Information interaction method and device, electronic equipment and storage medium

Abstract

The invention provides an information interaction method, and relates to the technical field of artificial intelligence, in particular to the technical fields of large models, computer vision, deep learning and the like. The specific implementation scheme is as follows: in response to a received dialogue request, processing the dialogue request by using a large language model to obtain a task function for the dialogue request; calling the expert model by executing the task function, and generating at least one piece of candidate reply information; screening reply information from the candidate reply information by using an evaluation model; and outputting the reply information. The invention further provides an information interaction device, electronic equipment and a storage medium.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to the fields of large models, computer vision, and deep learning. More specifically, the present disclosure provides an information interaction method, device, electronic device, and storage medium. Background Art

[0002] With the rapid development of artificial intelligence technology, Internet service platforms can train deep learning models based on actual business needs, and use pre-trained deep learning models to help process business information, thereby improving business service efficiency. Summary of the invention

[0003] The present disclosure provides an information interaction method, apparatus, device and storage medium.

[0004] According to one aspect of the present disclosure, there is provided an information interaction method, the method comprising: in response to a received dialogue request, processing the dialogue request using a large language model to obtain a task function for the dialogue request; calling an expert model by executing the task function to generate at least one candidate reply information; using an evaluation model to filter out reply information from the candidate reply information; and outputting the reply information.

[0005] According to another aspect of the present disclosure, an information interaction device is provided, which includes: a task planning module, which is used to respond to a received dialogue request, process the dialogue request using a large language model, and obtain a task function for the dialogue request; a generation module, which is used to call an expert model by executing the task function to generate at least one candidate reply information; a screening module, which is used to use an evaluation model to screen reply information from the candidate reply information; and an output module, which is used to output the reply information.

[0006] According to another aspect of the present disclosure, an artificial intelligence agent is provided, which is configured to execute the method provided by the embodiment of the present disclosure.

[0007] According to another aspect of the present disclosure, an electronic device is provided, comprising: 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, and the instructions are executed by the at least one processor so that the at least one processor can execute the method provided according to the present disclosure.

[0008] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided. The computer instructions are used to cause a computer to execute the method provided according to the present disclosure.

[0009] According to another aspect of the present disclosure, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the method provided according to the present disclosure is implemented.

[0010] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0012] Figure 1 A schematic diagram of an application scenario in which the information interaction method and device according to an embodiment of the present disclosure can be applied is shown schematically;

[0013] Figure 2 A schematic diagram of the system architecture of an intelligent agent according to an embodiment of the present disclosure is shown;

[0014] Figure 3 A flowchart of an information interaction method according to an embodiment of the present disclosure is schematically shown;

[0015] Figure 4 A flowchart schematically shows a task function determination process according to an embodiment of the present disclosure;

[0016] Figure 5 A flowchart schematically shows a reply information screening process according to an embodiment of the present disclosure;

[0017] Figure 6 is a block diagram of an information interaction device according to an embodiment of the present disclosure;

[0018] Figure 7 A schematic diagram of a structure of an intelligent agent according to an embodiment of the present disclosure is shown; and

[0019] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement embodiments of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0020] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0021] In the technical solutions disclosed in this disclosure, the collection, storage, use, processing, transmission, provision, disclosure and application of user personal information involved are in compliance with the relevant laws and regulations, necessary confidentiality measures are taken, and they do not violate public order and good morals. In the technical solutions disclosed in this disclosure, the user's authorization or consent is obtained before obtaining or collecting user personal information.

[0022] The embodiments of the present disclosure provide an information interaction method, device, electronic device and storage medium. The information interaction method includes: in response to a received dialogue request, using a large language model to process the dialogue request to obtain a task function for the dialogue request; calling an expert model by executing the task function to generate at least one candidate reply information; using an evaluation model to filter out reply information from the candidate reply information; and outputting the reply information.

[0023] According to the embodiments of the present disclosure, by constructing an intelligent agent framework for large model planning and multi-tool calling, the user conversation intention is understood and the user conversation request is diverted to different expert tools through reasonable planning of tasks, thereby reducing the access to general experts, such as large models such as eb3.5 / eb4, reducing the access frequency of large models, and reducing the overall system impact.

