A hybrid interactive system and method based on AI big model

Abstract

The present invention provides a hybrid interactive system and method based on an AI big model, which combines traditional natural language processing NLP technology with AI big model technology. Simple instructions are processed by traditional NLP technology, and complex instructions are processed by AI big models, which not only ensures a quick response to simple instructions, but also improves the accuracy of understanding complex instructions; the AI ​​big model is connected to an external interface, and the AI ​​big model is responsible for understanding user instructions and determining the scheduling plan of external tools. The external tools are responsible for returning real-time and accurate knowledge to the AI ​​big model, and then the AI ​​big model summarizes and concludes based on this knowledge, and finally returns real and reliable results to the user, thereby reducing the illusion phenomenon of the AI ​​big model. The system and method provided by the present invention can reduce deployment costs while ensuring user experience.

Description

Technical Field

[0001] The present invention belongs to the technical field of human-computer interaction, and in particular relates to a hybrid interaction system and method based on an AI big model. Background Art

[0002] Human-computer interaction technology is an important branch of computer science. It studies the interaction between humans and computers, including human input of commands to computers, computer feedback output to humans, and information exchange between humans and computers. With the development of computer technology, human-computer interaction technology is also constantly improving: from the initial command line interaction, to the graphical user interface, and then to the current natural language interaction, the human-computer interaction method is getting closer and closer to the interaction method between people, making the interaction between people and computers more natural and convenient.

[0003] Natural Language Processing (NLP) technology is one of the key technologies to achieve natural language interaction between humans and computers. It is an important branch of computer science and artificial intelligence. It mainly studies the use of computers to simulate the human language communication process, so that computers can understand and use the natural language used in people's lives, and realize natural language communication between humans and computers, thereby further realizing the goal of computers replacing people in some mental work. Among them, some mental work mainly includes searching for information, answering questions, extracting literature, compiling information, and all processing related to natural language information.

[0004] Traditional NLP technology mainly understands natural language based on rules or neural networks with fewer parameters. It has a high understanding accuracy for simple instructions, fast reasoning speed and low deployment cost. However, it is more complicated to process complex instructions and has a low understanding accuracy.

[0005] With the development of deep learning technology, AI big model technology has emerged. AI big model is based on large-scale neural network parameters and training data training, with strong modeling ability and strong ability to understand human natural language. It has obvious advantages over traditional NLP technology in tasks such as logical reasoning, task scheduling, multi-intention understanding, and reference resolution, but its reasoning speed is slow and the deployment cost is high.

[0006] However, both traditional NLP technology and AI big model technology have some problems. Although traditional NLP technology has advantages in processing simple instructions, it is weak in processing complex instructions; and although AI big models have advantages in processing complex instructions, their reasoning speed is slow and the deployment cost is high. Moreover, since their output is based on probability distribution, it is inevitable that inaccurate answers will appear for knowledge that is not in the training set or real-time content. When the text generated by the big model does not follow the original text (Faithfulness) or does not conform to the facts (Factualness), we can think that the model has "big model hallucination" (LLM Hallucination).

[0007] Therefore, how to process simple instructions to reduce latency, accurately process complex instructions, reduce the hallucination phenomenon of large AI models, and reduce deployment costs is an important issue facing current human-computer interaction technology. Summary of the invention

[0008] In response to the problems existing in the prior art, the present invention combines traditional NLP technology, AI big model technology and external interface to form a new hybrid interaction system and method, which not only solves the respective problems of traditional NLP technology and AI big model technology, but also realizes real-time updating of knowledge, while reducing the deployment cost of AI big models. It has high practical value and broad application prospects.

