Information sending method, device and equipment

Abstract

The invention provides an information sending method, device and equipment. The scheme comprises the steps of obtaining problem triggering information which is sent by a user terminal and relates to a target service field; afterwards, if a target question corresponding to the question triggering information exists in a preset question list, preset reply information of the target question can be sent to the user terminal, and the preset reply information is sent to the user terminal. The information is information obtained and stored by calling a large language model by an agent after the target question is input into the agent for the target business field in advance; and if the target question corresponding to the question triggering information does not exist in the preset question list, inputting the question triggering information into the intelligent agent aiming at the target business field, calling a large language model by the intelligent agent to generate reply information, and sending the reply information to the user terminal.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and more particularly to an information sending method, an information sending device and a computing device. Background Art

[0002] Agents can complete tasks through autonomous planning. The processing of tasks can be simply described as perception, planning and action. Perception refers to the information obtained by the agent from the environment; planning refers to the decision-making process made by the agent to complete the task. For example, the task can be split to autonomously complete the planning of the process; action refers to the action taken based on the environment and planning. The agent technology driven by large models is one of the important research contents in the field of large models and artificial intelligence. It helps to realize the wide application of large models in various industries. Specifically, the large model can be used as the core control module of the agent, and the task can be planned based on the ability of the large model, and the agent can be controlled to realize the processing of the task.

[0003] Currently, when large model-driven intelligent agents process tasks, the efficiency of task processing is low because it may involve complex reasoning or multiple calls to large models. It is difficult to meet users' requirements for timeliness in responding to questions, and the user interaction experience is poor. Summary of the invention

[0004] In view of this, the embodiments of the present application provide an information sending method, device and equipment to solve the problem that it is difficult to meet the user's requirements for the timeliness of question replies when the existing user application intelligent agent obtains question reply information, resulting in poor user interaction experience.

[0005] According to a first aspect of an embodiment of the present application, there is provided a method for sending information, including: Obtaining problem triggering information related to the target business field sent by the user terminal; If there is a target question corresponding to the question trigger information in the preset question list, the preset reply information of the target question is sent to the user terminal; wherein the preset reply information is information obtained and stored by the intelligent agent calling the large language model after the target question is input into the intelligent agent for the target business field in advance; If the target question corresponding to the question trigger information does not exist in the preset question list, the question trigger information is input into the agent for the target business field, and the agent calls the large language model to generate reply information, and sends the reply information to the user terminal.

[0006] According to a second aspect of an embodiment of the present application, there is provided an information sending device, including: An information acquisition module, used to acquire problem triggering information related to a target business field sent by a user terminal; A first information sending module is configured to send preset reply information of the target question to the user terminal if there is a target question corresponding to the question trigger information in the preset question list; wherein the preset reply information is information obtained and stored by the agent after the target question is input into the agent for the target business field in advance and then the agent calls the large language model; The second information sending module is used to input the question trigger information into the agent for the target business field, and then the agent calls the large language model to generate reply information if there is no target question corresponding to the question trigger information in the preset question list, and send the reply information to the user terminal.

[0007] According to a third aspect of an embodiment of the present application, there is provided an information sending device, including: at least one processor; and, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Obtaining problem triggering information related to the target business field sent by the user terminal; If there is a target question corresponding to the question trigger information in the preset question list, the preset reply information of the target question is sent to the user terminal; wherein the preset reply information is information obtained and stored by the intelligent agent calling the large language model after the target question is input into the intelligent agent for the target business field in advance; If the target question corresponding to the question trigger information does not exist in the preset question list, the question trigger information is input into the agent for the target business field, and the agent calls the large language model to generate reply information, and sends the reply information to the user terminal.

[0008] One embodiment of the present specification can at least achieve the following beneficial effects: by pre-setting preset questions in a preset question list, inputting the preset questions into an intelligent agent for the target business field, and then the intelligent agent calls the large language model to obtain preset reply information, and the preset questions are associated with the preset reply information and stored; when a question trigger information involving the target business field sent by a user terminal is received, if there is a target question corresponding to the question trigger information in the preset question list, the preset reply information corresponding to the target question can be returned to the user terminal, without calling the intelligent agent to generate reply information using the large language model after the user asks the question and before replying to the user, thereby improving the efficiency of users obtaining question reply information and enhancing the user interaction experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0010] Figure 1 A schematic diagram of an application scenario of a method for answering user questions based on an intelligent agent provided in an embodiment of this specification; Figure 2 A flowchart of a method for sending information provided in an embodiment of this specification; Figure 3 A schematic diagram of an interactive interface for a user to communicate with an agent in a dialogue page provided in an embodiment of this specification; Figure 4 A swimming lane diagram of a method for sending a question reply message to a user in response to a user's question in an actual application scenario provided by an embodiment of this specification; Figure 5 The embodiments of this specification provide corresponding to Figure 2 A structural schematic diagram of an information sending device; Figure 6 The embodiments of this specification provide corresponding to Figure 2 A structural diagram of an information sending device. DETAILED DESCRIPTION

[0011] Many specific details are described in the following description to facilitate a full understanding of the present application. However, the present application can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of the present application, so the present application is not limited by the specific implementation disclosed below.

[0012] The terms used in one or more embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of the present application. The singular forms of "a", "said" and "the" used in one or more embodiments of the present application and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of the present application refers to and includes any or all possible combinations of one or more associated listed items.

[0013] It should be understood that, although the terms first, second, etc. may be used to describe various information in one or more embodiments of the present application, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of the present application, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0014] 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 relevant laws, regulations and standards in the relevant regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0015] First, the terms involved in one or more embodiments of the present application are explained.

[0016] A large model refers to a deep learning model with large-scale model parameters, usually containing hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than 10 trillion model parameters. A large model can also be called a foundation model. It is pre-trained with large-scale unlabeled corpus to produce a pre-trained model with more than 100 million parameters. This model can adapt to a wide range of downstream tasks and has good generalization capabilities, such as large-scale language models (LLMs) and multi-modal pre-training models.

[0017] When the big model is used in practice, only a small number of samples are needed to fine-tune the pre-trained model and it can be applied to different tasks. The big model can be widely used in natural language processing (NLP), computer vision and other fields. Specifically, it can be applied to computer vision tasks such as visual question answering (VQA), image description (IC, Image Caption), image generation, as well as natural language processing tasks such as text-based sentiment classification, text summary generation, and machine translation. The main application scenarios of the big model include digital assistants, intelligent robots, search, online education, office software, e-commerce, intelligent design, etc.

[0018] 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. LLM has the ability of in-context learning, can learn complex patterns in language, and perform a wide range of tasks, including text summarization, translation, sentiment analysis, multi-round dialogue, etc. Among them, LLM can include ChatGPT model, T5 (Text-to-Text Transfer Transformer) model, PaLM model, LLaMA (Large Language ModelMetaAI) model, Tongyi Qianwen model, Bailing model, etc.

[0019] Agent, also known as agent model, agent model, AI agent model, AI model, artificial intelligence model, etc. Different from traditional machine learning models, agents are highly intelligent models that can perceive the environment, make decisions and perform corresponding tasks. The core of agents is their ability to think independently and call tools, so that they can gradually complete given goals. In the process of using agents, the basic models or related tools connected to the agents can be called according to the pre-defined role and path planning to gradually complete the specified tasks.

