Online consultation method, system and device, storage medium and program product

Abstract

The embodiment of the invention provides an online consultation method, system and device, a storage medium and a program product. The method comprises the following steps that in response to a received session request of a user client, a target customer service type is selected from a manual customer service type and an intelligent customer service type according to a preset selection strategy, and the session request is sent by the user client in response to a triggering operation of a user for media information published by a publisher; the manual customer service type and the intelligent customer service type both have the probability of being selected in a preset selection strategy; a session is created according to the target customer service type, the receiver of the session is the target customer service of the publisher, and the target customer service belongs to the target customer service type; and response data for the first target session request is sent to the user client, so that the user client displays a consultation page, and the consultation page comprises the session box. According to the technical scheme provided by the embodiment of the invention, the media information access experience of the user can be effectively improved, so that the clue conversion rate of the media information is improved.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to an online consultation method, system, device, storage medium and program product. Background Art

[0002] With the rapid development of Internet technology, more and more users or organizations hope to publish media information through online media platforms to gain attention.

[0003] Taking Internet advertising as an example of media information, Internet advertising has gradually become a mainstream advertising media. Compared with traditional advertising media (such as TV advertising and paper advertising), Internet advertising has the advantages of wide coverage, strong initiative and enthusiasm, relatively low cost, high cost performance and strong interactivity. Therefore, Internet advertising is also increasingly favored by advertisers (such as merchants, enterprises, and hospitals).

[0004] Advertisers publish advertisements on media platforms (e.g., social platforms, search engine platforms). After users see the advertisements on the media platforms, they click on the advertisements and enter the consultation page of the advertisements. On the consultation page, they consult with the advertiser's manual customer service. In this solution, the workload of the advertiser's manual customer service is heavy. Summary of the invention

[0005] In view of the above problems, the present application is proposed to provide an online consultation method, system, device, storage medium and program product that solve the above problems or at least partially solve the above problems.

[0006] In a first aspect of the present application, an online consultation method is provided, which is applied to a consultation page server, comprising:

[0007] In response to receiving a first target session request sent by a user client, selecting a target customer service type from a manual customer service type and an intelligent customer service type according to a preset selection strategy, wherein the first target session request is sent by the user client in response to a trigger operation of a user on media information published by a publisher; in the preset selection strategy, both the manual customer service type and the intelligent customer service type have a probability of being selected;

[0008] According to the target customer service type, a first target session is created, where the recipient of the first target session is the target customer service of the publisher, and the target customer service belongs to the target customer service type;

[0009] Response data for the first target session request is sent to the user client so that the user client displays a target consultation page, wherein the target consultation page includes a session box of the first target session.

[0010] A second aspect of the present application provides an online consultation method, which is applied to a user client and includes:

[0011] At a first moment, in response to a triggering operation on the media information published by the publisher, a first consultation page is displayed, wherein the first consultation page includes a conversation box of a first conversation, and the recipient of the first conversation is an intelligent customer service of the publisher;

[0012] At the second moment, in response to a trigger operation on the media information, a second consultation page is displayed, the second consultation page includes a conversation box for a second conversation, and the recipient of the second conversation is a manual customer service of the publisher;

[0013] The first session and the second session are two different sessions.

[0014] The third aspect of the present application provides an online consultation method, comprising:

[0015] The user client displays the media information published by the publisher;

[0016] The user client sends a first target session request to the consultation page server in response to the user's triggering operation on the media information;

[0017] The consultation page server responds to the first target session request and selects a target customer service type from the manual customer service type and the intelligent customer service type according to a preset selection strategy; in the preset selection strategy, both the manual customer service type and the intelligent customer service type have a probability of being selected;

[0018] The consultation page server creates a first target session according to the target customer service type, wherein the recipient of the first target session is the target customer service of the publisher, and the target customer service belongs to the target customer service type;

[0019] The consultation page server sends response data to the user client for the first target session request;

[0020] In response to receiving the response data, the user client displays a target consultation page including a session box of the first target session.

[0021] The fourth aspect of the present application provides an online consultation system, comprising: a user client and a consultation page server; wherein:

[0022] The user client is used to: display the media information published by the publisher; in response to a trigger operation on the media information, send a first target session request to the consultation page server;

[0023] The consultation page server is configured to: in response to receiving the first target session request, select a target customer service type from the manual customer service type and the intelligent customer service type according to a preset selection strategy; in the preset selection strategy, both the manual customer service type and the intelligent customer service type have a probability of being selected; create a first target session according to the target customer service type, the recipient of the first target session is the target customer service of the publisher, and the target customer service belongs to the target customer service type; send response data to the user client for the first target session request;

[0024] The user client is further configured to display a target consultation page in response to receiving the response data, wherein the target consultation page includes a session box of the first target session.

[0025] In a fifth aspect of the present application, an electronic device is provided. The electronic device includes: a memory and a processor, wherein:

[0026] The memory is used to store programs;

[0027] The processor is coupled to the memory, and is used to execute the program stored in the memory to implement any of the methods described above.

[0028] In a sixth aspect of the present application, a computer-readable storage medium storing a computer program is provided, wherein the computer program can implement any of the above-described methods when executed by a computer.

[0029] In a seventh aspect of the present application, a computer program product is provided, comprising a computer program, which implements any of the methods described above when executed by a processor.

[0030] The technical solution provided by the embodiment of the present application does not completely replace the manual customer service with the dialogue with the user by the intelligent customer service, but partially replaces the manual customer service with the dialogue with the user by the intelligent customer service. In this way, the technical solution provided by the embodiment of the present application can not only use the intelligent customer service to reduce the workload of the manual customer service and improve the customer service response speed, but also use the manual customer service to directly provide professional manual customer service services to some users to improve the quality of customer service. In this way, the user's media information access experience can be effectively improved, thereby improving the lead conversion rate of media information, and then improving the promotion effect of media information. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0032] Figure 1 A schematic diagram of an online consultation system provided in accordance with an embodiment of the present application;

[0033] Figure 2 A first group of interface schematic diagrams provided for an embodiment of the present application;

[0034] Figure 3a A schematic diagram of the mining process of the portrait data provided in one embodiment of the present application;

[0035] Figure 3b A schematic diagram of the response process of the intelligent customer service provided in one embodiment of the present application;

[0036] Figure 4 A flowchart of an online consultation method provided in one embodiment of the present application;

[0037] Figure 5 A flowchart of an online consultation method provided in one embodiment of the present application;

[0038] Figure 6 A second group of interface schematic diagrams provided for an embodiment of the present application;

[0039] Figure 7 A flowchart of an online consultation method provided in one embodiment of the present application;

[0040] Figure 8 A schematic diagram of a layered architecture provided for an embodiment of the present application;

[0041] Fig. 9 A structural block diagram of an electronic device provided in one embodiment of the present application;

[0042] Fig.10 A structural block diagram of an electronic device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below according to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0044] In addition, some of the processes described in the specification, claims and the above-mentioned figures of the present application include multiple operations that appear in a specific order, and these operations may not be executed in the order in which they appear in this article or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any execution order. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this article are used to distinguish different messages, devices, modules, etc., do not represent the order of precedence, and do not limit the "first" and "second" to be different types.

[0045] 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 used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0046] First, the vocabulary involved in the embodiments of the present application is explained. It can be understood that this explanation is for a clearer understanding of the embodiments of the present application, and does not necessarily constitute a limitation on the embodiments of the present application.

[0047] Large Language Model (LLM): refers to a deep learning model trained with a large amount of text data, which can generate natural language text or understand the meaning of language text. Large language models can handle a variety of natural language tasks, such as text classification, question answering, and dialogue, and are an important path to artificial intelligence. Among them, dialogue refers to the natural language communication between large language models and humans, including understanding natural language, generating natural language, etc.

[0048] Intelligent customer service system: refers to a customer service system that uses artificial intelligence technology to automatically handle customer inquiries.

[0049] Third-party consulting tool: refers to a consulting tool provided by a third party, such as an instant messaging tool. In the embodiment of the present application, the third party can be understood as other providers different from the publisher and the provider of the media platform for the publisher.

[0050] Agent: refers to an agent that can perceive the environment and take actions to achieve specific goals. It can be software, hardware or a system with autonomy, adaptability and interaction capabilities.

[0051] Taking advertising promotion as an example, in the existing scheme, the advertiser creates a consultation page and places an advertisement, the user clicks on the advertisement to enter the consultation page, asks questions on the consultation page, and the advertiser's manual customer service answers the questions and obtains advertising clue information through the guide information, thereby realizing advertising conversion. The current scheme has the following problems: the workload of the advertiser's manual customer service is heavy, the limited number of manual customer service during the peak visit period leads to the inability to respond to user questions in time, and the manual customer service cannot respond to user questions during the rest period. In order to solve or partially solve the above technical problems, the embodiment of the present application provides an online consultation method. In this method, it is not that the manual customer service is completely replaced by the intelligent customer service to talk to the user, but that the manual customer service is partially replaced by the intelligent customer service to talk to the user. In this way, the technical solution provided by the embodiment of the present application can not only use the intelligent customer service to reduce the workload of the manual customer service and improve the customer service response speed, but also use the manual customer service to directly provide professional manual customer service services to some users to improve the customer service quality. In this way, the user's advertising access experience can be effectively improved, thereby improving the lead conversion rate of the advertisement, and then improving the promotion effect of the advertisement.