[0009] The present invention provides a hybrid interactive system based on an AI large model, characterized by comprising:

[0010] The acquisition module is used to obtain user requests;

[0011] A natural language processing (NLP) module is used to process the user request and generate an NLP processing result;

[0012] A confidence evaluation module, used to determine whether the confidence of the NLP processing result output by the natural language processing (NLP) module meets a set threshold;

[0013] The AI ​​big model module is used to deeply understand and process the user request when the confidence level does not meet the threshold, and generate an external interface scheduling plan according to the understanding result; and after the external interface module returns the required information according to the external interface scheduling plan, summarize the acquired information and generate the final processing result;

[0014] An external interface module, used to obtain real-time and accurate information based on the external interface scheduling plan generated by the AI ​​large model module, and return it to the AI ​​large model module;

[0015] An output module is used to output the processing results of the system. If the confidence of the NLP processing result in the natural language processing NLP module is greater than a set threshold, the output module outputs the NLP processing result to the user; if the confidence is lower than the threshold, the output module outputs the final processing result generated by the AI ​​large model module to the user.

[0016] Furthermore, the external interface scheduling plan includes the external interface to be called and parameter information required for parsing the interface.

[0017] Furthermore, during the interaction with the user, the natural language processing NLP module and the AI ​​large model module simultaneously maintain the interaction context with the user to ensure the consistency of the interaction.

[0018] The present invention provides a hybrid interaction method based on an AI large model, characterized in that it comprises the following steps:

[0019] Get user request;

[0020] Processing the user request using a natural language processing (NLP) method and generating an NLP processing result;

[0021] Determine whether the confidence of the NLP processing result meets the set threshold;

[0022] If the confidence level satisfies the threshold, outputting the NLP processing result to the user;

[0023] If the confidence level does not meet the threshold, the AI ​​big model is used to deeply understand and process the user request, and an external interface scheduling plan is generated according to the understanding result; real-time and accurate information is obtained according to the external interface scheduling plan, and returned to the AI ​​big model; the AI ​​big model summarizes the information obtained from the external interface, generates the final processing result, and outputs it to the user.

[0024] Furthermore, the external interface scheduling plan includes the external interface to be called and parameter information required for parsing the interface.

[0025] Furthermore, during the interaction with the user, the natural language processing (NLP) method and the AI ​​big model method simultaneously maintain the interaction context with the user to ensure the consistency of the interaction.

[0026] The present invention provides a hybrid interactive system based on an AI large model, characterized by comprising:

[0027] The acquisition module is used to obtain user requests;

[0028] A natural language processing (NLP) module is used to process the user request and generate an NLP processing result;

[0029] The AI ​​big model module is used to deeply understand and process the user request, and generate an external interface scheduling plan according to the understanding result; and after the external interface module returns the required information according to the external interface scheduling plan, summarize the acquired information and generate the AI ​​processing result;

[0030] A decision module, used for determining whether to output the NLP processing result generated by the natural language processing NLP module or the AI ​​processing result generated by the AI ​​big model module to the user;

[0031] An external interface module, used to obtain real-time and accurate information based on the external interface scheduling plan generated by the AI ​​large model module, and return it to the AI ​​large model module;

[0032] The output module is used to output the processing result of the system according to the judgment of the decision module.

[0033] Furthermore, the judgment basis of the decision module is: if the confidence of the NLP processing result meets a preset threshold and the return result of the AI ​​large model module is not received within a predetermined time, the NLP processing result is output to the user; otherwise, the AI ​​processing result is output to the user.

[0034] Furthermore, the external interface scheduling plan includes the external interface to be called and parameter information required for parsing the interface.

[0035] Furthermore, during the interaction with the user, the natural language processing NLP module and the AI ​​large model module simultaneously maintain the interaction context with the user to ensure the consistency of the interaction.

[0036] The present invention provides a hybrid interaction method based on an AI large model, characterized in that it comprises the following steps:

[0037] Get user request;

[0038] Using a natural language processing (NLP) method to process the user request and obtain an NLP processing result;

[0039] The user request is deeply understood and processed by the AI ​​big model, and an external interface scheduling plan is generated according to the understanding result; real-time and accurate information is obtained according to the external interface scheduling plan, and returned to the AI ​​big model; the AI ​​big model summarizes the information obtained by the external interface and generates an AI processing result; if the AI ​​big model fails to complete the processing within the predetermined time, the processing is terminated;

[0040] If the confidence of the NLP processing result meets a preset threshold and the AI ​​big model does not generate the AI ​​processing result within a predetermined time, the NLP processing result is output to the user; otherwise, the AI ​​processing result is output to the user.