[0020] The agent can process tasks through multiple rounds of interaction with the environment. Each round of interaction information may include agent input information, i.e., information obtained by the agent from the environment, and may also include agent output information, i.e., information output by the agent to the environment. The environment may be anything outside the agent, for example, the environment may include user questions input by the user, document systems, API systems, other agents, etc., but is not limited thereto. Exemplarily, in the field of intelligent customer service, the agent may obtain the user's input information, generate reply information corresponding to the input information based on the information processing model, and output it to the user. In the process of interacting with the user, the agent may also implement operations such as API (Application Programming Interface) calls and document queries based on the information processing model, so as to better process the user's needs.

[0021] In the embodiments of this specification, the agent can be connected to the large language model, and can be called an agent driven by the large language model, or an agent controlled by the large language model. During the use of the agent, the agent can create appropriate prompts according to the given goals to stimulate the reasoning ability of the large language model. Among them, the prompts mentioned here are prompts or guides specifically for stimulating the reasoning ability of the large language model, which include a detailed description of the task, the required information type, and the output format. Specifically, after the user question is input into the agent, the user question can be parsed and understood by the agent, such as semantic recognition, determination of the question type, keyword extraction of the user question, etc., and then the knowledge base or API system can be called for information query based on the results of the analysis and understanding of the user question, and the reasoning prompt is generated based on the analysis and understanding results of the user question or based on the analysis and understanding results of the user question and the query result, and then the reasoning prompt is input into the large language model, so that the large language model generates a detailed and coherent reply to the user question based on the reasoning prompt and its powerful text generation ability. The knowledge base may specifically be a domain knowledge base, that is, a database having knowledge of a specific business field, wherein the specific business field may be the medical field, the financial field, the insurance field, the consumer field, and the like.

[0022] Prompt, also known as prompt engineering, is also called prompt word, prompt word template, prompt template, etc. It is a prompt or incentive used to guide user input or stimulate specific program response in the big model. It is a set of carefully designed questioning paradigms used in the design of instructions related to the big model. Through intention understanding and other methods, the user's questions are filled into it, and the instructions for inputting the big model are obtained and then sent to the big model. Well-designed prompts can often stimulate the stronger potential capabilities of the big model, thereby achieving better results than not using prompts.

[0023] In actual applications, in various business fields, such as consumer finance, healthcare, tourism, etc., the demand for detailed operations and business growth is becoming more and more urgent. The traditional decision-making model that relies on expert experience and intuition can no longer meet the needs of rapidly developing businesses. To this end, people use the intelligent agent technology driven by large language models to develop a business think tank based on large models, combining artificial intelligence, big data analysis and business knowledge management, etc., aiming to solve problems such as target user identification, fuzzy user portraits, unclear user intentions, and inaccurate operation space predictions in operations. Through knowledge engines and service engines, such intelligent agent technology can realize the automation of business knowledge reserves, intelligent analysis of data, and dynamic generation of strategies, making business operation decisions more efficient and accurate.

[0024] It is understandable that the agent can think and perform actions on behalf of the user. For example, in the scenario of helping users make travel plans, the agent can serve as the user's travel consultant, understand the user's preferences, perform travel information searches, arrange itineraries, place orders for air tickets / train tickets, hotels, etc. It can be seen that in the actual application process of products based on agent technology driven by large language models, due to the possibility of involving complex reasoning, one or more calls to large models, one or more queries to knowledge bases, one or more calls to multiple external tool systems, etc., the task processing time is long, and sometimes it may take minutes or hours to get a reply information. It is impossible to provide instant feedback, and it is impossible to meet the user's requirements for the timeliness of question responses. The user interaction experience is poor and it is difficult to promote and apply on a large scale.

[0025] In the related technology, in order to solve the problem of long waiting time for users of dialogue products such as intelligent agents driven by large language models, the computing power can be improved, that is, a large-scale GPU cluster can be provided, but this will significantly increase the computing cost and may be difficult to achieve the desired effect; or, the requirements can be broken down and developed for different scenarios, and the steps and processes can be simplified, but the development cycle is long, the development cost is high, and it may be difficult to achieve the desired effect; or, customized development can be carried out for different scenarios through pre-training or re-training to improve the response efficiency in each scenario in a targeted manner, but the development cycle is long, the development cost is high, and it may be difficult to achieve the desired effect.

[0026] In view of this, in the embodiments of this specification, a session acceleration scheme based on a cache system is proposed to improve the response speed to user questions and enhance the user interaction experience. Specifically, for different business fields, a series of preset user questions can be accurately predicted in advance, and these preset user questions can be input into the intelligent agent driven by a large language model in the corresponding business field, so that the intelligent agent can pre-determine the preset response information corresponding to each preset user question and save it in the cache system; afterwards, when the user has a real-time conversation, the user question can be matched with the preset user question list in the cache system, and if it is a hit, the corresponding preset response information is directly returned, without the need to call the intelligent agent in real time for reasoning and answering, which can greatly improve the response efficiency and enhance the user interaction experience.

[0027] The technical solutions provided by the embodiments of this specification are described in detail below in conjunction with the accompanying drawings.

[0028] Figure 1 A schematic diagram of an application scenario of a method for answering user questions based on an intelligent agent provided in an embodiment of this specification.

[0029] like Figure 1 As shown, on the server 102, an agent driven by a large language model may be deployed. On the user terminal 101, an application for accessing the agent deployed on the server 102 may be installed. When actually applied, the user may ask a user question through a dialogue page displayed on the user terminal 101 based on the application; then, the user terminal 101 may send a question trigger information reflecting the user question to the server 102, so that the server 102: if there is a preset target question corresponding to the user question, obtains the preset reply information corresponding to the user question from the cache center 103, and sends it to the user terminal 101; if there is no preset target question corresponding to the user question, calls the agent driven by the large language model to generate the reply information corresponding to the user question, and sends it to the user terminal 101.

[0030] Figure 1 The server 102 may include but is not limited to any device, equipment, platform, device cluster, etc. with computing and processing capabilities. Figure 1 The user terminal 101 may include but is not limited to a smart phone, a tablet computer, a laptop computer, a PDA, a personal computer, a smart home device, a vehicle-mounted device, etc. Figure 1 The user terminal 101 can interact with the user through a graphical user interface to call the agent in the server 102, thereby implementing the method provided in the embodiments of this specification. Figure 1 The cache center 103 can be a database deployed on any device, equipment, platform, device cluster, etc. with computing and processing capabilities.

[0031] In such Figure 1 In the application scenario shown, the server 102 can be connected to one or more user terminals 101 through a local area network connection, a wide area network connection, an Internet connection or other types of data networks. The server can also be connected to the cache center 103 through a local area network connection, a wide area network connection, an Internet connection or other types of data networks.

[0032] Although for ease of understanding, Figure 1 1 shows a solution in which the cache center 103 and the server 102 are separately arranged. However, in actual application, the cache center 103 can also be deployed on the server 102. Figure 1 As shown in , the agent driven by the large language model can be deployed on the server 102 interacting with the user terminal 101. However, in actual application, the agent driven by the large language model can also be deployed on other server devices connected to the server 102 by wire or wirelessly. Furthermore, the program corresponding to the agent and the program corresponding to the large language model can be deployed on the same or different server devices.