[0052] Before introducing the online consultation method provided in the embodiment of the present application, an online consultation system involved in the online consultation method is introduced.

[0053] Figure 1 This is a schematic diagram of an online consultation system provided by an embodiment of the present application. Figure 1 As shown, the online consultation system 100 includes: a user client 10 and a consultation page server 11.

[0054] The user client 10 is used to: display the media information published by the publisher; and send a first target session request to the consulting page server 11 in response to a trigger operation on the media information.

[0055] The user client 10 may be a software module, a hardware module or a combination of the two in a terminal device. The terminal device may be a mobile phone, a multimedia computer, a multimedia tablet, a desktop computer, a laptop computer, a notebook computer, a netbook computer, a tablet computer, an audio / video player, a digital camera / camcorder, an e-book device, a gaming device or any combination thereof.

[0056] In some embodiments, the user client 10 may be a browser client.

[0057] Wherein, media information refers to the media for transmitting information. Exemplarily, media information may include advertisements placed / published by advertisers, that is, media information is presented in the form of advertisements. When the media information is an advertisement, the publisher of the advertisement may be referred to as the advertiser.

[0058] like Figure 2 As shown in (a), the user client 10 displays media information 202 in a web page 200 provided by the media platform, for example, displaying titles, pictures, etc. related to the media information 202. Exemplarily, the web page includes a media information position, and the user client 10 can display the media information published by the publisher at the media information position included in the web page provided by the media platform.

[0059] Exemplarily, the media platform may also be referred to as a website platform.

[0060] Illustratively, a web page includes a web page.

[0061] In some embodiments, the webpage may be a recommended webpage (e.g., website homepage) provided by a media platform. In response to a user's operation to enter the recommended page, the recommended page is displayed, and the media information published by the publisher is displayed on the recommended page. In other embodiments, the page may be a search result page provided by a media platform. In response to a user's search operation performed in the search page provided by the media platform, the search result interface is displayed, and the media information published by the publisher is displayed on the search result interface. The content of the media information matches the search keyword provided by the search operation. Figure 2 As shown in (a) of FIG. 1 , the search result interface 200 displays a search box 201 and media information 202 published by the publisher. The search box 201 includes the search keyword "skin disease", and the advertisement content of the advertisement 202 published by the advertiser matches the search keyword "skin disease".

[0062] In response to a trigger operation, such as a click operation, on the media information displayed on the page provided by the media platform, the user client 10 sends a first target session request to the consulting page server 11. The first target session request may include an identifier of the publisher, which is used to uniquely identify the publisher.

[0063] In some embodiments, after the consultation page server 11 receives the first target session request, if the first target session request does not carry a user identifier, a user identifier is allocated to the user client 10, and the user identifier is returned to the user client 10 for storage by the user client 10. Exemplarily, the consultation page server 11 may carry the user identifier in the response data for the first target session request and send it to the user client 10. In this way, when the user client 10 subsequently sends a session request to the consultation page server 11, the user identifier may be carried in the session request. After the consultation page server 11 receives the first target session request, if the first target session request carries a user identifier, there is no need to allocate a user identifier to the user client.

[0064] Taking the user client 10 as a browser client as an example, the user client 10 can store the user identifier returned by the consultation page server in a cookie. Later, when the user client 10 needs to access the consultation page server again, the user identifier is obtained from the cookie and carried in the session request to be sent to the consultation page server. The lifetime or validity period of the cookie can be set according to actual needs, and the embodiment of the present application does not specifically limit this.

[0065] The consultation page server 11 is used to: respond to the first target session request and select a target customer service type from manual customer service types and intelligent customer service types according to a preset selection strategy.

[0066] In the preset selection strategy, both the manual customer service type and the intelligent customer service type have a probability of being selected. That is, in the preset selection strategy, the probability of the manual customer service type being selected is greater than 0 and less than 1, and the probability of the intelligent customer service type being selected is greater than 0 and less than 1. Optionally, the preset selection strategy (or hosting strategy) can be a selection strategy based on time period or a selection strategy based on request percentage. In some embodiments, these two selection strategies can be provided to the publisher for selection by the publisher.

[0067] In some embodiments, the consultation page server 11 can be used to: when the request time of the first session request is within a first preset time period, determine the intelligent customer service type as the target customer service type; when the request time of the first session request is within a second preset time period, determine the manual customer service type as the target customer service type; wherein there is no intersection between the first preset time period and the second preset time period.

[0068] The first preset time period and the second preset time period can be set according to actual needs, and the embodiment of the present application does not specifically limit this. For example, the first preset time period can be the rest time period of the manual customer service, such as: 0:00-8:00 and 17:00-24:00, and the second preset time period can be the working time period of the manual customer service, such as: 8:00-17:00.

[0069] In an embodiment of the present application, both the manual customer service type and the intelligent customer service type have a probability of being selected. When the request time is within the first preset time period, the intelligent customer service type will be selected, and when the request time is within the second preset time period, the manual customer service will be selected.

[0070] In some embodiments, the consultation page server 11 can be used to extract the target customer service type from the manual customer service type and the intelligent customer service type, taking the probability of drawing the intelligent customer service type as the first probability and the probability of drawing the manual customer service type as the second probability.

[0071] Among them, the first probability is greater than 0 and less than 1, and the second probability is greater than 0 and less than 1.

[0072] Exemplarily, the first probability and the second probability may be the same or different.

[0073] Exemplarily, the first probability is less than the second probability. In the embodiment of the present application, most conversations are provided with customer service by manual customer service, and a small part of the conversations are provided with customer service by intelligent customer service.

[0074] In an embodiment of the present application, a target customer service type is randomly selected from manual customer service types and intelligent customer service types.

[0075] In some embodiments, the consultation page server 11 can be used to: extract at least one historical question of the user from the historical conversation record, analyze the at least one historical question, determine the difficulty of the user's question, and select a target customer service type from the manual customer service type and the intelligent customer service type according to the difficulty of the user's question. For example: when the difficulty of the user's question is lower than a preset threshold, the intelligent customer service type can be determined as the target customer service type, and when the difficulty of the user's question is greater than or equal to the preset threshold, the manual customer service type is determined as the target customer service type.

[0076] After the target customer service type is selected, the consultation page server 11 is further used to: create a first target session according to the target customer service type.

[0077] The receiver of the first target conversation is the target customer service of the publisher, and the target customer service belongs to the target customer service type.

[0078] It should be noted that each session includes an initiator and a receiver of the session, wherein the initiator refers to the party that initiates the session, and the receiver refers to the party that receives the session.

[0079] The recipient of the first target session is the target customer service of the publisher, and the initiator is the user using the user client 10 .

[0080] When the target customer service type is an intelligent customer service type, the receiver of the first target session is the intelligent customer service of the publisher; when the target customer service type is a manual customer service type, the receiver of the first target session is the manual customer service of the publisher.

[0081] In some embodiments, the process of the consultation page server 11 creating the first target session includes: generating a session identifier of the first target session; for the first target session, constructing an association relationship between the session identifier of the first target session and the identifiers of the initiator and the receiver of the first target session. Among them, the session identifier is used to uniquely identify the session, the identifier of the initiator is used to uniquely identify the initiator, for example: the user identifier mentioned above, and the identifier of the receiver is used to uniquely identify the receiver, for example: the identifier of the intelligent customer service or the customer service identifier of the manual customer service. Since one publisher corresponds to one intelligent customer service, the identifier of the publisher can be used to identify the intelligent customer service, that is, the identifier of the intelligent customer service is the identifier of the publisher.

[0082] After the first target session is successfully created, the consultation page server 11 is further used to: send response data for the first target session request to the user client 10 .

[0083] The response data may include a session identifier of the first target session and web page data of the target consultation page.

[0084] The user client 10 is further configured to display a target consultation page in response to receiving the response data, wherein the target consultation page includes a session box of the first target session.

[0085] Among them, the user client 10 can generate and display the target consultation page according to the web page data of the target consultation page.

[0086] like Figure 2 As shown in (b) , the user client 10 displays a target consultation page 203 , and the target consultation page 203 includes a conversation box 204 of the first target conversation.

[0087] The user can then enter his or her own question in the conversation box, and the user client 10 sends the question to the target customer service, and then displays the reply message of the target customer service to the question in the conversation box 204 .