[0041] Furthermore, the external interface scheduling plan includes the external interface to be called and parameter information required for parsing the interface.

[0042] Furthermore, during the interaction with the user, the natural language processing (NLP) method and the AI ​​big model method simultaneously maintain the interaction context with the user to ensure the consistency of the interaction.

[0043] The present invention provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of implementing the above method are realized.

[0044] The present invention provides an electronic device, comprising:

[0045] one or more processors; and

[0046] A memory associated with the one or more processors, the memory is used to store program instructions, and when the program instructions are read and executed by the one or more processors, the steps of the aforementioned method are executed.

[0047] Compared with the prior art, the present invention has the following advantages and positive effects:

[0048] 1. Improved response speed and understanding accuracy: By combining traditional NLP technology and AI big model technology, simple instructions are processed by traditional NLP technology, and complex instructions are processed by AI big models, which not only ensures a quick response to simple instructions, but also improves the understanding accuracy of complex instructions.

[0049] 2. Reduced deployment costs: Traditional NLP technology has low deployment costs, while large AI models have high deployment costs. By combining the two, it is possible to reduce deployment costs while meeting the needs of processing complex instructions.

[0050] 3. Reduced the hallucination phenomenon of AI big models: By connecting AI big models with external interfaces, the AI ​​big models are responsible for understanding user instructions and determining the scheduling plan of external tools, and the external tools are responsible for returning real-time and accurate knowledge to the AI ​​big models. The AI ​​big models then summarize and conclude based on this knowledge, and finally return true and reliable results to users, thereby reducing the hallucination phenomenon of AI big models.

[0051] 4. Flexible framework shortens the development cycle: The method proposed in the present invention makes use of the powerful understanding ability of natural language of the AI ​​big model, and can replace the definition of the external interface without retraining, thereby realizing the rapid update of services and knowledge, shortening the development and launch cycle, and reducing costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] For a more complete understanding of the present invention, reference will now be made to the following description taken in conjunction with the accompanying drawings, in which:

[0053] Figure 1 A schematic diagram of a system according to Embodiment 1 of the present invention;

[0054] Figure 2 This is a flow chart of the method described in Embodiment 2 of the present invention;

[0055] Figure 3 is a schematic diagram of a system according to Embodiment 3 of the present invention;

[0056] Figure 4 Schematic diagram of the process of the method described in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0057] In order to make the problems, technical solutions and technical effects to be solved by the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments and obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0058] Before describing the exemplary embodiments of the present invention in more detail, it should be noted that although the flowchart of the present invention describes the operations as a sequential processing mode, many of the operation steps can be implemented in parallel or simultaneously, and the order of the operations can also be rearranged as needed. In addition, the processing can be terminated when the operation is completed, but there can also be additional steps not included in the drawings.

[0059] In the following description, the terms "first, second, third", etc. are merely used to distinguish similar objects and do not represent a specific ordering of the objects. It is understandable that "first, second, third", etc. can be interchanged with a specific order or sequence where permitted, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein.

[0060] Embodiment 1:

[0061] The present invention provides a hybrid interactive system based on an AI large model, see Figure 1 The system includes an acquisition module 101, a natural language processing NLP module 102, a confidence evaluation module 103, an AI large model module 104, an external interface module 105, and an output module 106.

[0062] Wherein, the acquisition module 101 is used to acquire a user request.

[0063] The user request may be input in the form of text, audio, video, etc.; and may be in the form of a sentence, paragraph, article, etc.

[0064] The natural language processing (NLP) module 102 is used to process the user request and generate an NLP processing result.

[0065] The processing tasks that the natural language processing NLP module 102 may complete include translation, understanding text content, establishing a database, generating summaries, and communicating with users on a computer interface. The output processing results have different forms depending on the task. For example, if the task is speech recognition, the output is a sentence or keyword that matches the speech; if it is a question-answering system, the output is an answer that is close to natural language; if it is a language translation, the output is a word, sentence or paragraph in another language; and so on.

[0066] The algorithm used by the natural language processing NLP module 102 is not limited in the present invention, and a traditional NLP module can be used.