[0024] Figure 1 A schematic diagram of an application scenario to which the information interaction method and device according to an embodiment of the present disclosure can be applied is schematically shown.

[0025] It should be noted that Figure 1 The examples shown are only examples of system architectures to which the embodiments of the present disclosure can be applied, in order to help those skilled in the art understand the technical content of the present disclosure, but do not mean that the embodiments of the present disclosure cannot be used in other devices, systems, environments or scenarios. For example, in another embodiment, the exemplary system architecture to which the information interaction method and apparatus can be applied may include a terminal device, but the terminal device may implement the information interaction method and apparatus provided by the embodiments of the present disclosure without interacting with the server.

[0026] like Figure 1As shown, the system architecture 100 according to this embodiment may include terminal devices 101, 102, 103, a network 104 and a server 105. The network 104 is used to provide a medium for a communication link between the terminal devices 101, 102, 103 and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc. Users may use the terminal devices 101, 102, 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the terminal devices 101, 102, 103, such as knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients and / or social platform software, etc. (only as examples).

[0027] The terminal devices 101 , 102 , and 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0028] The server 105 may be a server that provides various services, such as a background management server (only an example) that provides support for the content browsed by the user using the terminal devices 101, 102, and 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0029] Server 105 may also be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server may also be a server of a distributed system, or a server combined with a blockchain.

[0030] It should be noted that the information interaction method provided in the embodiment of the present disclosure can generally be executed by the server 105. Accordingly, the information interaction device provided in the embodiment of the present disclosure can generally be set in the server 105. The information interaction method provided in the embodiment of the present disclosure can also be executed by a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105. Accordingly, the information interaction device provided in the embodiment of the present disclosure can also be set in a server or server cluster that is different from the server 105 and can communicate with the terminal devices 101, 102, 103 and / or the server 105.

[0031] Alternatively, the information interaction method provided in the embodiment of the present disclosure may also be executed by the terminal device 101, 102 or 103. Accordingly, the information interaction apparatus provided in the embodiment of the present disclosure may also be arranged in the terminal device 101, 102 or 103.

[0032] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0033] Figure 2 The following schematically shows a schematic diagram of the system architecture of an intelligent agent according to an embodiment of the present disclosure.

[0034] like Figure 2 As shown, the embodiment of the present disclosure refers to the MOE (Mixture of Experts, MoE) mixed expert model structure to design the MOE-Agent framework. The MOE-Agent framework includes: pre-processing 210, planning brain 220, actuator 230, expert module 240, memory module 250, post-processing 260, special tool 270 and other components.

[0035] According to the embodiment of the present disclosure, the pre-processing 210 is responsible for uniformly processing data query functions that do not rely on other computational amounts, as well as some context cleaning functions, and preparing data that may be used by other modules in advance; the planning brain 210, as the core of the intelligent body, is responsible for overall task planning and scheduling, and determines the tasks that need to be performed; the executor 230 performs specific tasks according to the instructions of the planning brain, calls the corresponding expert module, executes the corresponding expert task and outputs batch subsequent reply results. The expert module 240 is a plurality of groups of specialized neural networks, each expert is responsible for processing a specific type of task or data. In the embodiment of the present disclosure, according to the business field, the expert module 240 includes marketing experts, answer experts, recommendation experts and general experts.

