AI content generation method and device of mixed AI model, equipment and medium

Abstract

The invention provides an AI content generation method and device for a mixed AI model, equipment and a medium, and the method comprises the steps: inputting user input information into a large language model, and obtaining analyzed input information; the analyzed input information is input into the distribution intelligent system, indication information is obtained, and the indication information comprises a field expert system type matched with an answer; according to the indication information, inputting the analyzed input information into a domain expert system of a corresponding type to obtain answering information; and inputting the solution information into the large language model to obtain translated solution information, and feeding back the translated solution information to the user. The large language model is not used for answering user input information, but plays a role in analysis and translation, so that the field expert system and the distribution intelligent system can understand conveniently, and the distribution intelligent system selects the matched field expert system according to the analyzed input information so as to carry out professional answering on technical contents of the professional field, so that the technical content of the professional field can be quickly and accurately answered. And the answering accuracy is ensured.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an AI content generation method, device, equipment and medium for a hybrid AI model. Background Art

[0002] In recent years, the field of artificial intelligence (AI) has made great progress, including the development of various technologies such as machine learning (ML), natural language processing (NLP), and expert systems. In particular, large language models (LLMs) have brought substantial progress in the field of AI.

[0003] A large language model is a well-trained deep learning model that can understand and generate text in a human-like way. For example, GPT-3 (Generative Pre-trained Transformer 3) is a natural language processing model based on deep learning. It has very strong language understanding and generation capabilities and can be applied to a variety of natural language tasks, such as text generation, question-answering systems, language translation, sentiment analysis, etc. Although GPT-3 has a wide range of knowledge, it is not precise enough. It cannot give accurate answers to some technical content in some professional fields. Moreover, GPT-3 is set up so that it cannot give negative suggestions. Even if the user gives wrong information, it will use its logical reasoning to explain and agree with the wrong information, which will mislead the user.

[0004] Therefore, it is necessary to propose an AI content generation method, device, equipment and medium of a hybrid AI model to solve the above problems. Summary of the invention

[0005] The purpose of the present invention is to provide an AI content generation method, device, equipment and medium of a hybrid AI model, so as to improve the problem that the existing large language model cannot provide accurate answers to some professional technical content.

[0006] In a first aspect, the present invention provides an AI content generation method of a hybrid AI model, wherein the hybrid AI model includes a large language model, a distribution intelligence system, and a plurality of domain expert systems of different domain types, and the method includes:

[0007] Inputting user input information into the large language model to obtain parsed input information;

[0008] Inputting the parsed input information into the distribution intelligent system to obtain indication information, wherein the indication information includes the type of domain expert system matching the answer;

[0009] According to the indication information, the parsed input information is input into the domain expert system of the corresponding type to obtain answer information;

[0010] The answer information is input into the large language model to obtain translated answer information to be fed back to the user.

[0011] The beneficial effect of the AI ​​content generation method of the hybrid AI model provided by the present invention is that the large language model is not used to answer user input information, but plays a role of parsing and translation. The user input information is input into the large language model for parsing so that the domain expert system and the allocation intelligent system can understand it. The allocation intelligent system selects a matching domain expert system based on the parsed input information to answer the user input information. The matching domain expert system can provide professional answers to the technical content of the professional field to ensure the accuracy of the answer.

[0012] In a possible embodiment, after inputting the answer information into the large language model to obtain the translated answer information and feeding it back to the user, the method further includes:

[0013] The user input information and the translated answer information are used to train the large language model, the distribution intelligence system and the domain expert system.

[0014] In a second aspect, the present invention provides an AI content generation method of a hybrid AI model, wherein the hybrid AI model includes a large language model, a business intelligence system, and a plurality of domain expert systems of different domain types, the method comprising:

[0015] inputting user input information into the business intelligence system;

[0016] If the business intelligence system is able to answer the question, the business intelligence system outputs answer information;

[0017] If the business intelligence system cannot answer the question, inputting the user input information into the large language model to obtain parsed input information, and inputting the parsed input information into the domain expert system to obtain answer information;

[0018] Alternatively, the user input information is input into the large language model to obtain parsed input information, the parsed input information is input into the distribution intelligent system to obtain indication information, and according to the indication information, the parsed input information is input into the domain expert system of the corresponding type to obtain answer information;

[0019] Alternatively, the business intelligence system processes the user input information to obtain parsed input information and indication information, and inputs the parsed input information into the domain expert system of the corresponding type according to the indication information to obtain answer information;

[0020] Wherein, the indication information includes the type of domain expert system matching the solution;

[0021] The answer information is input into the large language model to obtain translated answer information to be fed back to the user.