[0067] The confidence evaluation module 103 is used to determine whether the confidence of the NLP processing result output by the natural language processing NLP module 102 meets a set threshold.

[0068] The threshold can be set according to actual needs.

[0069] The AI ​​big model module 104 is used to deeply understand and process the user request when the confidence level does not meet the threshold, and generate an external interface scheduling plan based on the understanding result; and after the external interface module 105 returns the required information according to the external interface scheduling plan, summarize the obtained information and generate the final processing result.

[0070] The external interface scheduling plan needs to include the external interface to be called and parameter information required for parsing the interface.

[0071] If the confidence level of the NLP processing result output by the natural language processing NLP module 102 is less than the set threshold, it means that the natural language processing NLP module 102 cannot accurately understand and process the user request, and then the user request is passed to the AI ​​large model module 104 for processing.

[0072] AI big models have strong natural language understanding capabilities for humans, but they still have limitations that cannot be solved. These limitations include: inability to obtain the latest information and tendency to hallucinate, difficulty understanding low-resource languages, lack of precise mathematical calculation capabilities, inability to realize the passage of time, etc. To overcome these limitations, one possible option is to give it the ability to use external tools.

[0073] After the AI ​​big model module 104 deeply understands the user request, it generates an external interface scheduling plan based on the understanding result, determines the external interface to be called, and analyzes the parameter information required by the interface. The scheduling plan of the external interface is generated based on the existing algorithm in the prior art, and the present invention does not limit this.

[0074] After the external interface returns the information required for the plan to the AI ​​large model module 104 according to the external interface scheduling plan, the AI ​​large model module 104 summarizes the acquired information to generate the final processing result.

[0075] The external interface module 105 is used to obtain real-time and accurate information based on the external interface scheduling plan generated by the AI ​​large model module 104, and return it to the AI ​​large model module 104.

[0076] The output module 106 is used to output the processing result of the system. If the confidence of the NLP processing result in the natural language processing NLP module 102 is greater than a set threshold, the output module 106 outputs the NLP processing result to the user; if the confidence is lower than the threshold, the output module 106 outputs the final processing result generated by the AI ​​large model module 104 to the user.

[0077] In the system disclosed in this embodiment, the traditional NLP module is responsible for processing simple user requests, the AI ​​large model module is responsible for processing complex user requests, and the external interface module is responsible for providing real-time and accurate information to eliminate the illusion phenomenon of the AI ​​large model. During the interaction process, the traditional NLP module and the AI ​​large model module need to maintain the interaction context with the user at the same time to ensure the coherence of the interaction.

[0078] Embodiment 2:

[0079] Corresponding to the first embodiment, the present invention provides a hybrid interaction method based on an AI large model, see Figure 2 , specifically including the following steps:

[0080] Step S01: Obtain user request.

[0081] The user request may be input in the form of text, audio, video, etc.; and may be in the form of a sentence, paragraph, article, etc.

[0082] Step S02: Process the user request using a natural language processing (NLP) method and generate an NLP processing result.

[0083] The processing tasks that may be completed by the natural language processing (NLP) method include translation, understanding text content, building a database, generating summaries, and communicating with users on a computer interface. The output processing results have different forms depending on the task. For example, if the task is speech recognition, the output is a sentence or keyword that matches the speech; if it is a question-answering system, the output is an answer that is close to natural language; if it is a language translation, the output is a word, sentence or paragraph in another language; and so on.

[0084] The algorithm used by the natural language processing (NLP) is not limited in the present invention.

[0085] Step S03, determine whether the confidence of the NLP processing result meets the set threshold; if it meets the threshold, output the NLP processing result to the user; otherwise, proceed to step S04.

[0086] The threshold can be set according to actual needs.

[0087] Step S04: When the confidence level does not meet the threshold, the AI ​​big model is used to deeply understand and process the user request, and an external interface scheduling plan is generated according to the understanding result.

[0088] The external interface scheduling plan needs to include the external interface to be called and parameter information required for parsing the interface.

