service category required, which can be referred to as the initial location information and the initial service category.

[0038] In some alternative embodiments, collecting the first conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-visual channel includes: receiving the first audio information sent by the target user through the first audio-visual channel, using an audio-text conversion model to convert the first audio information into first text information, where the first text information includes the first question information; determining the first reply information of the AI dialogue model according to the first question information, using the audio-text conversion model to convert the first reply information into second audio information, and sending the second audio information to the target user through the first audio channel; obtaining the first question information and the first reply information as the first conversation content. This embodiment does not limit the audio-text conversion model. For example, it can be the Transcribe model, the Whisper model, the DeepSpeech model, the Transformer-Transducer model, etc.

[0039] In this embodiment, during the audio-text process, due to the large number of texts with the same pronunciation or different dialects for the same text, and according to different business scenarios, there will be special words outside traditional vocabulary, such as exclusive words developed for the target application, such as Tian Tian Exhibition Booth, An Xin Tou, etc. For this reason, a hot word mechanism is introduced, that is, during the audio-text conversion process, when audio content with the same pronunciation as the hot word appears, the audio content can be converted into the text of the corresponding hot word.

[0040] Based on this, optionally, hot words in the target application are pre-maintained in the audio-text conversion model. Then, using the audio-text conversion model to convert the first audio information into the first text information includes: inputting the hot words and the first audio information into the audio-text conversion model, extracting the semantic features of each hot word and the semantic features of the first audio information, and calculating the first correlation between the semantic features of each hot word and the semantic features of the audio information. If there is a case where the first correlation is greater than the first threshold, determine the first target correlation from the first correlations greater than the first threshold, and generate the first text information based on the semantic features of the hot word corresponding to the first target correlation. If there is no case where the first correlation is greater than the first threshold, generate the corresponding first text information according to the semantic features of the first audio information. Among them, the first target correlation includes one correlation or multiple correlations. In practical applications, the first target correlation can be any one of the first correlations greater than the first threshold. For example, there can be multiple first correlations greater than the first threshold, and these correlations have a correlation ranking. Then, the first target correlation can be the correlation ranked first, or the correlation ranked last, or the correlation ranked in the middle. This embodiment does not limit this. Preferably, the correlation ranked first can be selected as the first target correlation. In addition, in the following embodiments, if it involves the correlation between different information and the situation of determining the target correlation (target data) from the correlations (or other data) greater than a certain threshold, the relevant descriptions of this embodiment can also be referred to.

[0041] In the above or following embodiments, the semantic feature can be a semantic feature vector. The correlation between two semantic features represents the degree of similarity between the two semantic features. The greater the correlation, the closer the two semantic features are, and the higher the similarity between the corresponding two words. Conversely, the lower the similarity. When the semantic feature is a semantic feature vector, the correlation between two semantic feature vectors refers to the vector distance between the two semantic feature vectors. The smaller the vector distance, the more similar the two words corresponding to the two semantic feature vectors are.

[0042] Optionally, generating the first text information based on the semantic features of the hot word corresponding to the first target correlation includes: using the semantic features of the hot word corresponding to the first target correlation as the semantic features of the first audio information, and generating the text information corresponding to the semantic features of the hot word corresponding to the first target correlation as the first text information. Generating the corresponding first text information according to the semantic features of the first audio information includes: generating the text information corresponding to the semantic features of the first audio information as the first text information.

[0043] Optionally, the AI dialogue model corresponds to a knowledge base, and multiple question-answer pairs are stored in the knowledge base. Each question-answer pair includes question information and answer information, such asFigure 2b The content described above is exemplary Q&A pairs stored in the knowledge base. Since there are a large number of Q&A pairs stored in the knowledge base, in order to determine the reply information corresponding to the first question information of the target user, a preliminary screening can be performed first from the knowledge base, that is, corresponding to Figure 2b the knowledge base recall stage in it, and Q&A pairs related to the first question information are preliminarily screened out (the recall is not empty). Based on this, the knowledge base can be associated with a fuzzy matching query function. The fuzzy matching query function can be the ElasticSearch (ES) indexing technology integrated with a relevance calculation function. The relevance calculation function can be the BM25 value calculation function. That is, the knowledge base has a pre-created efficient index. In the knowledge base recall stage, through this index, the first question information of the target user can be quickly retrieved and matched. ElasticSearch is a search engine based on an inverted index and can quickly process large-scale text data. Specifically, an index needs to be pre-created to store the data in the knowledge base. The field design of the index should include questions, answers, keyword fields, etc. The question information and reply information in a Q&A pair are located in different fields. Then, the data in the knowledge base is imported into the ElasticSearch index. For example, the Bulk API of ElasticSearch can be used to batch import data to improve the import efficiency. The data can include frequently asked questions, product details, service specifications, etc. BM25 (Best Matching 25) is a ranking function based on term frequency used to evaluate the relevance between a document and a query. BM25 takes into account two factors: term frequency (TF) and inverse document frequency (IDF). TF measures the frequency of a word appearing in a document, and IDF measures the importance of a word in the entire corpus. BM25 adjusts the weights of TF and IDF to balance the influence of term frequency and inverse document frequency, so as to more accurately evaluate the relevance of a document. When the user's question is successfully converted into text, the system will initiate a query to the knowledge base through ElasticSearch. Using the match query or multi_match query of ElasticSearch, combined with the BM25 algorithm, the user's question is fuzzy matched and sorted by relevance. ElasticSearch will calculate the relevance (relevance score) of each document according to the BM25 algorithm and return the results sorted from high to low by relevance. Among the returned results, the document with the highest relevance will be preferentially considered as the basis for answering.