[0033] In addition, considering that the model parameters of the large model are huge and the computing resources of the user terminal are usually limited, the information sending method provided in the embodiment of this specification can be applied to the following: Figure 1 The application scenarios shown are not limited thereto. In the case where the operating resources of the user terminal device can meet the deployment and operating conditions of the large model, the embodiments of this specification can also be performed in the user terminal device.

[0034] In the present application, a method for sending information is provided. The present application also relates to an information sending device and a computing device, which are described in detail one by one in the following embodiments.

[0035] Figure 2 A flowchart of an information sending method provided in an embodiment of this specification.

[0036] From the program perspective, the execution subject of the process can be a program installed on the application server. It can be understood that the method can be executed by any device, equipment, platform, or device cluster with computing and processing capabilities.

[0037] like Figure 2 As shown, the process may include the following steps: Step 202: Acquire problem triggering information related to the target business field sent by the user terminal.

[0038] In actual application, the user can open the dialogue page in the user terminal through a browser, application, applet or other program tool, and then perform user operations in the dialogue page, so that the user terminal generates problem trigger information when sensing the user operation.

[0039] Furthermore, the dialogue page opened by the user may be a page for communicating with an agent. There may be a large number of agents in the program tool, and these agents may target different business areas respectively. That is, different agents may target different business areas. As an example, the business areas may include marketing activity planning, installment payment plan formulation, medical plan formulation, insurance plan planning, etc., but are not limited thereto. In actual application, the scope of the business area, the granularity of division, etc., may be set according to actual needs, and the aforementioned examples do not constitute a limitation on the scope of the business area.

[0040] When executing the solution described in the embodiment of this specification, the user can open a page related to a specific target business field. In actual application, the user can select one of the intelligent agents in the terminal display page to have a dialogue.

[0041] The question trigger information may be information generated in response to the user's dialogue behavior after the user enters the dialogue page with the agent. The user's dialogue behavior may include the behavior of performing touch operations in the dialogue page, the behavior of inputting text information or audio information in the dialogue page, etc. The user's dialogue behavior may not be limited to these examples, and any behavior that can reflect the user's dialogue intention may be regarded as the user's dialogue behavior.

[0042] In an embodiment of the present specification, different types of question trigger information may be generated in response to different dialogue behaviors of the user. For example, if the dialogue behavior of the user includes the behavior of the user performing a touch operation in the dialogue page, then the question trigger information may include control operation information generated in response to the user's touch operation, and the control operation information may reflect, for example, which control the user operated. For another example, if the dialogue behavior of the user includes the behavior of the user inputting text information or audio information in the dialogue page, then the question trigger information may include the actual text information or audio information output by the user, and may also include information obtained by performing data processing (for example, data encryption processing, data compression processing, data format processing, etc.) on the actual text information or audio information input by the user.

[0043] Step 204: If there is a target question corresponding to the question trigger information in the preset question list, the preset reply information of the target question is sent to the user terminal; wherein the preset reply information is the information obtained and stored by the intelligent agent calling the large language model after the target question is input into the intelligent agent for the target business field in advance.

[0044] Among them, the preset question list can be a pre-stored question list. In the embodiments of the present specification, for each business field, the questions related to each business field can be predetermined to obtain a list of questions corresponding to each business field. In practice, for a certain target business field, the questions raised by the user can be expected. Optionally, a large number of historical user question records for the target business field can be obtained, and a preset question list corresponding to the target business field can be determined based on these question records. Alternatively, a large language model can be used to input descriptive information for the target business field, or some examples of questions raised by users under the target business field can be input, so that the large language model can output a preset question list for the target business field. Alternatively, a preset question list for the target business field can be set by an expert. The method for determining the preset question list for the target business field may not be limited to the method of the aforementioned example.

[0045] Further, in the embodiments of the present specification, for the preset questions in the preset question list of the target business field, these preset questions can be input into the intelligent agent of the corresponding target business field in advance, and then the intelligent agent calls the large language model to obtain the preset reply information, and the preset question and the preset reply information can be associated and stored, so that when the user is in an online conversation, if the question raised by the user is the target question in the preset question list, the preset reply information corresponding to the target question generated and stored in advance can be directly obtained. Therefore, it can be avoided that the intelligent agent is called every time the user asks a question to perform question analysis, reasoning, call the large language model to answer, etc., reduce the waiting time of the user to obtain the reply information, and improve the user interaction experience.

[0046] Step 206: If the target question corresponding to the question trigger information does not exist in the preset question list, the question trigger information is input into the agent for the target business field, and the agent calls the large language model to generate reply information, and sends the reply information to the user terminal.

[0047] In an embodiment of the present specification, when it is determined that the question raised by the user is not a question in the preset question list, the intelligent agent can be called again to input the question raised by the user into the intelligent agent so that the intelligent agent can obtain reply information corresponding to the question raised by the user based on the large language model.

[0048] Based on this, reply information can be returned to the user regardless of whether the question raised by the user is a preset question in the preset question list. Moreover, if the question raised by the user is a preset question in the preset question list, the efficiency of returning reply information to the user can be greatly improved. Overall, the solution based on the embodiments of this specification can at least partially improve the efficiency of returning reply information to the user when the user asks questions regarding the target business field, and at least partially improve the user experience.

[0049] It should be understood that in the methods described in one or more embodiments of this specification, the order of some steps can be adjusted according to actual needs, or some steps can be omitted.

[0050] Figure 2 The method in the invention is to input the preset questions in the preset question list into the intelligent agent for the target business field in advance, and then the intelligent agent calls the large language model to obtain the preset reply information, and the preset question is associated with the preset reply information and stored. When the question trigger information related to the target business field sent by the user terminal is received, if there is a target question corresponding to the question trigger information in the preset question list, the preset reply information corresponding to the target question can be returned to the user terminal, without calling the intelligent agent to generate the reply information by using the large language model after the user asks the question and before replying to the user. Therefore, the efficiency of the user in obtaining the question reply information can be improved, and the user interaction experience can be improved.

[0051] based on Figure 2 The present specification also provides some improved implementation methods of the method, which are described below.

[0052] In an optional embodiment of the present specification, based on the control operation information of the user on the control in the dialog page, a target question corresponding to the control operation information can be determined from a preset question list.

[0053] Specifically, the step of obtaining the problem triggering information sent by the user terminal and involving the target business field may include: obtaining the control operation information sent by the user terminal; the control operation information is information generated by the user terminal in response to the user's operation on the problem control displayed on the dialog page. Accordingly, the preset problem list having a target problem corresponding to the problem triggering information may include: determining, according to the control operation information, a preset problem in the preset problem list that has a binding relationship with the problem control as the target problem.

[0054] Among them, the control operation information can reflect the user's operation on the problem control, wherein the user's operation on the problem control can specifically include the operation of clicking the control, the operation of long pressing the control, etc., but is not limited to these operation types.

[0055] Furthermore, the control operation information may carry the control identifier of the problem control. Optionally, in actual application, the control operation information may also carry information such as the control operation user identifier and the control operation time. In an embodiment of the present specification, after the control operation information is acquired, the control identifier may be parsed therefrom, and then, based on the control identifier (that is, the control identifier may be used as a query condition), a preset question having a binding relationship with the control identifier may be queried from a preset question list as the target question corresponding to the control operation information, that is, the user question is determined.