[0089] If the confidence level of the NLP processing result output by the natural language processing (NLP) method is less than the set threshold, it means that the natural language processing (NLP) method cannot accurately understand and process the user request, and the user request is then passed to the AI ​​big model for processing.

[0090] AI big models have strong natural language understanding capabilities for humans, but they still have limitations that cannot be solved. These limitations include: inability to obtain the latest information and tendency to hallucinate, difficulty understanding low-resource languages, lack of precise mathematical calculation capabilities, inability to realize the passage of time, etc. To overcome these limitations, one possible option is to give it the ability to use external tools.

[0091] After the AI ​​big model deeply understands the user request, it generates an external interface scheduling plan based on the understanding result, determines the external interface to be called, and analyzes the parameter information required by the interface. The scheduling plan of the external interface is generated based on the existing algorithm in the prior art, and the present invention does not limit this.

[0092] Step S05: Acquire real-time and accurate information according to the external interface scheduling plan, and return it to the AI ​​big model.

[0093] Step S06: The AI ​​big model summarizes the information obtained by the external interface, generates the final processing result, and outputs it to the user.

[0094] In the method disclosed in this embodiment, the traditional NLP method is responsible for processing simple user requests, the AI ​​big model is responsible for processing complex user requests, and the external interface is responsible for providing real-time and accurate information to eliminate the illusion phenomenon of the AI ​​big model. During the interaction process, the traditional NLP method and the AI ​​big model need to maintain the interaction context with the user at the same time to ensure the coherence of the interaction.

[0095] Embodiment three:

[0096] The present invention provides a hybrid interactive system based on an AI large model, see Figure 3 The system includes an acquisition module 201, a natural language processing NLP module 202, an AI large model module 203, a decision module 204, an external interface module 205, and an output module 206.

[0097] The acquisition module 201 is used to acquire a user request.

[0098] The user request may be input in the form of text, audio, video, etc.; and may be in the form of a sentence, paragraph, article, etc.

[0099] The natural language processing (NLP) module 202 is used to process the user request and generate an NLP processing result.

[0100] The processing tasks that the natural language processing NLP module 202 may complete include translation, understanding text content, building a database, generating summaries, and communicating with users on a computer interface. The output processing results have different forms depending on the task. For example, if the task is speech recognition, the output is a sentence or keyword that matches the speech; if it is a question-answering system, the output is an answer that is close to natural language; if it is a language translation, the output is a word, sentence or paragraph in another language; and so on.

[0101] The algorithm adopted by the natural language processing (NLP) module 202 is not limited in the present invention.

[0102] The AI ​​large model module 203 is used to deeply understand and process the user request and generate an external interface scheduling plan based on the understanding results; and after the external interface module 205 returns the required information according to the external interface scheduling plan, summarize the obtained information and generate the AI ​​processing result.

[0103] The external interface scheduling plan needs to include the external interface to be called and parameter information required for parsing the interface.

[0104] AI big models have strong natural language understanding capabilities for humans, but they still have limitations that cannot be solved. These limitations include: inability to obtain the latest information and tendency to hallucinate, difficulty understanding low-resource languages, lack of precise mathematical calculation capabilities, inability to realize the passage of time, etc. To overcome these limitations, one possible option is to give it the ability to use external tools.

[0105] After the AI ​​big model module 203 deeply understands the user request, it generates an external interface scheduling plan based on the understanding result, determines the external interface to be called, and analyzes the parameter information required by the interface. The scheduling plan of the external interface is generated based on the existing algorithm in the prior art, and the present invention does not limit this.

[0106] After the external interface returns the information required for the plan to the AI ​​large model module 203 according to the external interface scheduling plan, the AI ​​large model module 203 summarizes the acquired information and generates an AI processing result.

[0107] The decision module 204 is used to determine whether to output the NLP processing result generated by the natural language processing NLP module 202 or the AI ​​processing result generated by the AI ​​large model module 203 to the user.

[0108] Among them, if the confidence of the NLP processing result output by the natural language processing NLP module 202 meets the preset threshold, and the return result of the AI ​​large model module 203 is not received within the predetermined time, the NLP processing result is output to the user. At this time, the natural language processing NLP module 202 can accurately understand and process the user request, and the delay of the AI ​​large model module 203 is large, then the NLP processing result is directly output to the user.