[0044] Further optionally, after the Q&A pairs related to the first question information are preliminarily screened out, a relevance calculation model can be used for fine screening to obtain the fine-screened Q&A pairs. Among them, the relevance evaluation model can be, for example, the Bert model.

[0045] Based on the above content, according to the first question information, to determine the first reply information of the AI dialogue model, the optional implementation methods include: according to the first question information, using the fuzzy matching query function, screening out from the knowledge base the first reference Q&A pairs whose second relevance to the first question information is greater than the second threshold, that is Figure 2b the recall shown in is not empty; input the first question information and the first reference Q&A pairs into the relevance evaluation model to obtain the third relevance between the first question information and each first reference Q&A pair; if there is a second reference Q&A pair in the first reference Q&A pairs whose third relevance to the first question information is greater than the third threshold, then determine the second target relevance from the third relevance greater than the third threshold, and use the reply information included in the second reference Q&A pair corresponding to the second target relevance as the first reply information corresponding to the first question information, that is, if the output result of the Bert model in Figure 2 is "high relevance", then use the reply information corresponding to this relevance as the reply information corresponding to the first question information; where the second threshold is less than the third threshold.

[0046] In addition, if according to the first question information, using the fuzzy matching query function, no first reference Q&A pairs whose second relevance to the first question information is greater than the second threshold are screened out from the knowledge base, that is Figure 2b the recall shown in is empty, then directly call the AI dialogue model to generate the first reply information corresponding to the first question information.

[0047] In practical applications, the content included in the reference reply information contained in a reference Q&A pair in the knowledge base may be relatively large, which may affect the relevance to the first question information. For example, the first question information of the target user is: "May I ask how much the target domestic helper earns per day?" If the reference reply information is "The target domestic helper earns 100 yuan per day", then the relevance of this reference reply information to the first question information is very high. If the reference reply information is "The target domestic helper earns 100 yuan per day. She can cook, wash dishes, clean the house, and help take care of children when necessary", then the relevance of this reference reply information to the first question information is relatively low. Although the reference reply information contains the answer corresponding to the first question information, the content "She can cook, wash dishes, clean the house, and help take care of children when necessary" affects the relevance of this reference reply information to the first question information. In this case, the help of the AI dialogue model is needed to help determine whether the reference reply information contains effective information for answering the first question information and whether it can solve the doubts of the target user. Based on this, if there is a third reference Q&A pair in the first reference Q&A pair whose third relevance to the first question information is greater than the fourth threshold and less than the third threshold, that is, the relevance to the first question information is less than that of the second reference Q&A pair, but there is still a certain relevance to the first question information, then the first question information and the third reference Q&A pair are input into the AI dialogue model. Under the guidance of the third reference Q&A pair, that is, under the guidance of the question information and reply information contained in the third reference Q&A pair, it is determined whether there is a fourth reference Q&A pair that matches the first question information in each third reference Q&A pair, and the third reference Q&A pair serves as both a prompt word and an object to be selected; if it exists, the reply information in the fourth reference Q&A pair is used as the first reply information corresponding to the first question information; if it does not exist, the AI dialogue model prompt word and the first question information are input into the AI dialogue model, and under the guidance of the AI dialogue model prompt word, the first reply information corresponding to the first question information is generated. The Q&A pairs are initially recalled using the fuzzy matching query function, and then accurately recalled using the relevance calculation model. During the accurate recall process, if there are recall results with relatively low relevance, the AI dialogue model will be called to determine whether there is a reference Q&A pair that matches the first question information in the relevant reference Q&A pairs. If it exists, the reference reply information in this reference Q&A pair can be directly replied to the target user through the large AI model.

[0048] As can be seen from the above embodiments, the fuzzy matching query function corresponds to a correlation calculation function; according to the first question information of the target user, using the fuzzy matching query function, the first reference question-answer pairs with a second correlation greater than the second threshold between the first question information and the knowledge base are screened out from the knowledge base, including: using the first correlation calculation function, calculating the fourth correlation between the first question information and the question information and answer information contained in each question-answer pair in the knowledge base respectively, obtaining the fifth correlation between the first question information and the question information contained in each question-answer pair and the sixth correlation between the first question information and the answer information contained in each question-answer pair; according to the fifth correlation and the sixth correlation, performing weighted calculation to obtain the second correlation between the first question information and each question-answer pair; screening out the question-answer pairs with a second correlation greater than the second threshold between the first question information and the knowledge base as the first reference question-answer pairs. Among them, the weighted calculation can be weighted sum calculation, square difference calculation or square sum calculation, etc. When calculating the correlation using the first correlation calculation function, the influence of the question information and answer information in the question-answer pair on the correlation is considered at the same time, which can improve the accuracy of screening question-answer pairs and the accuracy of the answer information provided by the AI dialogue model.