[0056] In the embodiments of the present specification, each preset question in the preset question list may have corresponding preset reply information generated in advance by the agent. Therefore, after determining the target question corresponding to the control operation information sent by the user terminal from the preset question list, the pre-stored preset reply information can be obtained based on the target question and returned to the user terminal, so that the user terminal provides the preset reply information to the user. In this case, the user can obtain the reply information in a very short time after asking the question, and the interactive experience is good.

[0057] Figure 3 A schematic diagram of the interactive interface for a user to communicate with an intelligent agent in a conversation page provided in an embodiment of this specification.

[0058] like Figure 3 As shown in , a question control 301 may be displayed in the dialog page, and the user may trigger the question control 301 by performing an operation on the question control 301. In this way, it is equivalent to raising a user question corresponding to the question control 301 to the user terminal.

[0059] exist Figure 3 , two question controls 301 are shown, specifically, including a question control corresponding to “prediction question 1” and a question control corresponding to “prediction question 2”. Figure 3 The examples given in the figure are only examples. In actual application, the number and form of the question control 301 can be set according to actual needs, and are not limited to the following. Figure 3 An example is given in .

[0060] like Figure 3In the scenario shown, the predicted question 1 and predicted question 2 corresponding to the question control 301 can actually be questions of the next level associated with the first question and predetermined by the server. Specifically, a question list can be pre-stored in the server or in a cache center connected to the server. More specifically, a set of preset question trees can be pre-stored. In each question tree, multiple preset questions and the association relationship information between the multiple preset questions are included. Therefore, when the user Figure 3 After the first question is raised as shown in , the user terminal can request the server to send the first reply information corresponding to the first question. The server can determine the first reply information corresponding to the first question and return it to the user on the one hand, and determine the next level question associated with the first question on the other hand, for example, "prediction question 1" and "prediction question 2" and return them to the user, such as Figure 3 as shown in .

[0061] In addition, although Figure 3 The question control 301 is shown in the figure as being provided to the user while displaying the first reply information in response to the first question raised by the user, but it may not be limited to this in actual application. For example, the question control 301 may be actively pushed to the user without receiving any question raised by the user. For another example, the reply information to the question raised by the user and the question control 301 may be pushed to the user separately when a question raised by the user is received.

[0062] In an optional embodiment of the present specification, based on the user question information input by the user in the dialogue page, a target question matching the user question information can be searched from a preset question list.

[0063] Specifically, in the dialogue page, there may be an information input control, through which the user can input user question information, and in this way, the user question expressed by the user question information is presented to the user terminal. The information input control may include: Figure 3 At least one of the audio input control 302 and the text input control 303 shown in .

[0064] Furthermore, when the user directly inputs question information, it can be determined based on the user question information whether there is a preset question in the preset question list whose matching degree with the user question information satisfies the preset matching condition, and a matching result is obtained; if the matching result indicates that there is a preset question whose matching degree with the user question information satisfies the preset matching condition, then the preset question whose matching degree with the user question information satisfies the preset matching condition can be determined as the target question that matches the user question information, and then, the pre-stored preset reply information corresponding to the target question can be obtained and returned to the user; if the matching result indicates that there is no preset question whose matching degree with the user question information satisfies the preset matching condition, then the intelligent agent for the target business field can be called to determine the reply information corresponding to the user question information, and the reply information is returned to the user.

[0065] Specifically, the obtaining of problem trigger information related to the target business field sent by the user terminal may include: obtaining user problem information sent by the user terminal; the user problem information is information related to the target business field input by the user in the dialogue page; accordingly, there is a target problem corresponding to the problem trigger information in the preset problem list, which may include: based on the user problem information, determining the target problem matching the user problem information from the preset problem list.

[0066] The user question information may specifically include information in different forms such as text information and voice information.

[0067] Furthermore, the preset question list may specifically include a question tree set; determining the target question matching the user question information from the preset question list may specifically include: determining a user question tree based on the user question information; identifying a preset question subtree matching the user question tree from the question tree set; and determining the last-level node of the preset question subtree as the target question corresponding to the user question information.

[0068] In practical applications, the process of identifying a preset problem subtree from a problem tree set may adopt, for example, a brute force search algorithm. Specifically, each preset problem tree in the problem tree set may be traversed to determine a preset problem subtree in each preset problem tree, and then each traversed preset problem subtree may be compared with the user problem tree to identify a preset problem subtree that matches the user problem tree.

[0069] Furthermore, the process of determining the preset question subtree that matches the user question tree from all the traversed preset question subtrees can adopt, for example, a machine learning algorithm, a database matching algorithm, a large model matching algorithm, a custom rule matching algorithm, etc., but is not limited to this.

[0070] Optionally, a machine learning algorithm may be used to determine a preset question subtree that matches the user question tree. Specifically, a kernel function may be used to calculate the similarity between two trees, so as to determine a preset question tree whose similarity meets a preset condition from multiple preset question subtrees.

[0071] Optionally, a database matching algorithm may be used to determine a preset question subtree that matches the user question tree. Specifically, a feature vector may be used to calculate the similarity between two trees, so as to determine a preset question tree whose similarity meets a preset condition from multiple preset question subtrees.

[0072] Optionally, a large model matching algorithm may be used to determine a preset question subtree that matches the user question tree. Specifically, the reading comprehension ability of the large model may be used to compare the semantics of the two trees, so as to determine a preset question tree whose semantic similarity meets a preset condition from multiple preset question subtrees.

[0073] Optionally, a custom rule matching algorithm may be used to determine a preset question subtree that matches the user question tree. Specifically, rule matching may be performed by coding according to actual needs, so as to determine a preset question tree that meets the preset rule conditions from multiple preset question subtrees according to the rules configured by coding.

[0074] In an optional embodiment of the present specification, a graph matching algorithm may be used to identify a preset problem subtree that matches the user problem tree from a problem tree set. Specifically, the tree may be converted into a graph and then solved using a graph matching algorithm. More specifically, the user problem tree may be converted into a user problem graph; each preset problem tree in the problem tree set may be converted into a preset problem graph, and each preset problem subgraph in the preset problem graph may be determined; and then, a graph matching algorithm may be used to identify a preset problem subgraph that matches the user problem graph, thereby indirectly obtaining a preset problem subtree that matches the user problem tree.

[0075] In an optional embodiment, determining the user question tree based on the user question information may specifically include: extracting at least two sub-question information having an associated relationship from the user question information; and constructing a user question tree based on the at least two sub-question information.

[0076] For example, the user question information may be "travel plan to Harbin in December this year". From the user question information, two sub-question information "December" and "Harbin" with an associated relationship may be extracted, and these two sub-question information may constitute a user question tree. Optionally, the determined user question tree may include: a first-level node - "Harbin", a second-level node - "December". Or optionally, the determined user question tree may include: a first-level node - "December", a second-level node - "Harbin".

[0077] In an optional embodiment, determining the user question tree based on the user question information may specifically include: determining the user question tree according to the user question information and historical question information before the user question information.