[0088] In some embodiments, Figure 1 As shown, the online consultation system 100 may further include an intelligent customer service terminal 12. An intelligent agent of the publisher is running on the intelligent customer service terminal 12. The intelligent agent is a specific implementation form of the intelligent customer service.

[0089] When the target customer service type is an intelligent customer service type and the target customer service is an intelligent customer service, after the first target session is successfully created, the consultation page server 11 sends a notification message to the agent of the publisher, and the notification message is used to notify that the first target session has been created. Optionally, the notification message may include a session identifier of the first target session.

[0090] The following is a detailed description of the conversation process between the user and the publisher's intelligent customer service. The conversation process includes the following steps:

[0091] S11 , in response to a question raised by the user in the conversation box of the first target conversation (ie, the second question mentioned above), the user client 10 sends the question to the consultation page server 11 .

[0092] Exemplarily, the user client 10 may send the question, the session identifier of the first target session, and the user identifier of the user to the consultation page server.

[0093] S12. The consultation page server 11 sends the question to the publisher's agent.

[0094] Exemplarily, when creating the first target session, the consultation page server 11 will construct an association relationship between the session identifier of the first target session, the user identifier, and the identifier of the publisher (that is, the identifier of the publisher's intelligent customer service) for the first target session. Therefore, in step S12, the consultation page server 11 can search for the association relationship constructed when creating the first target session based on the session identifier of the first target session, and determine the identifier of the publisher based on the association relationship; and send the question to the publisher's intelligent agent based on the identifier of the publisher.

[0095] S13. After receiving the question, the publisher's intelligent agent uses the large language model to generate answer information for the question.

[0096] In some embodiments, the publisher's agent generates prompt words based on the question; wherein the prompt words include the above-mentioned question; the prompt words are used to prompt the large language model to generate answer information for the question; the publisher's agent inputs the prompt words into the large language model to obtain the answer information output by the large language model.

[0097] For example, in addition to the above questions, the prompt may also include information such as role information, thought chain (i.e., process), and output restrictions. Role information is used to assign roles to the large language model, thought chain is used to guide the large language model to decompose complex problems into step-by-step reasoning processes, and output restrictions are used to limit the output content of the large language model to ensure that the answer is accurate and concise, and avoid fabrication and random answers.

[0098] In order to improve the efficiency of generating prompt words, a prompt word template can be set in advance, so that the intelligent agent of the subsequent publisher can generate prompt words according to the question and the prompt word template.

[0099] Optionally, the retrieval-augmented generation (RAG) technology can be used to improve the quality and accuracy of the response of the large language model. In one embodiment, a vector knowledge base and / or a text knowledge base can be constructed based on documents related to the publisher (the document can be provided by the publisher or collected). Exemplarily, the document can be segmented to obtain multiple segmented texts; the multiple segmented texts are vectorized to obtain segmentation vectors of each segmented text; the multiple segmented texts are segmented to obtain multiple segmented words of each segmented text; the corresponding relationship between each segmented text and its segmented vector in the multiple segmented texts is established in the vector knowledge base; the corresponding relationship between each segmented text and its multiple segmented words in the multiple segmented texts is established in the text knowledge base. That is, the vector knowledge base includes: the corresponding relationship between the segmented text and the segmented vector obtained after the segmented text is vectorized, and the text knowledge base includes: the corresponding relationship between the segmented text and the multiple segmented words obtained by segmenting the segmented text.

[0100] After receiving the above question, the publisher's intelligent agent matches the question in the vector knowledge base to obtain the first target segmented text; and matches the question in the text knowledge base to obtain the second target text segmented text.

[0101] Among them, the vector knowledge base and the text knowledge base can be the same knowledge base, that is, the knowledge base includes: the correspondence between the segmented text and the segmented vectors obtained after vectorizing the segmented text, and the correspondence between the segmented text and the multiple segmented words obtained by segmenting the segmented text.

[0102] Exemplarily, the question can be vectorized to obtain a question vector; the question vector is matched with a vector in a vector knowledge base, and the segmented text corresponding to the matched vector is determined as the first target segmented text. The matched vector may be one or more, and this embodiment of the application does not specifically limit this. The retrieval method provided in the embodiment of the application is called vector retrieval.

[0103] Exemplarily, the question can be segmented to obtain keywords to be matched; the keywords to be matched are matched with the segmented words in the text knowledge base, and the segmented text corresponding to the matched segmented words is determined as the second target segmented text. Among them, the matched segmented words may be one or more, and this embodiment of the application does not specifically limit this. The retrieval method provided in the embodiment of the application is called full-text retrieval (Elaticsearch, ES).

[0104] In this embodiment, vector search and full-text search are used to improve search efficiency.

[0105] In some embodiments, the first target segmented text and / or the second target segmented text may be added to the prompt word, which may help improve the accuracy of the response of the large language model.

[0106] In some embodiments, in order to make the intelligent customer service understand the publisher better and provide more targeted consultation, the intelligent agent obtains the portrait data of the publisher in response to receiving the above question; based on the question and the portrait data, the large language model is requested to output the answer message of the question. Optionally, the portrait data of the publisher is added to the prompt word as context. Among them, the portrait data of the publisher may include the basic attributes of the publisher, such as: the publisher's name (such as hospital name, company name), department, contact information and other information.

[0107] In this embodiment, the large language model can determine whether the user's question is within the scope of the publisher based on the portrait data, and respond within the scope of the domain, thereby improving the service quality of the intelligent customer service.

[0108] In an optional implementation, the session messages of the historical sessions in which the publisher's manual customer service participated can be obtained; and the publisher's portrait data can be extracted from the session messages of the historical sessions. In this way, the publisher's portrait can be extracted from the manual customer service's historical reply messages to assist the intelligent customer service, so that the intelligent customer service can better understand the publisher and provide more targeted consultation.

[0109] Exemplarily, two intelligent agents based on a large language model can be designed. One intelligent agent is used to extract core information from conversation messages of historical conversations in which human customer service has participated, and the other intelligent agent is used to integrate and summarize the core information to produce a portrait of the publisher. The role, skills, thinking process, input and output, constraints, sample examples and other information of the intelligent agent can be defined through prompt engineering.

[0110] like Figure 3a As shown, the first agent 31 is used to extract multiple core information 33 from the conversation messages 32 of historical conversations in which the publisher's manual customer service has participated, and the second agent 34 is used to integrate and summarize the multiple core information 33 to generate portrait data of the publisher.

[0111] Among them, the first agent 31 can generate prompt words based on the conversation messages 32 of the historical conversations in which the publisher's manual customer service has participated and the corresponding prompt word template, and input the prompt words into the large language model, thereby obtaining multiple core information 33 output by the large language model. Among them, the prompt word template can be obtained through the prompt project. Among them, the first agent is used to process the conversation messages of the historical conversation. Each time the agent is requested, the conversation messages of the historical conversation are input, which include the messages sent by the users of the historical conversation and the messages sent by the manual customer service. The agent analyzes the conversation messages of the historical conversation and extracts the core information from them. Taking the publisher as a hospital as an example, the core information may include but is not limited to the name, department and department phone number of the hospital to which the manual customer service mentioned in the historical conversation belongs. For example, each time the agent is requested, one or more conversation messages of historical conversations can be input.

[0112] Among them, the second agent 32 can generate prompt words according to multiple core information 33 and corresponding prompt word templates, and input the prompt words into the large language model to obtain the portrait data output by the large language model. Among them, the prompt word template can be obtained through the prompt project. The second agent is used to process the information extracted by the first customer service. Each time the agent is requested, the core information from multiple historical sessions extracted by the first agent is input. The agent summarizes, removes duplicates, and summarizes this information to generate the portrait data of the publisher. Taking the publisher as a hospital as an example, a portrait of the hospital is generated, including the name of the hospital, the departments owned by the hospital, and the telephone numbers of each department, etc.

[0113] In this embodiment, the publisher's portrait data can be automatically generated based on the publisher's manual customer service's historical conversation messages, without the publisher having to manually input the portrait data, which can reduce the publisher's operating costs.

[0114] In some embodiments, in order to ensure the quality of the dialogue, a banned word list may be generated and added to the prompt words. In this way, the large language model may generate a reply message based on the banned word list to ensure that the reply message does not include words in the banned word list.

[0115] In practical applications, the prompt word templates mentioned in the above embodiments can be obtained by calling a unified interface. For example, different prompt word templates can be obtained by calling a unified interface based on different calling parameters, which is convenient for use.

[0116] S14. The publisher's agent sends a reply message to the consultation page server.

[0117] Exemplarily, the publisher's agent sends a reply message, a session identifier of the first target session, and an identifier of the publisher to the consulting service end.

[0118] Optionally, before sending a reply message to the consultation page server, the publisher's agent can perform risk control on the reply message, for example, calling a banned word review interface to review whether there are banned words in the reply message. If there are banned words, modify the banned words in the reply message, and then send the modified reply message to the consultation page server.