[0049] Optionally, input the first question information and the first reference question-answer pairs into a correlation evaluation model to obtain the third correlation between the first question information and each first reference question-answer pair, including: inputting the first question information and the first reference question-answer pairs into the first correlation evaluation model, extracting the semantic features of the first question information and the semantic features of the answer information contained in the first reference question-answer pairs, and calculating the correlation between the semantic features of the first question information and the semantic features of the answer information contained in the first reference question-answer pairs as the third correlation.

[0050] Further optionally, the first correlation evaluation model includes: a semantic feature extraction network layer and a correlation calculation network layer. Then, inputting the first question information and the first reference question-answer pairs into the first correlation evaluation model to obtain the third correlation between the first question information and each first reference question-answer pair includes: inputting the first question information and the first reference question-answer pairs into the semantic feature extraction network layer to extract the semantic features of the first question information and the semantic features of the answer information contained in the first reference question-answer pairs; inputting the semantic features of the first question information and the semantic features of the answer information contained in the first reference question-answer pairs into the correlation calculation network layer to obtain the third correlation between the first question information and each first reference question-answer pair.

[0051] In some alternative embodiments, collecting the first conversation content of the target user communicating with the customer service staff simulated by the AI conversation model through the first audio-video channel further includes: determining the second question information of the AI conversation model according to the historical conversation content in the current conversation between the target user and the AI conversation model; receiving the second audio information sent by the first terminal through the first audio-video channel, and using an audio-text conversion model to convert the first audio information into second text information, where the second text information includes a second reply information.

[0052] Correspondingly, the target service demand information includes: the initial location information of the user and the initial service category required; then, determining the service demand information of the target user according to the first conversation content includes: determining whether the first question information, the first reply information, the second question information, and the second reply information contain the initial location information of the user and the initial service category required; if the judgment result is yes, then using the initial location information and the initial service category required as the demand service information of the target user.

[0053] In an alternative embodiment, after obtaining the demand service information of the target user, the target service category and the target location information adapted to the demand service information of the target user can be selected from multiple standard service categories and multiple standard location information provided by the target application. Specifically, inputting the initial service category, the initial location information, the multiple standard service categories and the multiple standard location information provided by the target application into a relevance evaluation model, calculating the seventh relevance between the initial service category and the multiple target standard service categories and the eighth relevance between the initial location information and the multiple standard location information; if there is a target service category among the multiple standard service categories whose seventh relevance with the initial service category is greater than the sixth threshold, then determining the third target relevance from the seventh relevance greater than the sixth threshold, and using the standard service category corresponding to the third target relevance as the target service category; and, if there is a target location information among the multiple standard location information whose eighth relevance with the initial location information is greater than the seventh threshold, then determining the fourth target relevance from the eighth relevance greater than the seventh threshold, and using the standard location information corresponding to the fourth target relevance as the target location information of the user.

[0054] Furthermore, if the judgment result is no, that is, the first question information, the first reply information, the second question information, and the second reply information do not contain the initial location information of the user and the initial service category required, then call the AI conversation model to simulate the customer service staff to send a third question information to the user through the first audio-video channel, where the third question information is used to guide the user to provide a third reply information containing the initial location information and the initial service category.

[0055] For example, if the initial location information provided by the target user is "** Street", which belongs to "** City", and the initial service category is "Cleaning", which belongs to "Domestic Helper", and the standard location information provided by the target application does not include "** Street", but includes "** City", and the standard service category provided by the target application does not include "Cleaning", but includes "Domestic Helper", then the location information of the target user is "** City", and the target service category is "Domestic Helper".

[0056] Further optionally, the target application provides multiple standard service categories and multiple standard location information. Then, using the second correlation calculation model, calculate the seventh correlation between the initial location information and the multiple standard location information, and the eighth correlation between the initial service category and the multiple standard service categories; using the correlation calculation function, calculate the ninth correlation between the initial location information and the multiple standard location information, and the tenth correlation between the initial service category and the multiple standard service categories; according to the weights of the second correlation calculation model and the weights of the correlation calculation function, perform weighted calculation on the seventh correlation and the ninth correlation to obtain the first weighted correlation between the initial location information and the multiple standard location information; if there is a standard location information among the multiple standard location information whose first weighted correlation with the initial location information is greater than the fifth threshold, then determine the third target correlation from the first weighted correlations greater than the fifth threshold, and use the standard location information corresponding to the third target correlation as the target location information; according to the weights of the second correlation calculation model and the weights of the correlation calculation function, perform weighted calculation on the eighth correlation and the tenth correlation to obtain the second weighted correlation between the initial service category and the multiple standard service categories; if there is a standard service category among the multiple standard service categories whose second weighted correlation with the initial service category is greater than the sixth threshold, then determine the fourth target correlation from the second weighted correlations greater than the sixth threshold, and use the standard service category corresponding to the fourth target correlation as the target service category.