[0036] According to an embodiment of the present disclosure, the memory module 250 is responsible for storing and managing the information obtained by the agent during the interaction with the user. The memory module is generally divided into two parts: short-term memory and long-term memory. Among them, short-term memory is also called working memory, which stores the agent's recent sensory input, goals, and intermediate reasoning results. Short-term memory plays a key role in the decision-making process of the agent. It works closely with other functional modules of the agent to continuously update and process information to support the agent to make appropriate decisions and actions in different scenarios. The capacity of short-term memory is limited, and it can usually only hold a small amount of information, and the retention time is short, usually between a few seconds and a few minutes. Long-term memory stores the knowledge, experience, user characteristics and other information accumulated by the agent during the long-term operation. The capacity of long-term memory is large, and it can store a large amount of information, and it can be retained for a long time, and it can even be permanent. Long-term memory can be further subdivided into procedural memory, semantic memory and situational memory, among which procedural memory stores the rule set and program of the agent's behavior, involving the formation of skills and habits; semantic memory stores basic knowledge and world knowledge, including concepts, facts and language information. Situational memory stores the agent's past historical event stream or behavioral experience, involving memories of personal experiences and specific events. Long-term memory enables the agent to learn from experience, accumulate knowledge, and retrieve this information when needed to support decision-making and action. The memory module works in conjunction with other modules of the agent through operations such as memory reading, writing, and management, and together constitutes the cognitive architecture of the agent. In practical applications, the implementation of the memory module 250 involves database technology, vector database, natural language processing and other technologies to ensure the effective storage, rapid retrieval and accurate use of information. Through such a memory mechanism, the merchant agent can provide more personalized, coherent and efficient services.

[0037] In the disclosed embodiment, post-processing 260 is used to integrate and optimize the output of the expert module to generate a final response. Special tools 270 are a set of tools that provide specific functions or services and can be called by agents to support task execution. Special tools 270 specifically include local tools and remote tools, among which local tools are configured and generated based on prompt templates, basic large models and retrieval enhancement generation technology, and specifically include marketing dialogues, question answering, general answers, card recommendations and other tools; remote tools are implemented through remote service calls, including intent recognition, business point recognition, question rewriting, user portraits and business point portrait acquisition, etc.

[0038] According to the embodiments of the present disclosure, in order to better meet the requirements of framework flexibility, performance, iteration efficiency and other aspects in the commercial intelligent agent scenario, the MOE intelligent agent framework is designed with reference to the MOE large model architecture in the embodiments of the present disclosure. In terms of performance, the intelligent agent framework of the embodiment of the present disclosure uses the method of large model planning + multi-tool calling to reasonably plan the tasks and divert them to different expert tools, such as marketing experts, answering experts and recommendation experts, thereby reducing the access to general experts and the access frequency of the large model, thereby reducing the overall system response; in addition, by building cache capabilities covering all stages of the funnel such as planning, tools and end-to-end, the generation of high-quality questions and answers is accelerated through the reasonable use of cache, the resource consumption of large models is reduced, and the first-round response and long-tail time are effectively improved; through the planning stage, fast thinking, slow thinking and zero thinking are flexibly supported, among which, if it is determined that a reply is required in the user query, fast thinking is performed; slow thinking is when the intelligent agent pre-generates the speech when the user is silent during the conversation, for example, the marketing conversation expert pre-generates the guiding speech based on the above conversation to promote conversion; solves the performance problems caused by a single thinking mode and achieves a balanced benefit of performance and effect.

[0039] In terms of iteration cost, the framework abstracts the main components of the intelligent agent, such as planning brain, tool management, memory, reflection, etc. The business can quickly build the intelligent agent based on the framework, pay more attention to effect tuning, and reduce the iteration cost. Each component itself supports tuning and secondary development at the same time. For example, the planning stage can support users to use different large models and support prompt command tuning, increasing the flexibility of the MOE intelligent agent.

[0040] The following will be combined Figure 3~Figure 5 The information interaction method provided by the present disclosure is described in detail.

[0041] Figure 3 The flowchart of the information interaction method according to the embodiment of the present disclosure is schematically shown.

[0042] like Figure 3 As shown, the method 300 includes operations S310 to S340.

[0043] In operation S310, in response to the received dialogue request, the dialogue request is processed using a large language model to obtain a task function for the dialogue request.

[0044] According to the disclosed embodiment, the merchant agent is a new form of user interaction, with a dialogue robot as the main page carrier, providing services around the clock and accepting user consultations. Currently, as the digital employee of the advertiser, the merchant agent needs to take into account the user experience and customer conversion effect in various complex marketing dialogue scenarios under the premise of low response time. Solving user problems by calling tools through task planning is the core capability of the agent.

[0045] According to an embodiment of the present disclosure, when the intelligent agent receives a user dialogue request, it obtains information input by the user, namely dialogue input information. The dialogue input information can be information in multiple formats, for example, it can be one or more of text, audio, picture, video, and chart formats.