[0109] If the confidence of the NLP processing result does not meet the preset threshold, or the AI ​​large model module 203 returns the result within the predetermined time, the AI ​​processing result generated by the AI ​​large model module 203 is output to the user.

[0110] The external interface module 205 is used to obtain real-time and accurate information based on the external interface scheduling plan generated by the AI ​​large model module 203, and return it to the AI ​​large model module 203.

[0111] The output module 206 is used to output the processing results of the system.

[0112] The output module 206 outputs the NLP processing result or the AI ​​processing result to the user according to the judgment result of the decision module 204 .

[0113] In the system disclosed in this embodiment, the user request is sent to the traditional NLP and AI big model at the same time. If the confidence given by the traditional NLP module is greater than the set threshold and the AI ​​big model does not return a result within the specified time, the traditional NLP module is used for processing, otherwise the AI ​​big model is used for processing. The system improves the accuracy of understanding while ensuring that the system delay meets the requirements. Similarly, during the interaction process, the traditional NLP module and the AI ​​big model module need to maintain the interaction context with the user at the same time.

[0114] Embodiment 4:

[0115] Corresponding to the third embodiment, the present invention provides a hybrid interaction method based on an AI large model, see Figure 4 , specifically including the following steps:

[0116] Step S01: Obtain user request.

[0117] The user request may be input in the form of text, audio, video, etc.; and may be in the form of a sentence, paragraph, article, etc.

[0118] Step S02: Process the user request using a natural language processing (NLP) method and obtain an NLP processing result.

[0119] The processing tasks that may be completed by the natural language processing (NLP) method include translation, understanding text content, building a database, generating summaries, and communicating with users on a computer interface. The output processing results have different forms depending on the task. For example, if the task is speech recognition, the output is a sentence or keyword that matches the speech; if it is a question-answering system, the output is an answer that is close to natural language; if it is a language translation, the output is a word, sentence or paragraph in another language; and so on.

[0120] The algorithm used by the natural language processing (NLP) is not limited in the present invention.

[0121] Step S03: Use the AI ​​big model to deeply understand and process the user request, and generate an external interface scheduling plan based on the understanding results; obtain real-time and accurate information based on the external interface scheduling plan, and return it to the AI ​​big model; the AI ​​big model summarizes the information obtained by the external interface to generate AI processing results.

[0122] The external interface scheduling plan needs to include the external interface to be called and parameter information required for parsing the interface.

[0123] AI big models have strong natural language understanding capabilities for humans, but they still have limitations that cannot be solved. These limitations include: inability to obtain the latest information and tendency to hallucinate, difficulty understanding low-resource languages, lack of precise mathematical calculation capabilities, inability to realize the passage of time, etc. To overcome these limitations, one possible option is to give it the ability to use external tools.

[0124] After the AI ​​big model deeply understands the user request, it generates an external interface scheduling plan based on the understanding result, determines the external interface to be called, and analyzes the parameter information required by the interface. The scheduling plan of the external interface is generated based on the existing algorithm in the prior art, and the present invention does not limit this.

[0125] Step S04: If the confidence of the NLP processing result meets the preset threshold and the AI ​​big model does not generate the AI ​​processing result within the predetermined time, the NLP processing result is output to the user; if the confidence of the NLP processing result does not meet the preset threshold, or the AI ​​big model generates the AI ​​processing result within the predetermined time, the AI ​​processing result is output to the user.

[0126] In the method disclosed in this embodiment, the user request is sent to the traditional NLP and AI big model at the same time. If the confidence given by the traditional NLP module is greater than the set threshold and the AI ​​big model does not return a result within the specified time, the traditional NLP module is used for processing, otherwise the AI ​​big model is used for processing. The system improves the accuracy of understanding while ensuring that the system delay meets the requirements. Similarly, during the interaction process, the traditional NLP module and the AI ​​big model module need to maintain the interaction context with the user at the same time.

[0127] In addition, the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the steps of the methods described in the method embodiments such as Embodiment 2 and Embodiment 4 are implemented.