[0057] It should be noted that the first correlation calculation model and the second correlation calculation model can be the same model or different models. The first correlation calculation function and the second correlation calculation function can be the same function or different functions.

[0058] In some alternative embodiments, determining first candidate service information that meets the service demand information from the service information available from the server includes: screening out second candidate service information that is adapted to the target location information and the target service category from the service information available from the server; obtaining historical data of the second candidate service information, where the historical data at least includes: click-through rate and order volume; selecting, from the second candidate service information, candidate service information with historical data higher than a seventh threshold as third candidate service information; inputting the third service information, the click-through rate and order volume of each third service information into a multi-objective model to obtain the click-through rate and order rate of each third service information; and selecting, from the third service information, service information with a click-through rate and / or order rate meeting the seventh threshold as first service information. Among them, the multi-objective model may be an Mmoe model.

[0059] In addition, it should be noted that the AI dialogue model has pre-learned prompt words as shown in Figure 2c to guide the AI dialogue model to be closer to customer service staff and improve the user's communication experience.

[0060] In the technical solutions provided by the above embodiments of the present application, when the browsing duration of the target user browsing the target page through the first terminal exceeds the set duration, the AI dialogue model will be called to simulate a customer service staff to call the first terminal, and affect the answering instruction of the first terminal to establish a first audio-video channel with the first terminal, so that the AI dialogue model can communicate with the target user to enable the AI dialogue model to understand the needs of the target user in detail. That is, it can trigger the call service process based on the browsing duration of the target user, be able to respond to the potential needs of the target user in a timely manner, and provide more proactive and personalized services for users. After that, the AI dialogue model and the target user have a progressive multi-stage dialogue through the first audio-video channel, and gradually determine the service needs of the target user by collecting the dialogue content. The more accurate the service needs of the target user obtained gradually, the more accurate the reply information and service information recommendation of the AI dialogue model. Specifically, according to the content of the first-stage dialogue, candidate service information that meets the target service need information of the target user is determined, and according to the content of the second-stage dialogue related to the candidate service information, the target service information that the target user is interested in is determined. And when it is determined according to the content of the third-stage dialogue related to the target service information that the target user agrees to establish communication with the target merchant providing the target service information, the AI dialogue model simulates a customer service staff to call the second terminal of the target merchant to establish a second audio-video channel between the first terminal and the second terminal, realizing the quick link between the target user and the target merchant, facilitating further communication between the target merchant and the target user, and reducing the communication links and improving the communication efficiency. Through the staged dialogue, the reply of the AI dialogue model is more targeted, improving the accuracy and reply efficiency of the AI dialogue model's reply; by determining the target service information that is most suitable for the target user's interest through the multi-stage dialogue content, and promoting further communication between the target user and the target merchant through dual calls, the accuracy of service information recommendation, user experience, and the conversion rate of target service information can be improved.

[0061] In summary, the call service process is triggered based on the browsing duration of the browsing service page of the target user, and the service needs of the user are gradually determined by collecting multi-stage conversation content, so as to accurately make a reply and push candidate service information, better meet the user's needs, further screen out the second service information that the user is interested in, and improve the accuracy and conversion rate of service recommendation. By simulating the customer service staff to actively call the user through the AI dialogue model and establishing an audio-video channel for communication, the potential needs of the user can be responded to in a timely manner, providing a more proactive and personalized service experience for the user. After determining that the user agrees to communicate with the target merchant, the AI model automatically calls the merchant and establishes an audio-video channel, realizing a fast connection between the user and the merchant, reducing the communication links and improving the communication efficiency. Using the AI dialogue model to simulate the customer service staff reduces the dependence on manual customer service, reduces the labor cost of the enterprise, and at the same time ensures the continuity and stability of the service. Enhance user stickiness: Through proactive service and accurate recommendation, it can better meet the user's needs, improve the user's satisfaction and loyalty to the platform, and then enhance user stickiness and promote the long-term development of the platform.