[0119] like Figure 3b As shown, after receiving the question, the publisher's agent can perform the following steps:

[0120] a. Index enhancement.

[0121] b. Call the large language model.

[0122] c. Risk control.

[0123] In the above step a, the first target segmented text and the second target segmented text can be obtained through vector retrieval and full-text retrieval.

[0124] In the above step b, the question, the first target segmented text, the second target segmented text, the portrait data of the publisher and the banned word list can be added to the corresponding prompt word template to obtain the prompt word. Based on the prompt word, the large language model is called to output a reply message under the prompt of the prompt word.

[0125] In the above step c, the banned word review interface can be called to review whether there are banned words in the reply information. If there are banned words, the banned words in the reply information are modified.

[0126] S15. The consultation page server sends a reply message to the user client.

[0127] Exemplarily, when creating the first target session, the consultation page server 11 will construct an association relationship between the session identifier of the first target session, the user identifier, and the identifier of the publisher (that is, the identifier of the publisher's intelligent customer service) for the first target session. Therefore, in step S15, the consultation page server 11 can search for the association relationship constructed when creating the first target session based on the session identifier of the first target session, and determine the user identifier of the user based on the association relationship; and send the question to the user client based on the user identifier of the user.

[0128] S16: The user client displays the reply message in the conversation box of the first target conversation.

[0129] In some embodiments, the publisher's intelligent customer service can also optimize the user conversation experience through opening words, guiding words to promote conversion, and words after completing conversion.

[0130] Exemplarily, the publisher's agent generates an opening statement in response to receiving a notification message sent by the consulting page server 11 after the first target session is successfully created, and sends the opening statement to the consulting page server 11, which sends the opening statement to the user client 10, and the user client 10 displays the opening statement in the session box of the first target session. Figure 2 As shown in (b) in the figure, the opening line is "Hello, it's a pleasure to serve you. What could you like to consult?"

[0131] Exemplarily, when the number of interactions between the user and the publisher's intelligent customer service is greater than a first preset number threshold, a guidance phrase to promote conversion is generated, and the guidance phrase is sent to the consultation page server 11, which sends the guidance phrase to the user client 10, and the user client 10 displays the guidance phrase in the conversation box of the first target conversation. For example, the guidance phrase is "You can leave your contact information, and we will arrange for professionals to contact you later."

[0132] Exemplarily, when the user issues clue information in the conversation box of the first target conversation in response to the guiding speech, the publisher's agent generates a speech after the conversion is completed, and sends the converted speech to the consulting page server 11, which sends the converted speech to the user client 10, and the user client 10 displays the converted speech in the conversation box of the first target conversation. For example, the guiding speech is "We will arrange for professionals to contact you within 24 hours."

[0133] In actual applications, the publisher has the need to use a third-party consulting tool, that is, the publisher's customer service has the need to use a third-party consulting tool to uniformly manage conversations and media information clues. Figure 1 As shown, the online consultation system 100 may further include: a service end 12 of a third-party consultation tool.

[0134] When the target customer service type is a manual customer service type and the target customer service is a manual customer service, the process of the consultation page server 11 creating the first target session includes the following steps:

[0135] S21. Send a third target session request to a server of a third-party consulting tool.

[0136] Optionally, the third target session request may include an identifier of the first publisher.

[0137] Optionally, the third target session request may include a user identifier of a user using the user client 10 .

[0138] S22: Receive session information of a third target session returned by the server of the third-party consulting tool.

[0139] Among them, the third target session is created by the server of the third-party consulting tool in response to receiving the third target session request, the initiator of the third target session is the user using the user client 10, and the recipient of the third target session is the manual customer service of the publisher.

[0140] In one embodiment, in response to receiving the third target session request, the server of the third-party consulting tool may determine the publisher's manual customer service according to the publisher's identifier, and create the third target session after the publisher's manual customer service is determined.

[0141] In an optional implementation, according to the identifier of the publisher, the publisher's manual customer service is determined from one or more candidate manual customer service of the publisher. Exemplarily, a manual customer service whose current number of user receptions is less than a preset reception number threshold is selected from one or more candidate manual customer service of the publisher, and the manual customer service whose current number of user receptions is less than the preset reception number threshold is allocated to the user. The current number of user receptions of the manual customer service refers to the number of users who need the manual customer service to provide customer service.

[0142] The process of the third-party consulting tool creating a third target session includes: generating a session identifier of the third target session, and for the third target session, constructing an association relationship between the session identifier of the third target session, the user identifier of the user using the user client 10, and the customer service identifier of the publisher's manual customer service.

[0143] The session information of the third target session may include the session identifier of the third target session, the second user identifier, and the customer service identifier of the manual customer service of the publisher. The customer service identifier may be a customer service name or a customer service number.

[0144] S23: Create a first target session according to the session information of the third target session.

[0145] The initiator of the first target session is a user using the user client 10 , and the receiver of the first target session is the manual customer service.

[0146] Optionally, a first target session is created according to the customer service identifier of the publisher's manual customer service and the user identifier of the user using the user client 10. The creation process may include: generating a session identifier of the first target session, and building an association relationship among the session identifier of the first target session, the customer service identifier of the publisher's manual customer service, and the user identifier of the user using the user client 10 for the first target session.

[0147] S24: Establish a mapping relationship between the first target session and the third target session (ie, the second mapping relationship mentioned above).

[0148] The mapping relationship is used to use the message sent by the manual customer service in the third target session as the message sent by the manual customer service in the first target session and to use the message sent by the user in the first target session as the message sent by the user in the third target session.

[0149] Among them, the mapping relationship is used when the consultation page server 11 performs message forwarding. For example, when the consultation page server 11 receives a message sent by a user in the first target session from the user client 10, according to the mapping relationship, the message is forwarded to the server 13 of the third-party consulting tool as a message sent by the user in the third target session. In this way, the server 13 of the third-party consulting tool sends the message to the customer service client 14 of the publisher's manual customer service, and the customer service client displays the message in the session box of the third target session. For another example, when the consultation page server 11 receives a message sent by a manual customer service in the third target session from the server 13 of the third-party consulting tool, according to the mapping relationship, the message is forwarded to the user client 10 as a message sent by the manual customer service in the first target session. In this way, the user client 10 displays the message in the session box of the first target session.

[0150] In some embodiments, Figure 1 As shown, the online consultation system 100 may include a customer service client 14 of a publisher's manual customer service.

[0151] In actual applications, when the recipient of the first target session is the publisher's intelligent customer service (that is, the target customer service type is the intelligent customer service type and the target customer service is the intelligent customer service), it is possible that the intelligent customer service cannot answer the user's questions. In order to deal with these situations, after the first target session is created, the consultation page server 11 is used to perform the following steps:

[0152] S31. For the first target session, send a second target session request to a server of a third-party consulting tool.

[0153] Optionally, the second target session request may include an identifier of the issuing party.

[0154] Optionally, the second target session request may include a user identifier of a user using the user client 10 .

[0155] In an optional implementation, in response to the intelligent customer service being unable to reply to the user's question (ie, the first question mentioned above), a second target session request is sent to a server of a third-party consulting tool for the first target session.

[0156] Exemplarily, when the publisher's agent inputs the prompt words containing the question into the large language model and the large language model cannot respond, the above step S31 is executed.

[0157] In another optional implementation, when the number of interactions of the first target session is greater than the second preset number threshold, the above step S31 is performed. The number of interactions can be understood as the number of messages currently existing in the first target session, that is, the sum of the number of messages sent by the initiator of the first target session in the first target session and the number of messages sent by the receiver of the first target session in the first target session. The size of the second preset number threshold can be set according to actual needs, and the embodiment of the present application does not specifically limit this.

[0158] S32: Receive session information of the second target session returned by the server of the third-party consulting tool.

[0159] The second target session is created by the server of the third-party consulting tool in response to receiving the second target session request, the initiator of the second target session is the user, and the receiver of the second target session is the manual customer service of the publisher.

[0160] In an optional implementation, in response to receiving the second target session request, the server of the third-party consulting tool assigns a human customer service representative to the user from one or more candidate human customer service representatives of the publisher according to the identifier of the publisher in the second target session request, and creates a second target session based on the assigned human customer service representative. The initiator of the second target session is the user, and the recipient of the second target session is the assigned human customer service representative.

[0161] The service end of the third-party consulting tool can allocate a human customer service to the user from one or more candidate human customer service of the publisher according to the preset human customer service allocation strategy. Exemplarily, a human customer service whose current number of user receptions is less than a preset reception number threshold is selected from one or more candidate human customer service of the publisher, and the human customer service whose current number of user receptions is less than the preset reception number threshold is allocated to the user. The current number of user receptions of the human customer service refers to the number of users who need customer service services provided by the human customer service.

[0162] In some embodiments, the process of creating a second target session by the server of a third-party consulting tool may include: generating a session identifier of the second target session, and for the second target session, constructing an association relationship between the session identifier of the second target session, the customer service identifier of the publisher's manual customer service, and the user identifier of the user using the user client 10.