[0022] The beneficial effect of the AI ​​content generation method of the hybrid AI model provided by the present invention is that the large language model is not used to answer user input information, but plays a role in parsing and translation. The user input information is input into the large language model for parsing, so that the domain expert system and the business intelligence system can understand it. The business intelligence system can directly give the answer information. When the business intelligence system cannot answer, there are three solutions: the first is to first parse it with the large language model, and then input it into the domain expert system for answer; the second is to first parse it with the large language model, and then the distribution intelligence system selects a domain expert system suitable for answering for answering; the third is that the business intelligence system directly parses the user input information and selects a domain expert system suitable for answering for answering. All three solutions use the domain expert system to give answers, thereby ensuring the accuracy of the answers.

[0023] In a possible embodiment, after inputting the answer information into the large language model to obtain the translated answer information and feeding it back to the user, the method further includes:

[0024] The user input information and the translated answer information are used to train the large language model, the business intelligence system and a plurality of the domain expert systems.

[0025] In a third aspect, the present invention further provides an AI content generation device of a hybrid AI model, which includes a module / unit that executes any possible design method of the first aspect. These modules / units can be implemented by hardware, or by executing corresponding software implementations through hardware.

[0026] In a fourth aspect, the present invention further provides an AI content generation device of a hybrid AI model, which includes a module / unit for executing any possible design method of the second aspect. These modules / units can be implemented by hardware, or by executing corresponding software implementations by hardware.

[0027] In a fifth aspect, an embodiment of the present invention provides an electronic device, including a processor and a memory, wherein the memory is used to store one or more computer programs; when the one or more computer programs stored in the memory are executed by the processor, the electronic device can implement any possible design method of the first aspect.

[0028] In the sixth aspect, an embodiment of the present invention further provides a computer-readable storage medium, the computer-readable storage medium comprising a computer program, and when the computer program runs on an electronic device, the electronic device executes any possible design method of the first aspect above.

[0029] In a seventh aspect, an embodiment of the present invention further provides a method comprising a computer program product, which, when running on an electronic device, enables the electronic device to execute any possible design of any of the above aspects.

[0030] For the beneficial effects of the second to seventh aspects, reference may be made to the description of the first and second aspects. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 Schematic diagram of the flow of the AI ​​content generation method of the hybrid AI model of the present invention in the first embodiment.

[0032] Figure 2 This is a flowchart of the implementation of the AI ​​content generation method of the hybrid AI model of the present invention in the first embodiment.

[0033] Figure 3 Schematic diagram of the flow of the AI ​​content generation method of the hybrid AI model of the present invention in the second embodiment.

[0034] Figure 4 This is a flowchart of the implementation of the AI ​​content generation method of the hybrid AI model of the present invention in the second embodiment.

[0035] Figure 5 Schematic diagram of an AI content generation device of a hybrid AI model of the present invention in a first embodiment.

[0036] Figure 6 Schematic diagram of the AI ​​content generation device of the hybrid AI model of the present invention in the second embodiment.

[0037] Figure 7 It is a schematic structural diagram of the electronic device of the present invention. DETAILED DESCRIPTION

[0038] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention. Unless otherwise defined, the technical terms or scientific terms used herein should be understood by people with general skills in the field to which the present invention belongs. "Including" and similar words used in this article mean that the elements or objects appearing before the word include the elements or objects listed after the word and their equivalents, without excluding other elements or objects.

[0039] In a first embodiment, the present invention provides an AI content generation method of a hybrid AI model, wherein the hybrid AI model includes a large language model, a distribution intelligence system, and a plurality of domain expert systems of different domain types, see Figure 1 and Figure 2 , the method comprising:

[0040] S101: Inputting user input information into a large language model to obtain parsed input information.