[0062] Figure 3 Schematic diagram of the structure of the electronic device provided by the exemplary embodiment of the present application. As Figure 3 shown, it includes: a memory 30a and a processor 30b; the memory 30a is used for storing computer programs; the processor 30b is coupled to the memory 30a and is used for executing the computer program to implement the following steps:

[0063] In response to the operation of the target user browsing the service page of the target application through the first terminal, obtain the browsing duration of the target user on the service page; when the browsing duration exceeds the first time threshold, call the AI dialogue model based on artificial intelligence to simulate a customer service staff to call the first terminal. The service page includes an advertisement space displaying the first service information; in response to the answering instruction of the first terminal, establish a first audio-video channel between the first terminal and the server; collect the first conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel, and determine the service demand information of the target user according to the first conversation content; determine the first candidate service information that meets the service demand information from the service information provided by the server, and push the first candidate service information to the first terminal for the target user to view the first candidate service information during the communication with the customer service staff simulated by the AI dialogue model; collect the second conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel regarding the first candidate service information, and determine the second service information that the target user is interested in from the first candidate service information according to the second conversation content; collect the third conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel regarding the second service information, and when it is determined according to the third conversation content that the target user agrees to communicate with the target merchant providing the second service information, call the AI dialogue model to simulate a customer service staff to call the second terminal of the target merchant, and establish a second audio-video channel between the first terminal and the second terminal through the server for the target user and the target merchant to conduct service communication based on the second audio-video channel.

[0064] In some alternative embodiments, when the processor collects the first conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel, it is specifically configured to: receive the first audio information sent by the target user through the first audio-video channel, use an audio-text conversion model to convert the first audio information into first text information, and the first text information includes first question information; according to the first question information, determine the first reply information of the AI dialogue model, use the audio-text conversion model to convert the first reply information into second audio information, and send the second audio information to the target user through the first audio channel; obtain the first question information and the first reply information as the first conversation content.

[0065] In some alternative embodiments, hot words in the target application are pre-maintained in the audio-text conversion model; when the processor uses the audio-text conversion model to convert the first audio information into the first text information, it specifically is configured to: input the hot words and the first audio information into the audio-text conversion model, extract the semantic features of each hot word and the semantic features of the first audio information, and calculate the first correlation between the semantic features of each hot word and the semantic features of the audio information; if there is a case where the first correlation is greater than the first threshold, determine the first target correlation from the first correlations greater than the first threshold, and generate the first text information based on the semantic features of the hot word corresponding to the first target correlation; if there is no case where the first correlation is greater than the first threshold, generate the corresponding first text information according to the semantic features of the first audio information.

[0066] In some alternative embodiments, the AI dialogue model corresponds to a knowledge base, and multiple question-answer pairs are stored in the knowledge base. The knowledge base is associated with a fuzzy matching query function; when the processor determines the first reply information of the AI dialogue model according to the first question information, it specifically is configured to: according to the first question information, use the fuzzy matching query function to screen out the first reference question-answer pairs in the knowledge base whose second correlation with the first question information is greater than the second threshold; input the first question information and the first reference question-answer pairs into the first correlation evaluation model to obtain the third correlation between the first question information and each first reference question-answer pair; if there is a second reference question-answer pair in the first reference question-answer pairs whose third correlation with the first question information is greater than the third threshold, determine the second target correlation from the third correlations greater than the third threshold, and use the reply information included in the second reference question-answer pair corresponding to the second target correlation as the first reply information corresponding to the first question information; wherein, the second threshold is less than the third threshold; if there is a third reference question-answer pair in the first reference question-answer pairs whose third correlation with the first question information is greater than the fourth threshold and less than the third threshold, input the first question information and the third reference question-answer pair into the AI dialogue model, and under the guidance of the third reference question-answer pair, determine whether there is a fourth reference question-answer pair that matches the first question information in each third reference question-answer pair; if there is, use the reply information in the fourth reference question-answer pair as the first reply information corresponding to the first question information; if not, input the AI dialogue model prompt word and the first question information into the AI dialogue model, and under the guidance of the AI dialogue model prompt word, generate the first reply information corresponding to the first question information.

[0067] In some alternative embodiments, the fuzzy matching query function corresponds to a first correlation calculation function; when the processor uses the fuzzy matching query function to screen out from the knowledge base a first reference Q&A pair whose second correlation with the first question information is greater than a second threshold according to the first question information of the target user, it specifically is used for: using the first correlation calculation function to calculate respectively the fourth correlation between the first question information and the question information and the answer information included in each Q&A pair in the knowledge base, obtaining the fifth correlation between the first question information and the question information included in each Q&A pair and the sixth correlation between the first question information and the answer information included in each Q&A pair; performing a weighted calculation according to the fifth correlation and the sixth correlation to obtain the second correlation between the first question information and each Q&A pair; screening out from the knowledge base the Q&A pairs whose second correlation with the first question information is greater than the second threshold as the first reference Q&A pairs.

[0068] In some alternative embodiments, when the processor inputs the first question information and the first reference Q&A pairs into the first correlation evaluation model to obtain the third correlation between the first question information and each first reference Q&A pair, it specifically is used for: inputting the first question information and the first reference Q&A pairs into the first correlation evaluation model, extracting the semantic features of the first question information and the semantic features of the answer information included in the first reference Q&A pairs, and calculating the correlation between the semantic features of the first question information and the semantic features of the answer information included in the first reference Q&A pairs as the third correlation.