[0046] According to the embodiment of the present disclosure, before the planning brain of the intelligent agent performs unified task planning for the dialogue request, the pre-processing module in the intelligent agent queries the object portrait information according to the dialogue input information and the object identifier, wherein the object portrait information includes the object material information and configuration information. The object portrait information can be used in the subsequent object material retrieval stage and the response speech generation stage.

[0047] According to the embodiments of the present disclosure, the object profile information includes customer introduction information, such as basic customer information or business-related information. Configuration information includes data such as customer configuration and dialogue configuration. These configurations help the intelligent agent to communicate effectively with the user according to the preset processes and rules when interacting with the user. Pre-processing can reduce multiple queries or multiple generation of the same data, and reduce the frequency of access or call.

[0048] According to an embodiment of the present disclosure, in response to a received dialogue request, the planning brain of the intelligent agent performs unified task planning for the dialogue request, determines the task function for the dialogue request, and the large model uniformly initiates end-to-end task planning to ensure the accuracy of the overall task.

[0049] In operation S320, at least one candidate reply information is generated by executing the task function and calling the expert model.

[0050] According to an embodiment of the present disclosure, task function types correspond one-to-one to expert model types. The expert model is a domain-specific model pre-trained according to the task type. The expert model includes at least one of a marketing expert model, a solution expert model, a recommendation expert model, and a general expert model.

[0051] According to the embodiments of the present disclosure, domain-specific models focus on dialogue processing in specific fields, with faster response time and more accurate professional field capabilities. These specialized models are implemented by tailoring and optimizing large basic models, and can switch different specialized models to respond according to the user's context, thereby providing more efficient multi-round dialogue management capabilities.

[0052] According to the embodiments of the present disclosure, by executing the task function to call the corresponding expert model, after executing the speech generation task, a batch of candidate reply results are generated. A single round of request only executes one tool call at most, so that the agent response time is controllable, improving the accuracy and response speed of the dialogue.

[0053] In operation S330, the evaluation model is used to filter out reply information from the candidate reply information.

[0054] In operation S340, reply information is output.

[0055] According to the embodiment of the present disclosure, for the batch candidate reply results generated above, a series of business rule filtering plug-ins are called to filter the reply results that do not meet the customer or product requirements. The conversion effect estimation model of the historical dialogue request is used to estimate the conversion effect of each candidate reply result, and the top 1 result is selected as the output information based on the conversion effect ranking, and the output information is displayed to the user.

[0056] According to the embodiments of the present disclosure, the received dialogue requests are diverted to different expert models through unified planning of a large model and collaboration of multiple expert models. The activation and routing of multiple business experts are realized through the planning brain, thereby achieving more accurate task planning and more flexible effect tuning, and reducing the overall response time of the system.

[0057] Figure 4 The flowchart of the task function determination process according to the embodiment of the present disclosure is schematically shown.

[0058] like Figure 4 As shown, according to an embodiment of the present disclosure, in response to a received dialogue request, processing the dialogue request using a large language model to obtain a task function for the dialogue request may include operations S411 to S412.

[0059] In operation S411, the object material information is retrieved according to the dialogue input information.

[0060] According to an embodiment of the present disclosure, RAG (Retrieval-Augmented Generation) is an artificial intelligence technology that combines a retrieval model and a generation model. It retrieves relevant information from a large data set or knowledge base and uses this information to assist in text generation, thereby improving the accuracy and relevance of the output. RAG technology is particularly suitable for scenarios that require processing large amounts of data and complex queries, such as question-answering systems, content generation, and intelligent assistants.

[0061] According to an embodiment of the present disclosure, based on the RAG technology, object material information matching the dialogue is retrieved and generated according to the dialogue input information, thereby improving the accuracy and relevance of the generated object material information. The object material information includes the current customer profile information and question and answer corpus information, and the object material information is used to characterize the business information related to the dialogue request intention.