[0078] For example, the user question information may be "plan to travel for 5 days", and the historical question information before the user question information may be "travel plan to Harbin in December this year", then the user question tree may be determined based on the user question information and the historical question information. Optionally, the determined user question tree may include: a first-level node - "Harbin", a second-level node - "December", and a third-level node - "travel for 5 days". Or optionally, the determined user question tree may include: a first-level node - "December", a second-level node - "Harbin", and a third-level node - "travel for 5 days".

[0079] Furthermore, after the question trigger information is input into the agent for the target business field, the agent calls the large language model to generate reply information, which may specifically include: the agent generates large model prompt words based on the user question information; the large model prompt words are input into the large language model to obtain reply information determined by the large language model.

[0080] In actual application, the operation mode of the intelligent agent driven by the large language model may include: when the intelligent agent is called, it receives user question information, and then can perform analysis and reasoning based on the user question information, and generate prompt words for input to the large model, so that the large language model generates reply information for the user question information according to the prompt words.

[0081] In the embodiments of the present specification, further, after obtaining the reply information determined by the large language model, the method may also include: storing the user question information in association with the reply information.

[0082] In actual application, during an online conversation with a user, each time the intelligent agent calls the large language model to generate reply information, the generated reply information can be associated with the user question information and stored, so that when the user or other users ask user questions consistent with the user question information later, the pre-stored reply information can be directly obtained for return, thereby reducing the waiting time of the user or other users in subsequent inquiries and improving the user interaction experience.

[0083] In an embodiment of the present specification, the question list and the preset answer information corresponding to each preset question in the question list can be stored in a cache center; the cache center can specifically include a memory database, a cache database, a vector database, etc., but is not limited thereto.

[0084] In one or more embodiments of the present specification, a specific solution is further provided for obtaining pre-stored preset response information based on the determined target question.

[0085] Optionally, the preset question and the preset reply information corresponding to the preset question can be directly associated and stored. That is, the preset reply information can be directly queried based on the target question. In actual application, the preset question and the preset reply information can be stored in the same storage space. For example, the preset question and the preset reply information can be stored in the same storage device or equipment, or the preset question and the preset reply information can be stored in the same database.

[0086] Specifically, sending the preset reply information of the target question to the user terminal may include: acquiring, according to the target question, the preset reply information stored in association with the preset question; and sending the preset reply information to the user terminal.

[0087] Optionally, the preset question and the preset reply information corresponding to the preset question can be stored in an indirectly associated manner. That is, based on the target question, the intermediate information can be first queried, and then the preset reply information can be queried based on the intermediate information. In actual application, the preset question and the preset reply information can be stored in different storage spaces. For example, the preset question and the preset reply information can be stored in different storage devices or equipment, and the preset question and the preset reply information can be stored in different databases.

[0088] Specifically, obtaining preset reply information stored in association with the preset question according to the target question may include: determining reply query information corresponding to the preset question; the reply query information is used to indicate a query path for the preset reply information; and obtaining the preset reply information based on the reply query information.

[0089] The reply query information is information used to query the preset reply information, and can be used to indicate a query path for the preset reply information. More specifically, it can be identification information used to mark the preset reply information, or it can be address information used to point to the preset reply information. In actual application, the reply query information can include at least one of the identification information and the identification information of the preset reply information.

[0090] Furthermore, the data volume of the reply query information may be smaller than the data volume of the preset reply information. In other words, the storage space occupied by the reply query information is smaller than the storage space occupied by the preset reply information.

[0091] Based on the above scheme of the embodiments of the present specification, in actual application, if the content of the preset reply information is small and occupies less storage resources, the preset reply information can be directly associated with the preset question for storage, thereby, the query efficiency for the preset reply information is higher; if the content of the preset reply information is large and occupies more storage resources, the preset reply question and the preset reply information can be stored separately, and only the reply query information that occupies less storage resources can be directly associated with the preset question for storage, thereby, considering that the data volume of the reply query information is smaller and is usually standardized data, it can be more convenient for information management and query, and it is also convenient to improve the query efficiency when querying the preset reply information.

[0092] In one or more embodiments of the present specification, a specific scheme for predetermining a preset question list (a preset question tree set) is further provided, and a specific scheme for calling an intelligent agent to generate preset response information for preset questions in the preset question list is further provided.

[0093] Specifically, before determining the target question matching the user question information from a preset question list based on the user question information, the method may also include: obtaining a first question in the target business field; generating a first prompt word based on the first question and a first prompt word template; the first prompt word template is used to instruct a large language model to output reply information corresponding to the first question; inputting the first prompt word into the large language model to obtain first reply information output by the large language model; and storing the first question in association with the first reply information.

[0094] The first question may be a pre-prepared question. Specifically, the first question may be generated based on historical conversation records of a large number of users, or may be pre-generated by a large language model, or may be pre-set by a business person. The acquisition method of the first question is not limited to these methods.

[0095] Specifically, the first prompt word template can be used to instruct the large language model to output the reply information corresponding to the question. When the question is the first question, the first prompt word generated based on the first prompt word template can be used to instruct the large language model to output the first reply information corresponding to the first question.

[0096] Furthermore, the generating of the first prompt word based on the first question and the first prompt word template may specifically include: based on the first question, acquiring reference knowledge required for answering the first question from a domain knowledge base corresponding to the target business field; and generating the first prompt word based on the first question, the reference knowledge and the first prompt word template.

[0097] Specifically, the first prompt word template can be used to instruct the large language model to output a response corresponding to the text and image based on the reference knowledge. Thus, the first prompt word can be used to instruct the large language model to output first response information corresponding to the first question based on the parameter instruction.

[0098] In actual application, the first prompt word template may include role portrait information, input prompt information, output prompt information, etc., but is not limited to this. Among them, the role portrait information can be used to specify the role of the large language model so that the output of the large language model meets expectations, for example, "You are a professional travel consultant, please xxxx". The input prompt information may include task information, reference knowledge, examples, etc., to make the output of the large language model meet expectations, for example, "You can refer to the following information xxxx, you can refer to the following example xxxx". The output prompt information can be used to express requirements for the format and content of the information output by the large language model, for example, "Please specify a detailed plan, including xxxx", "Output in the format of xxxx", etc. Correspondingly, the first prompt word generated based on the first prompt word template may include role portrait information, input prompt information, output prompt information, etc., and may not be limited to this in actual application.

[0099] In actual application, the scheme of predetermining a list of preset questions and their corresponding preset answer information can be applied to a server, and more specifically, can be applied to an intelligent agent model running in the server.

[0100] For example, assuming that the first question may be "travel plan to Harbin in December", then, after analysis and reasoning by the intelligent agent, it can first obtain the reference knowledge required to formulate this travel plan from the domain knowledge base about travel plans based on the content of the first question, such as city tourist attractions information, city weather information, city traffic information, city consumption information, etc.; then, based on the first question and the reference knowledge obtained by the query, the first prompt word template can be combined to generate the first prompt word.

[0101] In an optional embodiment of the present specification, the large language model may be called more than once during the process of the intelligent agent generating preset response information corresponding to a preset question.

[0102] Optionally, before determining the target question matching the user question information from a preset question list based on the user question information, the method may also include: obtaining a first question in the target business field; generating a first prompt word based on the first question and a first prompt word template; inputting the first prompt word into a large language model to obtain first reply information output by the large language model; generating a second prompt word based on the first reply information and a second prompt word template; inputting the second prompt word into the large language model to obtain second reply information output by the large language model; and storing the first question in association with the second reply information.