[0163] S33: According to the session information of the second target session, the recipient of the first target session is replaced by the manual customer service.

[0164] The session information of the second target session may include, but is not limited to: a session identifier of the second target session, a customer service identifier of the publisher's manual customer service, and a user identifier of the user.

[0165] Optionally, the consultation page server replaces the receiver of the first target session with a manual customer service, that is, replaces the customer service identifier of the publisher's intelligent customer service in the association relationship created for the first target session with the customer service identifier of the manual customer service.

[0166] S34: Establish a mapping relationship between the first target session and the second target session (ie, the first mapping relationship mentioned above).

[0167] The mapping relationship is used to use the message sent by the manual customer service in the second target session as the message sent by the manual customer service in the first target session and to use the message sent by the user in the first target session as the message sent by the user in the second target session.

[0168] Exemplarily, a mapping relationship between the session identifier of the first target session and the session identifier of the second target session may be established.

[0169] Among them, the mapping relationship is used when the consultation page server 11 performs message forwarding. For example, when the consultation page server 11 receives a message sent by a user in the first target session from the user client 10, according to the mapping relationship, the message is forwarded to the server 13 of the third-party consulting tool as a message sent by the user in the second target session. In this way, the server 13 of the third-party consulting tool sends the message to the customer service client of the publisher's manual customer service, and the customer service client displays the message in the session box of the second target session. For another example, when the consultation page server 11 receives a message sent by the manual customer service in the second target session from the server 13 of the third-party consulting tool, according to the mapping relationship, the message is forwarded to the user client 10 as a message sent by the manual customer service in the first target session. In this way, the user client 10 displays the message in the session box of the first target session.

[0170] In some embodiments, when the consultation page server sends a second target session request to the server of the third-party consulting tool in response to the intelligent customer service being unable to answer the user's question (i.e., the first question mentioned above), the second target session request carries the question.

[0171] The consultation page server is also used to perform the following steps:

[0172] S41. Receive a reply message sent by the server of the third consulting tool.

[0173] The reply message is a message sent by the manual customer service in the second target session to reply to the question.

[0174] After the second target session is successfully created, the third consulting tool sends the question as a message sent by the user in the second target session to the customer service client of the manual customer service, so that the customer service client displays the question in the session box of the second target session. The customer service client sends the message sent by the manual customer service in the second target session to reply to the question to the server of the third consulting tool, and the server of the third consulting tool sends a reply message to the consulting page server.

[0175] S42: According to the mapping relationship, send the reply message to the user client as the message sent by the manual customer service in the first target session.

[0176] Exemplarily, based on the session identifier for identifying the second target session carried in the reply message and the mapping relationship, the reply message is sent to the user client as a message sent by the manual customer service in the first target session, and the user client displays the reply message in the session box of the first target session.

[0177] In this embodiment, the user does not need to ask the manual customer service questions that the intelligent customer service cannot answer again after switching from the intelligent customer service to the manual customer service, thereby achieving seamless switching between the intelligent customer service and the manual customer service, which can improve the user experience.

[0178] After the intelligent customer service is transferred to the manual customer service, in order to improve the service quality of the manual customer service, the conversation record between the user and the intelligent customer service in the first target conversation can also be provided to the manual customer service for reference by the manual customer service. In an optional embodiment, the second target conversation request carries the conversation record of the first target conversation, and the conversation record is used to be displayed in the target page of the customer service client 14 of the manual customer service, and the target page includes the conversation box of the second target conversation. After the second target conversation is successfully created, the service end of the third-party consulting tool can send the conversation record of the first target conversation to the customer service client of the manual customer service, and the customer service client displays the conversation record of the first target conversation in the target page, and the target page includes the conversation box of the second target conversation. Exemplarily, the customer service client displays the conversation record of the first target conversation in the conversation box of the second target conversation.

[0179] In order to facilitate the information statistics of third-party consulting tools, the above-mentioned consulting page server can also be used to perform the following steps:

[0180] S51. Send the session records of historical sessions whose recipient is the intelligent customer service of the publisher to the server of the third-party consulting tool.

[0181] like Figure 1 As shown, the consultation page server 11 sends the session records of historical conversations whose recipient is the publisher's intelligent customer service to the server 13 of the third-party consultation tool (i.e., step 1: message feedback). In this way, the server of the third-party consultation tool not only has the session records of historical conversations whose recipient is the publisher's manual customer service, but also has the session records of historical conversations whose recipient is the publisher's intelligent customer service. Based on this, the server 13 of the third-party consultation tool can determine the clue conversion rate of the media information. Subsequently, the server 13 of the third-party consultation tool sends the media information clue conversion rate to the media information delivery server 15 (i.e., step 2: conversion feedback), so that the media information delivery server 15 optimizes the media information delivery strategy (i.e., step 3: optimized delivery), thereby improving the media information delivery effect, and the media information delivery server 15 delivers media information to the user client 10 (i.e., step 4: delivery). The technical solution provided in the embodiment of the present application forms a data link closed loop of media information delivery, media information clue conversion, and conversion feedback.

[0182] Optionally, the consultation page server can also send the clue information obtained by the intelligent customer service through historical conversations to the server of the third-party consultation tool, so that the server of the third-party consultation tool does not need to search for clue information from the conversation records.

[0183] As an option, the publisher's intelligent agent may send the session records of historical sessions in which the recipient is the publisher's intelligent customer service to the server of the third-party consulting tool, and / or the publisher's intelligent agent may send the clue information obtained by the intelligent customer service through historical sessions to the server of the third-party consulting tool.

[0184] In some embodiments, the above-mentioned target consultation page may also include the publisher's corporate QR code, which is provided by a social application. Users can use the same social application to scan the corporate QR code to establish a connection between the user and the enterprise. Afterwards, the user can have a conversation with the publisher's intelligent customer service or manual customer service through the social application.

[0185] It should be noted that the above-mentioned online consultation system may include a plurality of different third-party consultation tool server ends to meet the needs of different publishers to use different third-party consultation tools.

[0186] It should be noted that the server involved in the embodiments of the present application may be a real server or a virtual server, etc. The server may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, and big data and artificial intelligence platforms.

[0187] The following uses advertisements as an example to describe an application scenario in which an intelligent customer first provides services in the same session, and then a human customer service replaces the intelligent customer service to provide services:

[0188] like Figure 1As shown, the advertisement posted by the advertiser is displayed on the interface of the user client 10. In response to the user's click operation on the advertisement, the user client 10 sends a first session request to the consultation page server 11. The first session request carries the user ID of the user and the ID of the advertiser. After receiving the first session request, the consultation page server 11 extracts the intelligent customer service type by sampling the manual customer service type and the intelligent customer service type with unequal probabilities. The consultation page server 11 creates a first session based on the intelligent customer service type and the advertiser's ID. The initiator of the first session is the user and the recipient is the advertiser's intelligent customer service. Then, based on the first session, response data (including the session ID of the first session) is generated and the response data is sent to the user client 10. The user client 10 displays the first consultation page based on the response data. The first consultation page displays the session box of the first session. The user can enter the first question in the conversation box of the first conversation, and the user client 10 sends the first question and the conversation identifier of the first conversation to the consultation page server 11, and the consultation page server 11 sends the first question to the advertiser's intelligent agent running on the intelligent customer service server 12, and the intelligent agent generates a prompt word according to the first question, inputs the prompt word into the LLM model, and obtains a first answer to the first question. The intelligent agent returns the first answer to the consultation page server 11, and the consultation page server 11 sends the first answer as a message sent by the advertiser's intelligent customer service in the first conversation to the user client 10, and the user client 10 displays the message in the first conversation. The user continues to enter the second question in the conversation box of the first conversation, and the user client 10 sends the first question and the conversation identifier of the first conversation to the consultation page server 11. The consultation page server 11 sends the first question to the advertiser's intelligent agent running on the intelligent customer service server 12. The intelligent agent generates a prompt word according to the first question and inputs the prompt word into the LLM model. The LLM model outputs that it cannot answer the second question. The intelligent agent sends a notification to the consultation page server 11 that it cannot answer the second question. The consultation page server 11 sends a second conversation request to the server 13 of the third-party consultation tool. The second conversation request carries the user's user identifier, the advertiser's identifier, and the historical conversation record in the first conversation, and the historical conversation record includes the second question. The server 13 of the third-party consulting tool selects a target human customer service from the advertiser's alternative human customer service based on the advertiser's identifier and creates a second session. The initiator of the second session is the user and the recipient is the target human customer service. The session notification and historical session records of the second session are sent to the customer service client of the target human customer service, so that the customer service client can display the session interface of the second session based on the session notification. The session interface includes a session box of the second session, and the above-mentioned historical session records are displayed in the session box.The customer service client enters a second reply to the second question in the conversation box, and the customer service client sends the second reply to the server 13 of the third-party consulting tool. The server 13 of the third-party consulting tool sends the session identifier of the second session, the customer service identifier of the target manual customer service, and the second reply to the consultation page server 11. The consultation page server 11 establishes a mapping relationship between the session identifier of the first session and the session identifier of the second session, modifies the recipient of the first session to the target manual customer service, sends the second reply to the user client 10, and notifies the user client 10 that the recipient of the first session is modified to the target manual customer service. The user client 10 may display the prompt message "Manual customer service will now provide services to you instead of intelligent customer service" and the second reply in the conversation box of the first session. Next, the user can continue to enter the third question in the conversation box, and the user client sends the third question and the session identifier of the first conversation to the consultation page server 11. The consultation page server 11 will determine the session identifier of the second conversation based on the session identifier of the first conversation and the above-mentioned mapping relationship, and send the third question and the session identifier of the second conversation to the server 13 of the third-party consulting tool. The server 13 of the third-party consulting tool, based on the session identifier of the second conversation, sends the third question to the recipient of the second conversation, that is, the customer service client 14 of the target manual customer service. The customer service client 14 of the target manual customer service displays the third question in the conversation box of the second conversation, and the target manual customer service enters the third answer to the third question in the conversation box. The customer service client 14 sends the third answer and the session identifier of the second conversation to the server 13 of the third-party consulting tool. The server 13 of the third-party consulting tool determines the session identifier of the first conversation based on the session identifier of the second conversation and the above-mentioned mapping relationship, thereby sending the third answer to the initiator of the first conversation, that is, the user client 10.