[0062] According to an embodiment of the present disclosure, the dialogue input information is preprocessed, and the preprocessing includes at least one of semantic understanding, intent recognition, business recognition, and input information vectorization; the object material information matching the dialogue input information is retrieved from the vector database. Since vectorization takes a long time, in order to improve processing efficiency, the dialogue input information vectorization operation is performed in parallel with intent recognition in the preprocessing stage.

[0063] In operation S412, the dialogue input information and the object material information are processed using the large language model to determine the task function.

[0064] According to an embodiment of the present disclosure, the large language model includes a task planning large model. Using the large model to process the dialogue input information and the object material information to determine the task function includes: obtaining the object historical dialogue information; inputting the object portrait information, the question and answer corpus information, the object historical dialogue information and the dialogue input information into the task planning large model, inferring and generating the name of the task function to be executed; and determining the task function according to the name of the task function to be executed.

[0065] According to the embodiment of the present disclosure, the retrieved customer material information, the historical conversation of this round, and the current user message are input to request the task planning model to infer the function name of the task to be executed in this round. The task planning model initiates end-to-end task planning in a unified manner, thereby ensuring the accuracy of the overall task.

[0066] According to the embodiments of the present disclosure, there are currently four types of functions; if the generated function is an invalid filtering type, it indicates that the customer's business does not match the user's needs, and a clarification script is generated to directly reply and end the current round of dialogue in advance, otherwise an expert call is executed to generate corresponding reply scripts.

[0067] Figure 5 The flowchart of the reply information screening process according to the embodiment of the present disclosure is schematically shown.

[0068] According to an embodiment of the present disclosure, using the evaluation model to filter out reply information from candidate reply information may include operations S531 to S533.

[0069] In operation S531, candidate reply information is filtered based on a preset rule.

[0070] According to the embodiments of the present disclosure, the number of generated candidate reply information can be configured and adjusted according to actual effects. For example, in a marketing scenario, a marketing expert generates 15 candidate reply results at a time and determines the optimal reply result from multiple candidate reply information; because in a question-and-answer scenario, users usually expect to get a clear and specific reply, at this time the answering expert and the recommending expert generate 1 candidate reply result.

[0071] According to an embodiment of the present disclosure, for batch candidate reply results generated upstream, a series of business rule filtering plug-ins are called to filter reply results that do not meet customer or product requirements.

[0072] According to an embodiment of the present disclosure, in the process of calling a series of business rule filtering plug-ins to filter the reply results that do not meet the customer or product requirements, the business rules involved include but are not limited to: Compliance rules: ensure that all reply contents comply with relevant laws and regulations, such as data protection regulations, copyright laws, etc. Quality control rules: ensure the accuracy, relevance and professionalism of the reply, and avoid providing wrong or misleading information. Customer service rules: according to customer service standards, ensure that the reply meets the expectations and needs of customers and provides timely and effective help. Product information rules: ensure that the product information provided is accurate, including price, specifications, availability, etc. Security rules: protect customer data security, do not disclose sensitive information, and comply with relevant laws and regulations such as the Data Security Law. Language and cultural rules: ensure that the reply content respects cultural differences and uses appropriate language and expressions. Marketing and promotion rules: ensure that marketing information is accurate, comply with promotion activity rules, and do not exaggerate or mislead consumers. Risk management rules: assess the risks that may be brought about by the reply content, such as legal risks, reputation risks, etc., and formulate corresponding preventive measures. Interaction design rules: ensure that the reply content is consistent with the user interaction design principles and provide a good user experience.

[0073] In operation S532, the conversion effect of the candidate reply information is evaluated based on the evaluation model.

[0074] In operation S533, reply information is determined according to the evaluation result.

[0075] According to the embodiments of the present disclosure, the chat continuation rate and conversion rate of each candidate reply information are determined; based on the chat continuation rate and conversion rate, the evaluation result of the candidate reply information is determined, wherein the chat continuation rate and conversion rate are used to characterize the conversion effect of the reply information. Specifically, the customer introduction information, the above information, the conversation characteristics, the time characteristics and the reply information obtained by the pre-processing module are used as input information, organized into prompt instructions for conversion effect identification, and input into the large model for inference calculation to obtain the chat continuation rate score and conversion rate score of each candidate reply information, and the evaluation result is determined according to the weighted score of the chat continuation rate score and the conversion rate score. By estimating the conversion effect of each candidate reply information, the best one is replied to the user, and the Top 1 result is selected as the final reply information based on the conversion effect ranking, and replied to the user. From the business perspective, the batch reply result set produced upstream is evaluated, and the Top 1 high-quality result is selected for reply, so as to approach the optimal reply for the business and improve the user experience.