[0103] Specifically, the first prompt word template can be used to instruct the large language model to output the first reply information corresponding to the question; the second prompt word template can be used to instruct the large language model to further output the second reply information corresponding to the question with reference to the first reply information. When the question is the first question, the first prompt word generated based on the first prompt word template can be used to instruct the large language model to output the first reply information corresponding to the first question; further, the second prompt word generated based on the first reply information and the second prompt word template can be used to instruct the large language model to generate the second reply information corresponding to the first question with reference to the first reply information.

[0104] Further, similar to the process of generating the first prompt word, the process of generating the second prompt word based on the first reply information may optionally include: based on the first reply information, obtaining reference knowledge required to answer the first question from a domain knowledge base corresponding to the target business field, and generating a second prompt word based on the first reply information and the reference knowledge.

[0105] The above only gives an example of the intelligent agent calling the large language model twice in the process of generating preset response information corresponding to the preset question. In actual application, the large language model can be called more times according to actual business needs.

[0106] In the embodiments of the present specification, each question in a preset question list (for example, a preset question tree) and the association relationship information between each question may be generated based on historical conversation records of a large number of users, or may be pre-generated by a large language model, or may be pre-set by an expert, but is not limited to these methods.

[0107] In at least some embodiments of the present specification, a large language model may be called by an intelligent agent to generate preset questions.

[0108] Optionally, after obtaining the first reply information output by the large language model, the method may also include: generating a third prompt word based on the first question, the first reply information and a third prompt word template; the third prompt word template is used to indicate the large language model predicts a question after the first question; inputting the third prompt word into the large language model to obtain a second question after the first question predicted by the large language model; and generating and storing association relationship information between the second question and the first question.

[0109] In actual application, the association relationship information between the second question and the first question may specifically be question tree structure information, for example, which may be used to indicate that the second question is a child node of the first question in the question tree.

[0110] For example, assuming that the first question may be "travel plan to Harbin in December", the first reply information may be "on the first day, at xx o'clock, take xx transportation from place A to place B;...; from xx o'clock to xx o'clock, visit scenic spot C;..."; then, the aforementioned first question and the first reply information may be combined with the third prompt word template to generate a third prompt word. The first prompt word may, for example, include the following content, "After the user asks {travel plan to Harbin in December}, the reply is {on the first day, at xx o'clock, take xx transportation from place A to place B;...; from xx o'clock to xx o'clock, visit scenic spot C;...}. Please predict the question that the user may ask next". As an example, the second question may be "how to buy tickets for scenic spot C" or "please help me buy tickets for scenic spot C", "are there other transportation routes", etc.

[0111] Based on the above examples, it can be seen that in the embodiments of the present specification, when determining the pre-prepared preset questions, a question tree can be automatically constructed with the help of an intelligent agent driven by a large language model. For example, one or more second-level questions can be predicted based on a first-level question, and it can be understood that one or more third-level questions can also be predicted based on any second-level question in one or more second-level questions, without being limited to this.

[0112] In one or more embodiments of the present specification, the target problem is used to instruct the agent to formulate a solution to the problem.

[0113] Among them, the agent, as an agent that can perceive the environment and take actions to achieve specific goals, can think and perform actions instead of people. As an example, the agent can help users make travel plans. The agent is equivalent to the user's travel consultant. It can replace the user's thinking, understand the user's needs and preferences, complete the search for travel-related information, arrange itineraries, and place orders for air tickets or hotels. In another example, the agent can help users develop marketing plans. The agent is equivalent to a business manager. It can think for the user, comprehensively understand the user's problems, and break down the steps required to solve the problem, and then execute each step in sequence, and then filter and summarize the information, and finally feedback to the user.

[0114] With reference to the above examples, it can be understood that when an intelligent agent faces a user question for instructing to formulate a solution to a problem, it takes a certain amount of time to understand the problem, decompose the problem, and call various models including large language models to solve the problem. This process is usually time-consuming and the user needs to wait for seconds, minutes, or even hours. In view of this, based on the solution provided in the embodiments of this specification, since a list of questions is pre-determined for the target business field and the intelligent agent is pre-called to determine and store the preset reply information corresponding to the preset question, when a user asks a question online, the reply information can be returned to the user in a timely manner if the corresponding question and the corresponding reply information are pre-stored, thereby improving the efficiency of information reply and enhancing the user interaction experience, at least partially solving the problem of slow reply of the intelligent agent and poor user interaction experience.

[0115] The various technical features in the above embodiments can be combined arbitrarily as long as there is no conflict or contradiction between the combinations of features. However, due to space limitations, they are not described one by one. Therefore, any combination of the various technical features in the above embodiments also falls within the scope of this specification.

[0116] Figure 4 A swimlane diagram of a method for sending a question reply message to a user in response to a user's question in an actual application scenario provided by an embodiment of the present specification.

[0117] like Figure 4 As shown in , it is possible to pre-load the question tree that may be involved in the session, use the intelligent agent to perform advance reasoning, and associate the reasoned response information with the question tree and store it in the cache center; when the user requests (that is, when the user asks a user question), the user question tree can be submitted to the cache center for matching to obtain the pre-stored answer.

[0118] Specifically, in step 402, a preset problem tree may be obtained in the server.

[0119] In step 404, the intelligent agent may be used to call a large language model to generate preset response information corresponding to each question in the preset question tree.

[0120] In step 406, the preset question tree and the preset response information may be associated and stored in a cache center.

[0121] In step 408, the user terminal may receive the user's dialogue operation, such as the user's triggering operation on the question control, the user's input of question information in text or voice form, etc. In actual application, the user may use the conversation product to interact with the intelligent robot in natural language and wait for the intelligent robot to return a reply message.

[0122] Step 410: The user terminal may send question trigger information to the server in response to the dialogue operation.

[0123] Step 412: The server may determine the user problem tree according to the problem triggering information.

[0124] In step 414, the server may call the cache center acceleration service to match the user's question tree with the preset question tree of the cache center.

[0125] Step 416: If there is a preset target question matching the user question tree in the cache center, preset response information corresponding to the target question may be returned.

[0126] In step 418, the server may send the preset reply information to the user terminal.

[0127] Step 420: The user terminal may output the preset reply information to the user. In actual application, the intelligent robot may display the preset reply information to the user in the form of natural language.

[0128] Step 422: If there is no preset target question matching the user question tree in the cache center, result information indicating a miss may be returned.

[0129] In step 424, after receiving the result information indicating a miss, the server may utilize the agent to call the large language model to generate reply information corresponding to the user question.

[0130] Step 426: The server may send the preset reply information to the user terminal.

[0131] Step 428: The user terminal may output the reply information to the user. In actual application, the intelligent robot may display the reply information to the user in the form of natural language.

[0132] In step 430, the user question tree may be associated with the reply information and stored.

[0133] like Figure 4 As shown in , the cache center in the embodiment of this specification can, on the one hand, provide a storage service for writing the question tree and the corresponding reply information; on the other hand, it can provide a matching and recall service for the reply information corresponding to the question tree. In actual application, when matching and recalling, for example, a tree matching or graph matching algorithm can be used, which greatly improves the reply efficiency compared to the process of inputting the user question into the intelligent agent for analysis and reasoning and giving the reply information.