[0041] S102: Input the parsed input information into the distribution intelligent system to obtain indication information, wherein the indication information includes the type of domain expert system matching the solution.

[0042] S103: According to the instruction information, the parsed input information is input into a corresponding type of domain expert system to obtain answer information.

[0043] S104: Input the answer information into the large language model to obtain the translated answer information to feed back to the user.

[0044] Specifically, the scale of distributed intelligence systems is smaller than that of large language models.

[0045] In this embodiment, the large language model is responsible for "logical reasoning" and "speaking human language", that is, it only translates but does not answer. The domain expert system is responsible for "domain experts quickly locate knowledge points" and "quickly create a new world view", that is, it is responsible for providing professional answers. The allocation intelligent system is responsible for selecting domain expert systems suitable for answers, allowing the selected domain expert systems to answer, without having to input each domain expert system one by one to get the final answer. When answering each time, there is no need to call each domain expert system, saving computing resources and time, and improving response speed.

[0046] In a preferred embodiment, before inputting the user input information into the large language model to obtain the parsed input information, it also includes: training a number of domain expert systems of different domain types; training an allocation intelligent system, the allocation intelligent system is used to determine which domain expert system should answer the user input information based on the user input information.

[0047] In a specific embodiment, several domain expert systems of different domain types are trained, including: using several professional knowledge bases of different domain types to pre-train several domain expert systems of different domain types, wherein the professional knowledge bases and domain expert systems are pre-trained one by one. The purpose of this step is to allow the domain expert system to master the basic professional knowledge of the corresponding domain type on the basis of the prompt voice ability. Through the knowledge distillation method, the professional knowledge of different domain types is understood and translated by the large language model, and the output content understanding and the expression of generated content are respectively transferred to the domain expert system of the corresponding domain type. Specifically, the large language model is bridged with the artificial input of the specific domain professional knowledge, and the logical content is generated after the language is organized. Then, the generated content is used as the target to train the corresponding domain expert system to simulate the output process. During the training process, the domain expert system can learn the artificial "memory" and the understanding and generation capabilities of the large language model, and also has the professional knowledge of the professional knowledge base in the specific field.

[0048] In a specific embodiment, inputting user input information into a large language model to obtain parsed input information includes: inputting user input information into the large language model, the large language model extracting key information from the user input information, and parsing it into a form suitable for understanding by a domain expert system, to obtain the parsed input information. In this embodiment, the large language model completes this step by ignoring common sense in its own existing knowledge base and only simulating the semantic annotation method of language experts.

[0049] In a preferred embodiment, if the domain expert system cannot provide an answer, the large language model generates logical answer information. In this embodiment, the domain expert system and the large language model can be effectively combined to serve the professional content generation together. The large language model is responsible for understanding and generating human natural language, while the domain expert system relies on the professional knowledge in the professional knowledge base of the specific field to provide in-depth and rich knowledge in the field.

[0050] In a possible embodiment, after inputting the answer information into the large language model and obtaining the translated answer information to feed back to the user, it also includes: using the user input information and the translated answer information to train the large language model, the distribution intelligent system and the domain expert system, wherein the training results of the domain expert system include the domain knowledge boundary and scope, the domain knowledge entry and the interactive prompt (prompt) mode. For example, the domain knowledge boundary and scope can be the metaverse worldview or basic setting, and the domain knowledge entry can be the detailed setting applied to the rule details or individual cases. In this embodiment, based on the interaction samples between the large language model, the distribution intelligent system, the domain expert system and the user, the large language model, the distribution intelligent system and the domain expert system are continuously trained and optimized. By continuously learning and adjusting the model parameters, not only can the domain expert system further optimize its performance on the professional knowledge base in a specific field, but also can allow the large language model to learn how to work together, better process the user's input, and thus more accurately generate the output expected by the user. For example, when processing specific domain knowledge such as games and the worldview of the metaverse, related content can be processed and generated more effectively.

[0051] The unique training and interaction process of the hybrid AI model provides a new and efficient way to build and apply domain-specific knowledge bases. In other words, the hybrid AI model is characterized by its dynamic nature, which can adjust and improve its knowledge base based on user input information and the output of the domain expert system. This dynamic knowledge base building and optimization process provides a powerful tool for dealing with complex problems and tasks, especially in the context of games or metaverse worldviews.