[0069] In some alternative embodiments, the target service demand information includes: the initial location information of the user and the initial service category required; when the processor collects the first conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel, it is also used for: determining the second question information of the AI dialogue model according to the historical conversation content in the current round of conversation between the target user and the AI dialogue model; receiving the second audio information sent by the first terminal through the first audio-video channel, and using the audio-text conversion model to convert the first audio information into the second text information, where the second text information includes the second answer information.

[0070] Correspondingly, when the processor determines the target service demand information of the target user according to the first conversation content, it specifically is used for: judging whether the first question information, the first answer information, the second question information, and the second answer information include the initial location information of the user and the initial service category required; if the judgment result is yes, then taking the initial location information and the initial service category as the service demand information of the user; if the judgment result is no, then calling the AI dialogue model to simulate the customer service staff to send the third question information to the user through the first audio-video channel, where the third question information is used to guide the user to provide the third answer information including the initial location information and the initial service category.

[0071] Further optionally, if the target application provides a plurality of standard service categories and a plurality of standard location information, the processor is further configured to use the second correlation evaluation model to calculate a seventh correlation between the initial location information and the plurality of standard location information, and an eighth correlation between the initial service category and the plurality of standard service categories; use the second correlation calculation function to calculate a ninth correlation between the initial location information and the plurality of standard location information, and a tenth correlation between the initial service category and the plurality of standard service categories; according to the weights of the second correlation evaluation model and the second correlation calculation function, perform weighted calculation on the seventh correlation and the ninth correlation to obtain a first weighted correlation between the initial location information and the plurality of standard location information; if there is standard location information among the plurality of standard location information whose first weighted correlation with the initial location information is greater than a fifth threshold, determine a third target correlation from the first weighted correlations greater than the fifth threshold, and use the standard location information corresponding to the third target correlation as the target location information; according to the weights of the second correlation evaluation model and the second correlation calculation function, perform weighted calculation on the eighth correlation and the tenth correlation to obtain a second weighted correlation between the initial service category and the plurality of standard service categories; if there is a standard service category among the plurality of standard service categories whose second weighted correlation with the initial service category is greater than a sixth threshold, determine a fourth target correlation from the second weighted correlations greater than the sixth threshold, and use the standard service category corresponding to the fourth target correlation as the target service category.

[0072] In some alternative embodiments, when the processor determines the first candidate service information that meets the service demand information from the service information provided by the server, it is specifically configured to: screen out the second candidate service information that is adapted to the target location information and the target service category from the service information provided by the server; obtain the historical data of the second candidate service information, where the historical data at least includes: click volume and order volume; select the candidate service information with historical data higher than a seventh threshold from the second candidate service information as the third candidate service information; input the third service information, the click volume and the order volume of each third service information into the multi-target model to obtain the click-through rate and the order rate of each third service information; select the service information with the click-through rate and / or the order rate meeting the seventh threshold from the third service information as the first service information.

[0073] Further, as Figure 3 shown, the server further includes: other components such as a communication component 30c, a display 30d, a power supply component 30e, and an audio component 30f. Figure 3 Only some components are schematically shown, and it does not mean that the electronic device only includes Figure 3 the components shown.

[0074] The detailed implementation manners and beneficial effects of the electronic device provided in the embodiments of the present application have been described in detail in the foregoing embodiments, and will not be elaborated herein.

[0075] An exemplary embodiment of the present application further provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to implement the steps in the foregoing method embodiments.

[0076] An exemplary embodiment of the present application further provides a computer program product, which includes computer programs / instructions that, when executed by a processor, cause the processor to be able to implement the steps in the foregoing method embodiments.

[0077] The above-mentioned memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk or an optical disc.

[0078] The above-mentioned communication component is configured to facilitate communication between the device where the communication component is located and other devices in a wired or wireless manner. The device where the communication component is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G / LTE, 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component further includes a near field communication (NFC) module to facilitate short-range communication. For example, the NFC module can be implemented based on radio frequency identification (RFID) technology, infrared data association (IrDA) technology, ultra-wideband (UWB) technology, Bluetooth (BT) technology and other technologies.

[0079] The above-mentioned display includes a screen, which may include a Liquid Crystal Display (LCD) and a Touch Panel (TP). If the screen includes a touch panel, the screen can be implemented as a touch screen to receive input signals from a user. The touch panel includes one or more touch sensors to sense touches, swipes, and gestures on the touch panel. The touch sensors can sense not only the boundaries of touch or swipe actions but also detect the duration and pressure associated with the touch or swipe operation.

[0080] The above-mentioned power supply component provides power for various components of the device where the power supply component is located. The power supply component may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power for the device where the power supply component is located.

[0081] The above-mentioned audio component can be configured to output and / or input audio signals. For example, the audio component includes a microphone (MIC), which is configured to receive external audio signals when the device where the audio component is located is in an operating mode, such as a call mode, a recording mode, and a voice recognition mode. The received audio signals can be further stored in a memory or transmitted via a communication component. In some embodiments, the audio component further includes a speaker for outputting audio signals.