[0189] Figure 4 1 is a flow chart of an online consultation method provided in an embodiment of the present application. The method can be applied to the above-mentioned online consultation system. Figure 4 As shown, the method includes the following steps:

[0190] 401. The user client displays the media information published by the publisher.

[0191] 402. The user client sends a first target session request to the consultation page server in response to a trigger operation on the media information.

[0192] 403. The consultation page server responds to the first target session request and selects a target customer service type from manual customer service types and intelligent customer service types according to a preset selection strategy.

[0193] Among them, in the preset selection strategy, both the manual customer service type and the intelligent customer service type have the probability of being selected.

[0194] 404. The consultation page server creates a first target session according to the target customer service type.

[0195] The receiver of the first target conversation is the target customer service of the publisher, and the target customer service belongs to the target customer service type.

[0196] 405. The consultation page server sends response data to the user client for the first target session request.

[0197] 406. In response to receiving the response data, the user client displays a target consultation page.

[0198] Wherein, the target consultation page includes a conversation box of the first target conversation.

[0199] The specific implementation of the above steps 401-406 can refer to the corresponding contents in the above embodiments, which will not be repeated here.

[0200] In some embodiments, the target customer service type is an intelligent customer service type, and the target customer service is an intelligent customer service;

[0201] After the first target session is successfully created, the method further includes:

[0202] 407. The consultation page server sends a notification message to the publisher's agent.

[0203] The notification message is used to notify that the first target session has been created, and the agent runs on the intelligent customer service server.

[0204] The specific implementation of the above step 407 can refer to the corresponding content in the above embodiments, which will not be repeated here.

[0205] In some embodiments, the above method may further include:

[0206] 408. The consultation page server receives the second question sent by the user client and asked by the user in the first target session.

[0207] 409. The consultation page server sends the second question to the publisher's agent.

[0208] 410. The consultation page server receives a reply message returned by the agent in response to the second question.

[0209] 411. The consultation page server sends the reply message to the user client as a message sent by the intelligent customer service in the first target session.

[0210] 412. The user client displays the reply message in the conversation box of the first target conversation.

[0211] The specific implementation of the above steps 408-412 can refer to the corresponding content in the above embodiments, which will not be repeated here.

[0212] In some embodiments, the above method may further include:

[0213] 413. In response to receiving the second question, the agent requests the large language model to output an answer message to the question based on the question and the portrait data of the publisher.

[0214] Exemplarily, the portrait data is extracted from historical reply messages of the publisher's manual customer service using the large language model.

[0215] 414. The intelligent agent sends the reply message to the consultation page server.

[0216] The specific implementation of the above steps 413-414 can refer to the corresponding content in the above embodiments, which will not be repeated here.

[0217] It should be noted here that: for the contents of each step not fully described in detail in the method provided in the embodiment of the present application, please refer to the corresponding contents in the above embodiment, which will not be repeated here. In addition, in addition to the above steps, the method provided in the embodiment of the present application may also include other parts or all of the steps in the above embodiments, which can be specifically referred to the corresponding contents in the above embodiments, which will not be repeated here.

[0218] Figure 5 Schematic diagram of the process of an online consultation method provided in an embodiment of the present application. The method can be applied to the user client in the above-mentioned online consultation system. Figure 5 As shown, the method includes the following steps:

[0219] 501. At a first moment, in response to a triggering operation on media information released by a publisher, a first consultation page is displayed.

[0220] The first consultation page includes a conversation box of a first conversation, and the recipient of the first conversation is the intelligent customer service of the publisher.

[0221] In some embodiments, at a first moment, in response to a triggering operation by a user on media information displayed in a web page provided by a media platform, a first consultation page is displayed.

[0222] like Figure 2As shown, at the first moment, in response to the user's triggering operation on the media information 202 displayed in the web page 200 provided by the media platform, a first consultation page 203 is displayed, and the first consultation page 203 includes a conversation box 204 of the first conversation, and the recipient of the first conversation is the intelligent customer service of the publisher.

[0223] 502. At a second moment, in response to a triggering operation on the media information, a second consulting page is displayed.

[0224] The second consultation page includes a conversation box for a second conversation, and the recipient of the second conversation is a manual customer service representative of the publisher;

[0225] The first session and the second session are two different sessions.

[0226] In some embodiments, at the second moment, in response to a triggering operation by a user on media information displayed in a web page provided by the media platform, a second consultation page is displayed.

[0227] like Figure 6 As shown, at the second moment, in response to the user's triggering operation on the media information 602 displayed in the web page 601 provided by the above-mentioned media platform, a second consultation page 603 is displayed, and the second consultation page 603 includes a conversation box 604 for the second conversation, and the recipient of the second conversation is the publisher's manual customer service.

[0228] In practical applications, Figure 2 Page 200 and Figure 6 The web page 601 in may be provided by the same media platform.

[0229] In the technical solution provided by the embodiment of the present application, when a user enters the consultation page of the media information at different times, he is not always received by the publisher's intelligent customer service, but sometimes by the publisher's manual customer service, and sometimes by the publisher's intelligent customer service. In other words, in this solution, the intelligent customer service does not completely replace the manual customer service to talk to the user, but the intelligent customer service partially replaces the manual customer service to talk to the user. In this way, the technical solution provided by the embodiment of the present application can not only use intelligent customer service to reduce the workload of manual customer service and improve the customer service response speed, but also use manual customer service to provide professional manual customer service services to some users to improve the quality of customer service. In this way, the user's media information access experience can be effectively improved, thereby improving the lead conversion rate of media information, and then improving the promotion effect of media information.

[0230] In some embodiments, in order to facilitate session management, the consultation page server in the online consultation system will maintain the session state of the session. After creating a session, the consultation page server will set the session state of the session to an open state; when the session does not generate a new session message within a preset time (for example, half an hour), the consultation page server can set the session state to a closed state. In response to the user's trigger operation on the media information, the user client sends a session request to the consultation page server. The consultation page server searches for a target session in an open state based on the user's user ID and the publisher's ID carried in the session request, wherein the initiator of the target session is the user and the receiver is the manual customer service or intelligent customer service of the publisher; if such a target session exists, then based on the target session, the response data for the session request is sent to the user client, so that the user client displays the consultation page, and the consultation page includes the session box of the target session; if such a target session does not exist, a new session is created, and the response data for the session request is sent to the user client, so that the user client displays the consultation page, and the consultation page includes the session box of the newly created session.

[0231] In some embodiments, the second time is later than the first time, and the state of the first session at the second time is a closed state. That is, relative to the first session, the second session is a newly created session at the second time.

[0232] In some other embodiments, the first time is later than the second time, and the state of the second session at the first time is closed. That is, relative to the second session, the first session is a newly created session at the first time.

[0233] In some embodiments, the first moment is within a first preset time period, the second moment is within a second preset time period, and there is no intersection between the first preset time period and the second preset time period. Figure 1 The corresponding contents in the illustrated embodiment will not be repeated here.

[0234] In some embodiments, at the third moment, the recipient of the first conversation is changed to the manual customer service of the publisher. The third moment is later than the first moment. For the specific change process, please refer to Figure 1 The corresponding contents in the illustrated embodiment will not be repeated here.