[0076] According to the embodiments of the present disclosure, various components of the intelligent body can be tuned according to the actual conversion effect of the generated reply speech to reduce the iteration cost.

[0077] It can be understood that the method of the present disclosure is described above, and the device of the present disclosure will be described below.

[0078] Figure 6 It is a block diagram of an information interaction device according to an embodiment of the present disclosure.

[0079] like Figure 6 As shown, the device 600 may include a task planning module 610 , a generation module 620 , a screening module 630 and a control module 640 .

[0080] The task planning module 610 is used to respond to the received dialogue request and process the dialogue request using the large language model to obtain a task function for the dialogue request.

[0081] The generating module 620 is used to call the expert model by executing the task function to generate at least one candidate reply information.

[0082] The screening module 630 is used to screen out reply information from candidate reply information by using the evaluation model.

[0083] The output module 640 is used to output reply information.

[0084] In some embodiments, the task planning module includes: a retrieval submodule, used to retrieve object material information based on dialogue input information, and the object material information is used to characterize business information related to the dialogue request intent; a task function determination submodule, used to use a large language model to process the dialogue input information and object material information to determine the task function.

[0085] In some embodiments, the large language model includes a task planning large model, and the task function determination submodule includes: an acquisition unit, used to acquire object historical dialogue information; an input unit, used to input object material information, object historical dialogue information and dialogue input information into the task planning large model, and infer and generate the name of the task function to be executed; and a first determination unit, used to determine the task function according to the name of the task function to be executed.

[0086] In some embodiments, the retrieval submodule includes: a preprocessing unit, used to preprocess the dialogue input information, the preprocessing including at least one of semantic understanding, intent recognition, business recognition and input information vectorization; a matching unit, used to retrieve object material information matching the dialogue input information in the vector database.

[0087] In some embodiments, the screening module includes: a filtering submodule for filtering candidate reply information based on preset rules; an evaluation submodule for evaluating the conversion effect of the candidate reply information based on an evaluation model; and a determination submodule for determining the reply information based on the evaluation results.

[0088] In some embodiments, the evaluation submodule includes: a second determination unit for determining the chat continuation rate and conversion rate of each candidate reply information; and a third determination unit for determining the evaluation result of the candidate reply information based on the chat continuation rate and the conversion rate.

[0089] In some embodiments, the task function type corresponds one-to-one to the expert model type. The expert model is a domain-specific model pre-trained according to the task type. The expert model includes at least one of a marketing expert model, a solution expert model, a recommendation expert model, and a general expert model.

[0090] In some embodiments, the apparatus further comprises:

[0091] The pre-processing module is used to query the object portrait information according to the dialogue input information and the object identification, wherein the object portrait information includes the object data information and the configuration information.

[0092] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0093] Figure 7 The structural block diagram of an intelligent agent according to an embodiment of the present disclosure is schematically shown.

[0094] In the embodiments of the present disclosure, inspired by the von Neumann structure in modern computer theory, such as Figure 7 As shown, the intelligent agent 700 may include five core modules: an input module 710 , a control module 720 , a storage module 730 , a calculation module 740 and an output module 750 .

[0095] The input module 710 is responsible for receiving or sensing information such as queries, requests, instructions, signals or data from the outside world (such as users or external environment), and converting them into a format that can be understood and processed by the intelligent agent 700. The input module 710 is the primary link for the intelligent agent 700 to interact with the outside world. It enables the intelligent agent 700 to efficiently and accurately obtain the necessary "sensory" information from the outside world and respond to this information.

[0096] In an example, the input module 710 may input the conversation request described above.

[0097] In the example, the control module 720 is the core support for the ability of the agent 700 to handle complex tasks. The control module 720 can execute the information interaction method described above.