[0082] Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-readable storage media (including but not limited to disk memory, Compact Disc Read-Only Memory (CD-ROM), optical memory, etc.) containing computer-usable program code.

[0083] The present application is described with reference to the flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, such that the instructions executed by the processor of the computer or other programmable data processing devices generate for implementing in the process Figure 1 one process or multiple processes and / or blocks Figure 1means for the functions specified in one or more blocks.

[0084] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction means that implements the process Figure 1 one process or more processes and / or blocks Figure 1 the functions specified in one or more blocks.

[0085] These computer program instructions may also be loaded onto a computer or other programmable data processing device, such that a series of operational steps are performed on the computer or other programmable device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable device provide for implementing the process Figure 1 one process or more processes and / or blocks Figure 1 the steps of the functions specified in one or more blocks.

[0086] In a typical configuration, a computing device includes one or more processors (Central Processing Unit, CPU), an input / output interface, a network interface, and memory.

[0087] The memory may include non-permanent memory in the form of computer-readable media, random access memory (Random Access Memory, RAM) and / or non-volatile memory such as read-only memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.

[0088] Computer-readable media include permanent and non-permanent, removable and non-removable media that can store information by any method or technology. Information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change random access memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, CD-ROM, digital versatile disc (DVD) or other optical storage, magnetic cassettes, tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory media such as modulated data signals and carrier waves.

[0089] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0090] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. An information processing method, characterized in that, Applied to the server side, the method includes: In response to the operation of the target user browsing the service page of the target application through the first terminal, obtain the browsing duration of the target user on the service page; When the browsing duration exceeds the first time threshold, call the AI dialogue model based on artificial intelligence to simulate a customer service staff to call the first terminal, and the service page includes an advertisement space displaying the first service information; In response to the answering instruction of the first terminal, establish a first audio-video channel between the first terminal and the server side; Collect the first conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel, and determine the service demand information of the target user according to the first conversation content; Determine the first candidate service information that meets the service demand information from the service information that can be provided by the server side, and push the first candidate service information to the first terminal for the target user to view the first candidate service information during the communication with the customer service staff simulated by the AI dialogue model; Collect the second conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel for the first candidate service information, and determine the second service information that the target user is interested in from the first candidate service information according to the second conversation content; Collect the third conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel for the second service information, and when it is determined according to the third conversation content that the target user agrees to communicate with the target merchant providing the second service information, call the AI dialogue model to simulate the customer service staff to call the second terminal of the target merchant, and establish a second audio-video channel between the first terminal and the second terminal through the server side for the target user to communicate with the target merchant based on the second audio-video channel.

2. The method according to claim 1, wherein Collecting the first conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-video channel includes: Receive the first audio information sent by the target user through the first audio-video channel, and use an audio-text conversion model to convert the first audio information into first text information, where the first text information includes the first question information; According to the first question information, determine the first reply information of the AI dialogue model, use the audio-text conversion model to convert the first reply information into second audio information, and send the second audio information to the target user through the first audio channel; Obtain the first question information and the first reply information as the first conversation content.

3. The method according to claim 2, wherein The hot words in the target application are pre-maintained in the audio-text conversion model; using the audio-text conversion model to convert the first audio information into first text information includes: Input the hot word and the first audio information into the audio-text conversion model, extract the semantic features of each hot word and the semantic features of the first audio information, and calculate the first correlation between the semantic features of each hot word and the semantic features of the audio information; If there is a case where the first correlation is greater than the first threshold, determine the first target correlation from the first correlations greater than the first threshold, and generate the first text information based on the semantic features of the hot word corresponding to the first target correlation; If there is no case where the first correlation is greater than the first threshold, generate the corresponding first text information according to the semantic features of the first audio information.

4. The method according to claim 2, wherein The AI dialogue model corresponds to a knowledge base, and multiple question-answer pairs are stored in the knowledge base. The knowledge base is associated with a fuzzy matching query function; determining the first reply information of the AI dialogue model according to the first question information includes: According to the first question information, use the fuzzy matching query function to screen out the first reference question-answer pairs from the knowledge base whose second correlation with the first question information is greater than the second threshold; Input the first question information and the first reference question-answer pairs into the first correlation evaluation model to obtain the third correlation between the first question information and each first reference question-answer pair; If there is a second reference question-answer pair in the first reference question-answer pairs whose third correlation with the first question information is greater than the third threshold, determine the second target correlation from the third correlations greater than the third threshold, and use the reply information included in the second reference question-answer pair corresponding to the second target correlation as the first reply information corresponding to the first question information; where the second threshold is less than the third threshold; If there is a third reference question-answer pair in the first reference question-answer pairs whose third correlation with the first question information is greater than the fourth threshold and less than the third threshold, input the first question information and the third reference question-answer pair into the AI dialogue model, and under the guidance of the third reference question-answer pair, determine whether there is a fourth reference question-answer pair that matches the first question information in each third reference question-answer pair; If there is, use the reply information in the fourth reference question-answer pair as the first reply information corresponding to the first question information; If not, input the AI dialogue model prompt word and the first question information into the AI dialogue model, and generate the first reply information corresponding to the first question information under the guidance of the AI dialogue model prompt word.