[0235] Figure 7 1 is a flow chart of an online consultation method provided in an embodiment of the present application. The method can be applied to the consultation page server in the above-mentioned online consultation system. Figure 7 As shown, the method includes the following steps:

[0236] 701. In response to receiving a first target session request sent by a user client, select a target customer service type from a manual customer service type and an intelligent customer service type according to a preset selection strategy.

[0237] Among them, the first target session request is sent by the user client in response to a trigger operation of the user on the media information published by the publisher; in the preset selection strategy, the manual customer service type and the intelligent customer service type both have a probability of being selected.

[0238] 702. Create a first target session according to the target customer service type.

[0239] The receiver of the first target conversation is the target customer service of the publisher, and the target customer service belongs to the target customer service type;

[0240] 703. Send response data for the first target session request to the user client so that the user client displays a target consultation page.

[0241] Wherein, the target consultation page includes a conversation box of the first target conversation.

[0242] The specific implementation of the above steps 701-703 can be found in the above Figure 1 The corresponding contents in the illustrated embodiment will not be repeated here.

[0243] In some embodiments, in the above 701, “in response to receiving the first target session request sent by the user client, selecting a target customer service type from the manual customer service type and the intelligent customer service type according to a preset selection strategy” can be implemented by the following steps:

[0244] 7011a. When the request time of the first conversation request is within a first preset time period, determining the intelligent customer service type as the target customer service type;

[0245] 7012a. When the request time of the first conversation request is within a second preset time period, determining the manual customer service type as the target customer service type;

[0246] There is no intersection between the first preset time period and the second preset time period.

[0247] The specific implementation of the above steps 7011a-7012a can be found in the above Figure 1 The corresponding contents in the illustrated embodiment will not be repeated here.

[0248] In other embodiments, in 701 above, “in response to receiving the first target session request sent by the user client, selecting a target customer service type from the manual customer service type and the intelligent customer service type according to a preset selection strategy” can be implemented by the following steps:

[0249] 7011b. Taking the probability of selecting the intelligent customer service type as the first probability and the probability of selecting the manual customer service type as the second probability, extract the target customer service type from the manual customer service type and the intelligent customer service type, wherein the first probability is different from the second probability.

[0250] The specific implementation of the above step 7011b can be found in the above Figure 1 The corresponding contents in the illustrated embodiment will not be repeated here.

[0251] In some embodiments, the target customer service type is an intelligent customer service type, and the target customer service is an intelligent customer service; after the first target session is created, the method may further include the following steps:

[0252] 704. For the first target session, send a second target session request to a server of a third-party consulting tool.

[0253] 705: receiving session information of the second target session returned by the server of the third-party consulting tool,

[0254] The second target session is created by the server of the third-party consulting tool in response to receiving the second target session request, the initiator of the second target session is the user, and the receiver of the second target session is the manual customer service of the publisher.

[0255] 706. According to the session information of the second target session, replace the recipient of the first target session with the manual customer service.

[0256] 707. Establish a first mapping relationship between the first target session and the second target session.

[0257] The first mapping relationship is used to use the message sent by the manual customer service in the second target session as the message sent by the manual customer service in the first target session and to use the message sent by the user in the first target session as the message sent by the user in the second target session.

[0258] The specific implementation of the above steps 704-707 can be found in the above Figure 1 The corresponding contents in the illustrated embodiment will not be repeated here.

[0259] In some embodiments, the above-mentioned step 704 of “sending a second target session request to a server of a third-party consulting tool for the first target session” may include the following steps:

[0260] 7041. In response to the intelligent customer service being unable to answer the user's first question, sending a second target session request to a server of a third-party consulting tool for the first target session.

[0261] In some embodiments, the second target session request carries the first question. The method may further include the following steps:

[0262] 708. Receive a reply message sent by the server of the third consulting tool.

[0263] The reply message is a message sent by the manual customer service in the second target session to reply to the first question.

[0264] 709. According to the mapping relationship, send the reply message to the user client as the message sent by the manual customer service in the first target session.

[0265] In some embodiments, the second target session request carries a session record of the first target session, and the session record is used to be displayed in a session box of the second target session displayed by the customer service client of the manual customer service.

[0266] In some embodiments, the above method may further include the following steps:

[0267] 710. Send the session records of historical sessions in which the recipient is the intelligent customer service of the publisher to the server of the third-party consulting tool.

[0268] In some embodiments, the target customer service type is a manual customer service type, and the target customer service is a manual customer service.

[0269] In the above 702, “creating a first target session according to the target customer service type” may include the following steps:

[0270] 7021. Send a third target session request to the server of the third-party consulting tool.

[0271] 7022. Receive session information of the third target session returned by the server of the third-party consulting tool.

[0272] The third target session is created by the server of the third-party consulting tool in response to receiving the third target session request, the initiator of the third target session is the user, and the receiver of the third target session is the manual customer service of the publisher.

[0273] 7023. Create a first target session according to the session information of the third target session.

[0274] The initiator of the first target session is the user, and the receiver of the first target session is the manual customer service.

[0275] 7024. Establish a second mapping relationship between the first target session and the third target session.

[0276] The second mapping relationship is used to use the message sent by the manual customer service in the third target session as the message sent by the manual customer service in the first target session and to use the message sent by the user in the first target session as the message sent by the user in the third target session.

[0277] The specific implementation of the above steps 7021-7024 can be found in the above Figure 1 The corresponding contents in the illustrated embodiment will not be repeated here.

[0278] It should be noted here that: for the contents of each step not fully described in detail in the method provided in the embodiment of the present application, please refer to the corresponding contents in the above embodiment, which will not be repeated here. In addition, in addition to the above steps, the method provided in the embodiment of the present application may also include other parts or all of the steps in the above embodiments, which can be specifically referred to the corresponding contents in the above embodiments, which will not be repeated here.

[0279] The following will be combined Figure 8 The system layered architecture involved in the online consultation method provided in the embodiment of the present application includes a model layer, a storage layer, a platform layer and an application layer.

[0280] The model layer provides a large language model, the storage layer provides vector database (database required for vector retrieval), full-text retrieval (ES), session storage, metadata management and other capabilities, the platform layer can provide tenant management, intelligent agent management, tool management, knowledge base management and session management, and the application layer provides intelligent customer service capabilities, image mining capabilities and data insight capabilities. Session management includes global session management and multi-round session management.

[0281] In the development process of intelligent customer service, the open source technology framework - LangChain4j can be used to initially build an intelligent question-and-answer platform using LangChain4j, and explore other application scenarios based on the platform, one of which is data insight; build a platform for specific application scenarios based on the open source technology framework to improve the applicability of functions to meet specific needs. Among them, LangChain4j is the Java version of LangChain, which is currently widely used. Secondary development is carried out on this basis to build the required intelligent question-and-answer platform. Through secondary development, the capabilities of LangChain4j can be enhanced, including global session memory capabilities in distributed environments, and unified question-and-answer interfaces that support multiple prompt templates. Based on the platform, the conversation capabilities of intelligent customer service can be realized, and later, other intelligent applications can also be supported based on this platform.

[0282] Fig. 9 FIG. 1 is a schematic diagram showing the structure of an electronic device provided by an embodiment of the present application. Fig. 9 As shown, the electronic device includes a memory 1101 and a processor 1102. The memory 1101 can be configured to store various other data to support operations on the electronic device. Examples of these data include instructions for any application or method for operating on the electronic device. The memory 1101 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read only memory (EEPROM), an erasable programmable read only memory (EPROM), a programmable read only memory (PROM), a read only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0283] The memory 1101 is used to store programs;

[0284] The processor 1102 is coupled to the memory 1101 and is used to execute the program stored in the memory 1101 to implement the above Figure 5 The illustrated embodiments provide methods.

[0285] Further, if Fig. 9As shown, the electronic device also includes: a communication component 1103, a display 1104, a power component 1105, an audio component 1106 and other components. Fig. 9 Only some components are shown schematically, which does not mean that the electronic device only includes Fig. 9 Components shown.

[0286] Accordingly, the embodiment of the present application further provides a computer-readable storage medium storing a computer program, which can achieve the above-mentioned Figure 5 The illustrated embodiments provide method steps or functions.

[0287] The present application also provides a computer program product, including a computer program, which can achieve the above-mentioned Figure 5 The illustrated embodiments provide method steps or functions.

[0288] Fig.10 FIG. 1 is a schematic diagram showing the structure of an electronic device provided by an embodiment of the present application. Fig.10 As shown, the electronic device includes a memory 1101 and a processor 1102. The memory 1101 can be configured to store various other data to support operations on the electronic device. Examples of these data include instructions for any application or method for operating on the electronic device. The memory 1101 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as a static random access memory (SRAM), an electrically erasable programmable read only memory (EEPROM), an erasable programmable read only memory (EPROM), a programmable read only memory (PROM), a read only memory (ROM), a magnetic memory, a flash memory, a magnetic disk or an optical disk.