[0098] In the example, the control module 720 will continuously interact with the storage module 730, the computing module 740 and / or the output module 750 during operation. However, it should be noted that in the embodiment of the present disclosure, the control module 720 acts as a single initiator to initiate communication with the storage module 730, the computing module 740 and / or the output module 750, and there is no communication coupling between the storage module 730, the computing module 740 and the output module 750.

[0099] In the example, the performance of the control module 720 may be closely related to the large model based on the agent 700. In order to give full play to the capabilities of the large language model, the internal structure of the control module 720 may be designed to be highly configurable and extensible to cope with various types of tasks and requirements in real scenarios.

[0100] The storage module 730 may be responsible for memorizing information such as historical conversations, etc. The object historical conversation information and object material information as mentioned above may be included in the storage module 730 .

[0101] In the example, after the agent 700 obtains the dialogue request, the agent 700 can use the intention recognition model to determine the user intention from the dialogue input information. The agent 700 can retrieve the data resources related to the user intention from the storage module 730 and feed it back to the control module 720. Then, the control module 720 can use the large language model to process the dialogue request, obtain the task function for the dialogue request, call the expert model by executing the task function, generate at least one candidate reply information, and pass the reply information to the output module 750.

[0102] The operation module 740 can be regarded as a predefined tool library. The above-mentioned renderer and display control can be included in the operation module 740.

[0103] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0104] In some embodiments, an electronic device includes: at least one processor; and a memory connected to the at least one processor in communication. The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform a method provided in accordance with the present disclosure (e.g., the above-mentioned method 200).

[0105] In some embodiments, a non-transitory computer-readable storage medium stores computer instructions, where the computer instructions are used to enable a computer to execute a method provided according to the present disclosure (eg, the above-mentioned method 200).

[0106] In some embodiments, a computer program product includes a computer program, and when the computer program is executed by a processor, the computer program implements the method provided according to the present disclosure (such as the above method 200). Figure 8 Provide further explanation.

[0107] Figure 8 A schematic block diagram of an example electronic device 800 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0108] like Figure 8 As shown, the device 800 includes a computing unit 801, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 802 or a computer program loaded from a storage unit 808 to a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0109] A number of components in the device 800 are connected to the I / O interface 805, including: an input unit 806, such as a keyboard, a mouse, etc.; an output unit 807, such as various types of displays, speakers, etc.; a storage unit 808, such as a disk, an optical disk, etc.; and a communication unit 809, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 809 allows the device 800 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0110] The computing unit 801 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 801 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSP), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 801 performs the various methods and processes described above, such as the information interaction method. For example, in some embodiments, the information interaction method may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 808. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 800 via the ROM 802 and / or the communication unit 809. When the computer program is loaded into the RAM 803 and executed by the computing unit 801, one or more steps of the information interaction method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to execute the information interaction method in any other appropriate manner (eg, by means of firmware).

[0111] Various embodiments of the systems and techniques described above herein may be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard parts (ASSPs), system on chip systems (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0112] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0113] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. More specific examples of machine-readable storage media may include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories, read-only memories, erasable programmable read-only memories (EPROM) or flash memories, optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0114] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a cathode ray tube (CRT) display or a liquid crystal display (LCD)) for displaying information to the user; and a keyboard and a pointing device (e.g., a mouse or a trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0115] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a Local Area Network (LAN), a Wide Area Network (WAN), and the Internet.

[0116] A computer system may include clients and servers. Clients and servers are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship to each other.

[0117] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0118] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An information interaction method, comprising: In response to the received dialogue request, the dialogue request is processed using the large language model to obtain a task function for the dialogue request; By executing the task function and calling the expert model, at least one candidate reply information is generated; Using the evaluation model, filter out reply information from the candidate reply information; as well as The reply information is output.

2. The method according to claim 1, wherein: The process of processing the dialogue request by using the large language model to obtain a task function for the dialogue request includes: Retrieving target material information according to the dialogue input information, wherein the target material information is used to represent business information related to the dialogue request intention; and The dialogue input information and the object material information are processed using the large language model to determine a task function.