5. The method according to claim 4, wherein The fuzzy matching query function corresponds to a first correlation calculation function; according to the first question information of the target user, using the fuzzy matching query function to screen out the first reference question-answer pairs from the knowledge base whose second correlation with the first question information is greater than the second threshold includes: Using the first correlation calculation function, calculate the fourth correlation between the first question information and the question information and answer information included in each Q&A pair in the knowledge base respectively, to obtain the fifth correlation between the first question information and the question information included in each Q&A pair and the sixth correlation between the first question information and the answer information included in each Q&A pair; According to the fifth correlation and the sixth correlation, perform weighted calculation to obtain the second correlation between the first question information and each Q&A pair; Screen out the Q&A pairs in the knowledge base whose second correlation with the first question information is greater than the second threshold as the first reference Q&A pairs.

6. The method according to claim 4, wherein Input the first question information and the first reference Q&A pairs into the first correlation evaluation model to obtain the third correlation between the first question information and each first reference Q&A pair, including: Input the first question information and the first reference Q&A pairs into the first correlation evaluation model, extract the semantic features of the first question information and the semantic features of the answer information included in the first reference Q&A pairs, and calculate the correlation between the semantic features of the first question information and the semantic features of the answer information included in the first reference Q&A pairs as the third correlation.

7. The method according to claim 5, characterized in that, The target service demand information includes: the initial location information of the user and the required initial service category; collecting the first conversation content of the target user communicating with the customer service staff simulated by the AI dialogue model through the first audio-visual channel, further including: Determine the second question information of the AI dialogue model according to the historical conversation content in the current round of conversation between the target user and the AI dialogue model; Receive the second audio information sent by the first terminal through the first audio-visual channel, and use the audio-text conversion model to convert the first audio information into second text information, where the second text information includes the second answer information; Correspondingly, determine the target service demand information of the target user according to the first conversation content, including: Judge whether the first question information, the first answer information, the second question information, and the second answer information contain the initial location information of the user and the required initial service category; If the judgment result is yes, use the initial location information and the initial service category as the service demand information of the user; If the judgment result is no, call the AI dialogue model to simulate the customer service staff to send the third question information to the user through the first audio-visual channel, where the third question information is used to guide the user to provide the third answer information containing the initial location information and the initial service category.

8. The method according to claim 7, wherein The target application provides multiple standard service categories and multiple standard location information, further including: Use the second correlation evaluation model to calculate the seventh correlation between the initial location information and the multiple standard location information and the eighth correlation between the initial service category and the multiple standard service categories; Using the second correlation calculation function, calculate the ninth correlation between the initial location information and the multiple standard location information, and the tenth correlation between the initial service category and the multiple standard service categories; According to the weights of the second correlation evaluation model and the weights of the second correlation calculation function, perform weighted calculation on the seventh correlation and the ninth correlation to obtain the first weighted correlation between the initial location information and the multiple standard location information; if there is standard location information among the multiple standard location information whose first weighted correlation with the initial location information is greater than the fifth threshold, then determine the third target correlation from the first weighted correlations greater than the fifth threshold, and use the standard location information corresponding to the third target correlation as the target location information; According to the weights of the second correlation evaluation model and the weights of the second correlation calculation function, perform weighted calculation on the eighth correlation and the tenth correlation to obtain the second weighted correlation between the initial service category and the multiple standard service categories; if there is standard service category among the multiple standard service categories whose second weighted correlation with the initial service category is greater than the sixth threshold, then determine the fourth target correlation from the second weighted correlations greater than the sixth threshold, and use the standard service category corresponding to the fourth target correlation as the target service category.

9. The method according to claim 8, characterized in that, Determining first candidate service information that meets the service demand information from the service information provided by the server, including: From the service information provided by the server, filter out second candidate service information that is adapted to the target location information and the target service category; Obtain the historical data of the second candidate service information, where the historical data at least includes: click-through rate and order volume; Select candidate service information with historical data higher than the seventh threshold from the second candidate service information as third candidate service information; Input the third service information, the click-through rate and order volume of each third service information into a multi-objective model to obtain the click-through rate and order rate of each third service information; Select service information with a click-through rate and / or order rate that meets the seventh threshold from the third service information as the first service information.

10. An electronic device, characterized in that, Including: A memory and a processor; The memory is used to store a computer program; the processor is coupled to the memory and is used to execute the computer program to implement the steps in the method according to any one of claims 1-9.

11. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, the processor is caused to implement the steps in the method according to any one of claims 1-9.

12. A computer program product, characterized in that, The computer program product includes computer programs / instructions, and when the computer programs / instructions are executed by the processor, the processor is caused to be able to implement the steps in the method according to any one of claims 1-9.

Images