[0289] The memory 1101 is used to store programs;

[0290] The processor 1102 is coupled to the memory 1101 and is used to execute the program stored in the memory 1101 to implement the above Figure 7 The illustrated embodiments provide methods.

[0291] Further, if Fig.10As shown, the electronic device also includes: a communication component 1103, a display 1104, a power component 1105, an audio component 1106 and other components. Figure 5 Only some components are shown schematically, which does not mean that the electronic device only includes Fig.10 Components shown.

[0292] Accordingly, the embodiment of the present application further provides a computer-readable storage medium storing a computer program, which can achieve the above-mentioned Figure 7 The illustrated embodiments provide method steps or functions.

[0293] The present application also provides a computer program product, including a computer program, which can achieve the above-mentioned Figure 7 The illustrated embodiments provide method steps or functions.

[0294] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0295] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM (Read Only Memory) / RAM (Random Access Memory), a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0296] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it, and multiple embodiments can be combined with each other; although the present application has been described in detail with reference to the aforementioned embodiments, ordinary technicians in this field should understand that they can still modify the technical solutions recorded in the aforementioned embodiments, or replace some of the technical features therein with equivalents; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. An online consultation method, applied to a consultation page server, characterized in that: include: In response to receiving a first target session request sent by a user client, selecting a target customer service type from a manual customer service type and an intelligent customer service type according to a preset selection strategy, wherein the first target session request is sent by the user client in response to a trigger operation of a user on media information published by a publisher; in the preset selection strategy, both the manual customer service type and the intelligent customer service type have a probability of being selected; According to the target customer service type, a first target session is created, where the recipient of the first target session is the target customer service of the publisher, and the target customer service belongs to the target customer service type; Response data for the first target session request is sent to the user client so that the user client displays a target consultation page, wherein the target consultation page includes a session box of the first target session.

2. The method according to claim 1, characterized in that In response to receiving the first target session request sent by the user client, selecting a target customer service type from a manual customer service type and an intelligent customer service type according to a preset selection strategy, including: When the request time of the first conversation request is within a first preset time period, determining the intelligent customer service type as the target customer service type; When the request time of the first conversation request is within a second preset time period, determining the manual customer service type as the target customer service type; There is no intersection between the first preset time period and the second preset time period.

3. The method according to claim 1, characterized in that The step of selecting a target customer service type from a manual customer service type and an intelligent customer service type according to a preset selection strategy in response to receiving a first target session request sent by a user client includes: The probability of selecting the intelligent customer service type is taken as the first probability, and the probability of selecting the manual customer service type is taken as the second probability, and the target customer service type is extracted from the manual customer service type and the intelligent customer service type.

4. The method according to any one of claims 1 to 3, characterized in that The target customer service type is an intelligent customer service type, and the target customer service is an intelligent customer service; after the first target session is created, the method further includes: For the first target session, sending a second target session request to a server of a third-party consulting tool; Receiving session information of a second target session returned by the server of the third-party consulting tool, where the second target session is created by the server of the third-party consulting tool in response to receiving the second target session request, the initiator of the second target session is the user, and the receiver of the second target session is the manual customer service of the publisher; According to the session information of the second target session, replacing the recipient of the first target session with the manual customer service; Establishing a first mapping relationship between the first target session and the second target session; The first mapping relationship is used to use the message sent by the manual customer service in the second target session as the message sent by the manual customer service in the first target session and to use the message sent by the user in the first target session as the message sent by the user in the second target session.

5. The method according to claim 4, characterized in that For the first target session, sending a second target session request to a server of a third-party consulting tool includes: In response to the intelligent customer service being unable to answer the first question of the user, a second target session request is sent to a service end of a third-party consulting tool for the first target session.

6. The method according to claim 5, characterized in that The second target session request carries the first question; The method further comprises: receiving a reply message sent by the server of the third consulting tool, wherein the reply message is a message sent by the manual customer service in the second target session to reply to the first question; According to the mapping relationship, the reply message is sent to the user client as the message sent by the manual customer service in the first target session.

7. The method according to claim 5, characterized in that The second target session request carries a session record of the first target session, and the session record is used to be displayed in a session box of the second target session displayed by the customer service client of the manual customer service.

8. The method according to any one of claims 1 to 3, characterized in that Also includes: The session records of historical sessions whose recipient is the publisher's intelligent customer service are sent to a server of a third-party consulting tool.

9. The method according to any one of claims 1 to 3, characterized in that The target customer service type is a manual customer service type, and the target customer service is a manual customer service; The step of creating a first target session according to the target customer service type includes: Sending a third target session request to a server of a third-party consulting tool; Receiving session information of a third target session returned by the server of the third-party consulting tool, wherein the third target session is created by the server of the third-party consulting tool in response to receiving the third target session request, the initiator of the third target session is the user, and the receiver of the third target session is the manual customer service of the publisher; Creating a first target session according to the session information of the third target session, where the initiator of the first target session is the user and the receiver of the first target session is the manual customer service; Establishing a second mapping relationship between the first target session and the third target session; The second mapping relationship is used to use the message sent by the manual customer service in the third target session as the message sent by the manual customer service in the first target session and to use the message sent by the user in the first target session as the message sent by the user in the third target session.

10. The method according to any one of claims 1 to 3, characterized in that The publisher is an advertiser, and the media information is an advertisement.

11. An online consultation method, applied to a user client, characterized in that: include: At a first moment, in response to a triggering operation on the media information published by the publisher, a first consultation page is displayed, wherein the first consultation page includes a conversation box of a first conversation, and the recipient of the first conversation is an intelligent customer service of the publisher; At the second moment, in response to a trigger operation on the media information, a second consultation page is displayed, the second consultation page includes a conversation box for a second conversation, and the recipient of the second conversation is a manual customer service of the publisher; The first session and the second session are two different sessions.

12. The method according to claim 11, characterized in that The second time is later than the first time, and the state of the first session at the second time is closed.

13. The method according to claim 11, characterized in that The first time is later than the second time, and the state of the second session at the first time is closed.

14. The method according to claim 11, characterized in that The first moment is within a first preset time period, the second moment is within a second preset time period, and there is no intersection between the first preset time period and the second preset time period.

15. The method according to any one of claims 11 to 14, characterized in that At the third moment, the recipient of the first conversation is changed to the manual customer service of the publisher; The third moment is later than the first moment.

16. An online consultation method, characterized in that: include: The user client displays the media information published by the publisher; The user client sends a first target session request to the consultation page server in response to the user's triggering operation on the media information; The consultation page server responds to the first target session request and selects a target customer service type from the manual customer service type and the intelligent customer service type according to a preset selection strategy; In the preset selection strategy, both the manual customer service type and the intelligent customer service type have a probability of being selected; The consultation page server creates a first target session according to the target customer service type, wherein the recipient of the first target session is the target customer service of the publisher, and the target customer service belongs to the target customer service type; The consultation page server sends response data to the user client for the first target session request; In response to receiving the response data, the user client displays a target consultation page including a session box of the first target session.

17. The method according to claim 16, characterized in that The target customer service type is an intelligent customer service type, and the target customer service is an intelligent customer service; After the first target session is successfully created, the method further includes: The consultation page server sends a notification message to the publisher's agent, where the notification message is used to notify that the first target session has been created, wherein the agent runs on the intelligent customer service server.

18. The method according to claim 17, characterized in that Also includes: The consultation page server receives a second question sent by the user client and issued by the user in the first target session; The consultation page server sends the second question to the agent of the publisher; In response to receiving the second question, the agent requests the large language model to output an answer message to the question according to the second question and the portrait data of the publisher; The consultation page server receives the reply message sent by the agent; The consultation page server sends the reply message to the user client as a message sent by the intelligent customer service in the first target session; The user client displays the reply message in the conversation box of the first target conversation.

19. The method according to claim 18, characterized in that The portrait data is extracted from conversation messages of historical conversations in which the publisher's manual customer service participated using the large language model.

20. An online consultation system, characterized in that: include: User client and consultation page server; among them, The user client is used to: display the media information published by the publisher; in response to a trigger operation on the media information, send a first target session request to the consultation page server; The consultation page server is configured to: in response to receiving the first target session request, select a target customer service type from the manual customer service type and the intelligent customer service type according to a preset selection strategy; in the preset selection strategy, both the manual customer service type and the intelligent customer service type have a probability of being selected; create a first target session according to the target customer service type, the recipient of the first target session is the target customer service of the publisher, and the target customer service belongs to the target customer service type; send response data to the user client for the first target session request; The user client is further configured to display a target consultation page in response to receiving the response data, wherein the target consultation page includes a session box of the first target session.

21. An electronic device, characterized in that: include: A memory and a processor, wherein The memory is used to store programs; The processor, coupled to the memory, is configured to execute the program stored in the memory to implement the method according to any one of claims 1 to 19.

22. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a computer, the method according to any one of claims 1 to 19 can be implemented.

23. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the method according to any one of claims 1 to 19 is implemented.

Images