3. The method according to claim 2, wherein: The large language model includes a task planning large model, and the use of the large model to process the dialogue input information and the object material information to determine the task function includes: Get the object's historical conversation information; Inputting the object material information, the object historical dialogue information and the dialogue input information into the task planning model, and inferring and generating the function name of the task to be executed; and The task function is determined according to the name of the task function to be executed.

4. The method according to claim 2, wherein: The object material information retrieval according to the dialogue input information includes: Preprocessing the dialogue input information, wherein the preprocessing includes at least one of semantic understanding, intent recognition, business recognition, and input information vectorization; and The target material information matching the dialogue input information is retrieved from the vector database.

5. The method according to claim 1, wherein: The using the evaluation model to filter out the reply information from the candidate reply information includes: Filtering the candidate reply information based on preset rules; Evaluate the conversion effect of candidate response information based on the evaluation model; and Determine the response information based on the evaluation results.

6. The method according to claim 5, wherein: The evaluation of the conversion effect of the candidate reply information based on the evaluation model includes: Determine the follow-up rate and conversion rate of each candidate response message; and An evaluation result of the candidate reply information is determined according to the chat continuation rate and the conversion rate.

7. The method according to any one of claims 1 to 6, wherein: The task function type corresponds to the expert model type one by one. The expert model is a domain-specific model pre-trained according to the task type. The expert model includes at least one of a marketing expert model, a solution expert model, a recommendation expert model and a general expert model.

8. The method according to claim 7, wherein: Before the dialog request is processed by using the large language model to obtain a task function for the dialog request, the method further includes: The object portrait information is queried according to the dialogue input information and the object identification, wherein the object portrait information includes object profile information and configuration information.

9. An information interaction device, comprising: A task planning module, used for responding to a received dialogue request, processing the dialogue request using a large language model, and obtaining a task function for the dialogue request; A generating module, configured to generate at least one candidate reply information by calling the expert model by executing the task function; A screening module, used to screen out reply information from the candidate reply information by using an evaluation model; An output module is used to output the reply information.

10. The device according to claim 9, wherein: The mission planning module includes: A retrieval submodule, used to retrieve object material information according to the dialogue input information, wherein the object material information is used to represent business information related to the dialogue request intention; and The task function determination submodule is used to use the large language model to process the dialogue input information and the object material information to determine the task function.

11. The device according to claim 10, wherein: The large language model includes a task planning large model, and the task function determination submodule includes: An acquisition unit, used to acquire historical conversation information of an object; An input unit, used to input the object material information, the object historical dialogue information and the dialogue input information into the task planning model, and infer and generate a function name of the task to be executed; and The first determining unit is used to determine the task function according to the name of the task function to be executed.

12. The device according to claim 10, wherein: The retrieval submodule includes: a preprocessing unit, configured to preprocess the dialogue input information, wherein the preprocessing includes at least one of semantic understanding, intent recognition, service recognition, and input information vectorization; and The matching unit is used to retrieve object material information matching the dialogue input information in the vector database.

13. The device according to claim 9, wherein: The screening module comprises: A filtering submodule, used to filter the candidate reply information based on preset rules; An evaluation submodule, for evaluating the conversion effect of the candidate reply information based on the evaluation model; and The determination submodule is used to determine the reply information according to the evaluation result.

14. The device according to claim 13, wherein: The evaluation submodule includes: A second determining unit is used to determine the chat continuation rate and conversion rate of each candidate reply information; and The third determining unit is used to determine the evaluation result of the candidate reply information according to the chat continuation rate and the conversion rate.

15. The device according to any one of claims 9 to 14, wherein: The task function type corresponds to the expert model type one by one. The expert model is a domain-specific model pre-trained according to the task type. The expert model includes at least one of a marketing expert model, a solution expert model, a recommendation expert model and a general expert model.

16. The device according to claim 15, wherein: Also includes: The pre-processing module is used to query the object portrait information according to the dialogue input information and the object identification, wherein the object portrait information includes the object data information and the configuration information.

17. An artificial intelligence agent configured to perform the method according to any one of claims 1 to 8.

18. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 8.

19. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 8.

20. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 8.

Images