[0128] The present invention also provides an electronic device, comprising:

[0129] one or more processors; and

[0130] A memory associated with the one or more processors, the memory being used to store program instructions, and the program instructions, when read and executed by the one or more processors, execute the steps of the method described in the aforementioned method embodiments such as Embodiment 2 and Embodiment 4.

[0131] It should be noted that a part of the present invention can be applied as a computer program product, such as a computer program instruction, when it is executed by an intelligent electronic device (such as a smart mobile phone or a tablet computer, etc.), the method and / or technical solution according to the present invention can be called or provided through the operation of the intelligent electronic device. The program instruction for calling the method of the present invention may be stored in a fixed or removable recording medium, and / or transmitted through a data stream in a broadcast or other signal-bearing medium, and / or stored in a working memory of an intelligent electronic device that operates according to the program instruction. Here, according to an embodiment of the present invention, a device is included, the device including a memory for storing computer program instructions and a processor for executing program instructions, wherein when the computer program instruction is executed by the processor, the device is triggered to run the method and / or technical solution based on the aforementioned multiple embodiments of the present invention.

[0132] It is obvious to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential features of the present invention. Therefore, no matter from which aspect, the embodiments should be regarded as exemplary and non-restrictive, and the scope of the present invention is defined by the appended claims rather than the description of the above embodiments, and it is intended that all changes falling within the meaning and scope of the equivalent elements of the claims are included in the present invention. Any figure mark in the claims should not be regarded as limiting the claims involved. In addition, it is obvious that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices stated in the system claim can also be implemented by one unit or device through software or hardware. The words first, second, etc. are used to indicate names, and do not indicate any particular order.

Claims

1. A hybrid interactive system based on an AI big model, characterized in that: include: The acquisition module is used to obtain user requests; A natural language processing (NLP) module is used to process the user request and generate an NLP processing result; The AI ​​big model module is used to deeply understand and process the user request, and generate an external interface scheduling plan according to the understanding result; and after the external interface module returns the required information according to the external interface scheduling plan, summarize the acquired information and generate an AI processing result; A decision module, used to determine whether to output the NLP processing result generated by the natural language processing NLP module or the AI ​​processing result generated by the AI ​​large model module to the user; The decision-making module is based on: If the confidence of the NLP processing result meets the preset threshold and the return result of the AI ​​large model module is not received within the predetermined time, the NLP processing result is output to the user; Otherwise, output the AI ​​processing result to the user; An external interface module, used to obtain real-time and accurate information based on the external interface scheduling plan generated by the AI ​​large model module, and return it to the AI ​​large model module; The output module is used to output the processing result of the system to the user according to the judgment of the decision module.

2. The system according to claim 1, characterized in that: During the interaction with the user, the natural language processing NLP module and the AI ​​large model module simultaneously maintain the interaction context with the user to ensure the consistency of the interaction.

3. A hybrid interaction method based on AI big model, characterized in that: The following steps are involved: Get user request; Using a natural language processing (NLP) method to process the user request and obtain an NLP processing result; The user request is deeply understood and processed by the AI ​​big model, and an external interface scheduling plan is generated according to the understanding result; real-time and accurate information is obtained according to the external interface scheduling plan, and returned to the AI ​​big model; the AI ​​big model summarizes the information obtained by the external interface and generates an AI processing result; If the confidence of the NLP processing result meets a preset threshold and the AI ​​big model does not generate the AI ​​processing result within a predetermined time, the NLP processing result is output to the user; otherwise, the AI ​​processing result is output to the user.

4. The method according to claim 3, characterized in that: During the interaction with the user, the natural language processing (NLP) method and the AI ​​big model simultaneously maintain the interaction context with the user to ensure the consistency of the interaction.

5. A computer-readable storage medium having a computer program stored thereon, wherein when the program is executed by a processor, the steps of implementing the method described in any one of claims 3 and 4 are implemented.

6. An electronic device comprising: one or more processors; as well as A memory associated with the one or more processors, the memory being used to store program instructions, wherein the program instructions, when read and executed by the one or more processors, execute the steps of the method described in any one of claims 3 and 4.

Images