[0134] Based on at least some of the above embodiments of the present specification, in the face of the problem that it takes a long time for the intelligent agent to answer user questions and the user interaction experience is not good, there is no need to increase the GPU computing cost, nor is there any need for further fine development or pre-training or re-training of machine intelligent agents or large models. Instead, a pre-inference method is used to pre-load questions, and a question tree matching mechanism is used to ensure recall during real-time conversations. In this way, a large amount of cost can be saved, while meeting the user's requirements for timeliness, improving the user's interaction experience, and having great commercial production value.

[0135] Based on the same idea, the embodiments of this specification also provide a device corresponding to the above method.

[0136] Figure 5 The embodiments of this specification provide corresponding to Figure 2 A structural schematic diagram of an information sending device.

[0137] like Figure 5 As shown, the device may include: The information acquisition module 502 is used to acquire the problem triggering information related to the target business field sent by the user terminal; The first information sending module 504 is used to send preset reply information of the target question to the user terminal if there is a target question corresponding to the question trigger information in the preset question list; wherein the preset reply information is information obtained and stored by the agent after the target question is input into the agent for the target business field in advance and then the agent calls the large language model; The second information sending module 506 is used to input the question trigger information into the agent for the target business field, and then the agent calls the large language model to generate reply information if there is no target question corresponding to the question trigger information in the preset question list, and send the reply information to the user terminal.

[0138] based on Figure 5 The present specification also provides some specific implementation schemes of the method, which are described below.

[0139] Optionally, the information acquisition module 502 can be specifically used to: obtain control operation information sent by the user terminal; the control operation information is information generated by the user terminal in response to the user's operation on the question control displayed on the dialogue page; the first information sending module 504 can be specifically used to: according to the control operation information, determine the preset question in the preset question list that has a binding relationship with the question control as the target question.

[0140] Optionally, the information acquisition module 502 can be specifically used to: obtain user question information sent by the user terminal; the user question information is information related to the target business field entered by the user in the dialogue page; the first information sending module 504 can be specifically used to: based on the user question information, determine the target question that matches the user question information from a preset question list.

[0141] Optionally, the preset question list may specifically include a question tree set; determining the target question matching the user question information from the preset question list may specifically include: determining a user question tree based on the user question information; identifying a preset question subtree matching the user question tree from the question tree set; and determining the last-level node of the preset question subtree as the target question corresponding to the user question information.

[0142] Optionally, determining the user question tree based on the user question information may specifically include: extracting at least two sub-question information having an associated relationship from the user question information; and constructing a user question tree based on the at least two sub-question information.

[0143] Optionally, determining the user problem tree based on the user problem information may specifically include: determining the user problem tree according to the user problem information and historical problem information before the user problem information.

[0144] Optionally, the second information sending module 506 can be specifically used to: generate a large model prompt word based on the user question information by the agent; input the large model prompt word into the large language model to obtain reply information determined by the large language model.

[0145] Optionally, the information sending device may also be used to: store the user question information and the reply information in association with each other.

[0146] Optionally, the second information sending module 506 may be specifically configured to: acquire preset reply information stored in association with the preset question according to the target question; and send the preset reply information to the user terminal.

[0147] Optionally, obtaining preset reply information stored in association with the preset question according to the target question may specifically include: determining reply query information corresponding to the preset question; the reply query information is used to indicate a query path for the preset reply information; and obtaining the preset reply information based on the reply query information.

[0148] Optionally, the information sending device can also be used to: obtain a first question in the target business field; generate a first prompt word based on the first question and a first prompt word template; the first prompt word template is used to instruct the large language model to output reply information corresponding to the first question; input the first prompt word into the large language model to obtain the first reply information output by the large language model; and store the first question in association with the first reply information.

[0149] Optionally, generating the first prompt word based on the first question and the first prompt word template may specifically include: based on the first question, acquiring reference knowledge required to answer the first question from a domain knowledge base corresponding to the target business field; generating the first prompt word based on the first question, the reference knowledge and the first prompt word template.

[0150] Optionally, the information sending device can also be used to: obtain a first question in the target business field; generate a first prompt word based on the first question and a first prompt word template; input the first prompt word into a large language model to obtain first reply information output by the large language model; generate a second prompt word based on the first reply information and a second prompt word template; input the second prompt word into the large language model to obtain second reply information output by the large language model; and store the first question in association with the second reply information.

[0151] Optionally, the information sending device can also be used to: generate a third prompt word based on the first question, the first reply information and a third prompt word template; the third prompt word template is used to indicate the large language model predicts a question after the first question; input the third prompt word into the large language model to obtain a second question after the first question predicted by the large language model; generate and store the association relationship information between the second question and the first question.

[0152] Optionally, the target problem is used to instruct the agent to formulate a solution to the problem.

[0153] It is understood that the above modules refer to computer programs or program segments for executing one or more specific functions. In addition, the distinction between the above modules does not mean that the actual program codes must also be separated.

[0154] The above is a schematic scheme of an information sending device of this embodiment. It should be noted that the technical scheme of the information sending device and the technical scheme of the above information sending method belong to the same concept, and the details not described in detail in the technical scheme of the information sending device can be referred to the description of the technical scheme of the above information sending method.

[0155] Based on the same idea, the embodiments of this specification also provide a device corresponding to the above method.

[0156] Figure 6 The embodiments of this specification provide corresponding to Figure 2 A schematic diagram of the structure of an information sending device. Figure 6 As shown, the device 600 may include: at least one processor 610; and, A memory 630 in communication with the at least one processor; wherein, The memory 630 stores instructions 620 executable by the at least one processor 610. The instructions are executed by the at least one processor 610 to enable the at least one processor 610 to: Obtaining problem triggering information related to the target business field sent by the user terminal; If there is a target question corresponding to the question trigger information in the preset question list, the preset reply information of the target question is sent to the user terminal; wherein the preset reply information is information obtained and stored by the intelligent agent calling the large language model after the target question is input into the intelligent agent for the target business field in advance; If the target question corresponding to the question trigger information does not exist in the preset question list, the question trigger information is input into the agent for the target business field, and the agent calls the large language model to generate reply information, and sends the reply information to the user terminal.

[0157] Based on the same idea, the embodiment of this specification also provides a computer-readable medium corresponding to the above method. The computer-readable medium stores computer-readable instructions, which can be executed by a processor to implement the following method: The above is a schematic scheme of a computing device of this embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the above information sending method belong to the same concept, and the details not described in detail in the technical scheme of the computing device can be referred to the description of the technical scheme of the above information sending method.

[0158] An embodiment of the present application further provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the steps of the information sending method as described above.

[0159] The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical scheme of the storage medium and the technical scheme of the above information sending method belong to the same concept, and the details not described in detail in the technical scheme of the storage medium can be referred to the description of the technical scheme of the above information sending method.

[0160] An embodiment of the present specification also provides a computer program product, including a computer program / instruction, which implements the steps of the above-mentioned information sending method when executed by a processor.

[0161] The above is an illustrative solution of a computer program product of this embodiment. It should be noted that the technical solution of the computer program product and the technical solution of the above information sending method belong to the same concept, and the details not described in detail in the technical solution of the computer program product can be referred to the description of the technical solution of the above information sending method.

[0162] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments. In particular, for the device and equipment embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments. The devices, equipment and methods provided in the embodiments of this specification correspond to each other, so the devices and equipment also have beneficial technical effects similar to the corresponding methods. Since the beneficial technical effects of the methods have been described in detail above, the beneficial technical effects of the corresponding devices and equipment will not be repeated here.

[0163] The above describes specific embodiments of the present application. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the accompanying drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0164] In the 1990s, it was very clear whether the improvement of a technology was hardware improvement (for example, improvement of the circuit structure of diodes, transistors, switches, etc.) or software improvement (improvement of the method flow). However, with the development of technology, many improvements of the method flow today can be regarded as direct improvements of the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that the improvement of a method flow cannot be implemented with a hardware entity module. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming themselves, without having to ask chip manufacturers to design and make dedicated integrated circuit chips. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one kind of HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0165] The controller may be implemented in any suitable manner, for example, the controller may take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) executable by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller may also be implemented as part of the control logic of the memory. It is also known to those skilled in the art that, in addition to implementing the controller in a purely computer-readable program code manner, the controller may be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, such a controller may be considered as a hardware component, and the devices for implementing various functions included therein may also be considered as structures within the hardware component. Or even, the devices for implementing various functions may be considered as both software modules for implementing the method and structures within the hardware component.

[0166] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0167] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing the present application, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0168] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0169] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0170] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0171] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0172] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0173] Memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.

[0174] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disc read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined in this article, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0175] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0176] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0177] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for sending information, comprising: Obtaining problem triggering information related to the target business field sent by the user terminal; If there is a target question corresponding to the question trigger information in the preset question list, the preset reply information of the target question is sent to the user terminal; wherein the preset reply information is information obtained and stored by the intelligent agent calling the large language model after the target question is input into the intelligent agent for the target business field in advance; If the target question corresponding to the question trigger information does not exist in the preset question list, the question trigger information is input into the agent for the target business field, and the agent calls the large language model to generate reply information, and sends the reply information to the user terminal.

2. The method according to claim 1, wherein obtaining the problem triggering information related to the target business field sent by the user terminal specifically comprises: Obtain control operation information sent by the user terminal; The control operation information is information generated by the user terminal in response to the user's operation on the question control displayed in the dialogue page; The preset question list contains target questions corresponding to the question triggering information, specifically including: According to the control operation information, a preset question in a preset question list that has a binding relationship with the question control is determined as a target question.

3. The method according to claim 1, wherein obtaining the problem triggering information related to the target business field sent by the user terminal specifically comprises: Obtain user question information sent by the user terminal; The user question information is the information related to the target business field entered by the user in the dialogue page; The preset question list contains target questions corresponding to the question triggering information, specifically including: Based on the user question information, a target question matching the user question information is determined from a preset question list.

4. The method according to claim 3, wherein the preset question list specifically includes a question tree set; and determining the target question matching the user question information from the preset question list specifically includes: Based on the user problem information, determining a user problem tree; From the problem tree set, identifying a preset problem subtree that matches the user problem tree; The last level node of the preset question subtree is determined as the target question corresponding to the user question information.

5. The method according to claim 4, wherein determining the user problem tree based on the user problem information specifically comprises: Extracting at least two sub-question information having a correlation relationship from the user question information; Based on the at least two sub-question information, a user question tree is constructed.

6. The method according to claim 4, wherein determining the user problem tree based on the user problem information specifically comprises: A user question tree is determined according to the user question information and historical question information before the user question information.

7. The method according to claim 1, wherein the inputting of the question trigger information into the agent for the target business field and the agent invoking a large language model to generate a reply message specifically comprises: The intelligent agent generates a large model prompt word based on the user question information; The large model prompt word is input into the large language model to obtain the reply information determined by the large language model.

8. The method according to claim 7, after obtaining the reply information determined by the large language model, further comprising: The user question information and the reply information are associated and stored.

9. The method according to claim 1, wherein sending the preset answer information of the target question to the user terminal specifically comprises: According to the target question, obtaining preset answer information stored in association with the preset question; The preset reply information is sent to the user terminal.

10. The method according to claim 9, wherein obtaining, according to the target question, preset answer information stored in association with the preset question comprises: Determine the reply query information corresponding to the preset question; The reply query information is used to indicate a query path for the preset reply information; Based on the reply query information, the preset reply information is obtained.

11. The method according to claim 3, before determining a target question matching the user question information from a preset question list based on the user question information, further comprising: Obtaining a first problem in the target business area; Based on the first question and the first prompt word template, generate a first prompt word; The first prompt word template is used to instruct the large language model to output reply information corresponding to the first question; Inputting the first prompt word into a large language model to obtain first reply information output by the large language model; The first question is associated with the first reply information and stored.

12. The method according to claim 11, wherein generating the first prompt word based on the first question and the first prompt word template specifically comprises: Based on the first question, obtaining reference knowledge required to answer the first question from a domain knowledge base corresponding to the target business domain; A first prompt word is generated based on the first question, the reference knowledge and a first prompt word template.

13. The method according to claim 3, before determining the target question matching the user question information from a preset question list based on the user question information, further comprising: Obtaining a first problem in the target business area; Based on the first question and the first prompt word template, generate a first prompt word; Inputting the first prompt word into a large language model to obtain first reply information output by the large language model; Generate a second prompt word based on the first reply information and the second prompt word template; Inputting the second prompt word into the large language model to obtain second reply information output by the large language model; The first question is associated with the second reply information and stored.

14. The method according to claim 11, after obtaining the first reply information output by the large language model, further comprising: generating a third prompt word based on the first question, the first reply information and a third prompt word template; The third prompt word template is used to instruct the large language model to predict a question after the first question; Inputting the third prompt word into a large language model to obtain a second question after the first question predicted by the large language model; Generate and store association relationship information between the second question and the first question.

15. The method according to any one of claims 1 to 14, wherein: The target problem is used to instruct the agent to formulate a solution to the problem.

16. An information sending device, comprising: An information acquisition module, used to acquire problem triggering information related to a target business field sent by a user terminal; A first information sending module is configured to send preset reply information of the target question to the user terminal if there is a target question corresponding to the question trigger information in the preset question list; wherein the preset reply information is information obtained and stored by the agent after the target question is input into the agent for the target business field in advance and then the agent calls the large language model; The second information sending module is used to input the question trigger information into the agent for the target business field, and then the agent calls the large language model to generate reply information if there is no target question corresponding to the question trigger information in the preset question list, and send the reply information to the user terminal.

17. An information sending device, comprising: at least one processor; as well as, a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to: Obtaining problem triggering information related to the target business field sent by the user terminal; If there is a target question corresponding to the question trigger information in the preset question list, the preset reply information of the target question is sent to the user terminal; wherein the preset reply information is information obtained and stored by the intelligent agent calling the large language model after the target question is input into the intelligent agent for the target business field in advance; If the target question corresponding to the question trigger information does not exist in the preset question list, the question trigger information is input into the agent for the target business field, and the agent calls the large language model to generate reply information, and sends the reply information to the user terminal.

Images