[0052] In some specific embodiments, the domain expert system may be Prolog (Programming in logic), a knowledge graph, a vector database (VDB), a large language model (LLM), etc. For example, the large language model may be a GPT-3 model or a GPT-4 model, which are large-scale natural language processing models trained by deep learning technology, have powerful text understanding, generation and reasoning capabilities, and can be used for various natural language processing tasks, such as text generation, translation, question answering and dialogue systems. Among them, Prolog is a programming language and reasoning system based on first-order logic, which is particularly suitable for knowledge representation and reasoning. Prolog is widely used in expert systems and artificial intelligence applications, and can be used to build a rule-based knowledge base and answer queries or questions through reasoning processes. Knowledge graph is a graphical structure for representing and organizing relationships between entities. It is usually composed of nodes (entities) and edges (relationships) to form a directed graph. Knowledge graphs are used to store and represent large amounts of structured data so that machines can understand and process the relationships between entities, thereby supporting natural language understanding and the answering of complex questions. Vector Database (VDB) is a database oriented to vector data analysis. It has the functions of storing, organizing, managing, retrieving and analyzing vector data. It is a new type of special database.

[0053] The following is a specific explanation of the AI ​​content generation method of the hybrid AI model in the first embodiment based on the AI ​​hybrid model of GPT-3 and Prolog, wherein GPT-3 (Generative Pre-trained Transformer 3) is used as a large language model, the first Prolog system is used as a domain expert system, and the second Prolog system is used as a distribution intelligent system.

[0054] 1. Input user input information (for example, game instructions or queries) into the GPT-3 model. The GPT-3 model understands the input natural language and extracts key information.

[0055] 2. The GPT-3 model translates key information into a form that the Prolog system can understand. For example, if the user input is "I want to know the health value of my character", the GPT-3 model may translate it into a Prolog query, for example, health_status(character, Status).

[0056] 3. Input the translated key information into the second Prolog system (i.e., the distribution intelligent system) to obtain a domain expert system suitable for answering the user input information, which is the game domain.

[0057] 4. Input the translated key information into the first Prolog system in the game field (i.e., the domain expert system in the game field), which contains a professional knowledge base established in a specific domain game environment, including various rules and information about the game. The first Prolog system uses this information to answer queries.

[0058] 5. Return the answer of the first Prolog system in the game field to the GPT-3 model, which translates it into natural language output that humans can understand. For example, if the second Prolog system returns Status = 100, the GPT-3 model may translate it into "Your character's health is 100".

[0059] In a second embodiment, the present invention further provides an AI content generation method of a hybrid AI model, wherein the hybrid AI model includes a large language model, a business intelligence system, and a number of domain expert systems of different domain types, see Figure 3 and Figure 4 , the method comprising:

[0060] S301: Input user input information into the business intelligence system.

[0061] S302: If the business intelligence system can provide an answer, the business intelligence system outputs answer information.

[0062] S303: If the business intelligence system cannot answer the question, the user input information is input into the large language model to obtain parsed input information, and the parsed input information is input into the domain expert system to obtain answer information;

[0063] Alternatively, the user input information is input into the large language model to obtain parsed input information, the parsed input information is input into the distribution intelligent system to obtain indication information, and according to the indication information, the parsed input information is input into the corresponding type of domain expert system to obtain answer information;

[0064] Alternatively, the business intelligence system processes the user input information to obtain parsed input information and instruction information, and inputs the parsed input information into a corresponding type of domain expert system according to the instruction information to obtain answer information;

[0065] The indication information includes the type of domain expert system matching the solution.

[0066] S304: Input the answer information into the large language model to obtain the translated answer information to feed back to the user.

[0067] Specifically, the scale of business intelligence systems and distribution intelligence systems is smaller than that of large language models.

[0068] In this embodiment, the business intelligence system is responsible for processing problems and knowledge in the field, similar to the role of cache in efficient network technology. When a user asks a question, it will first try to find the answer from the cache of the business intelligence system. If the cache can provide an answer, then the obtained answer is directly returned to the user, which can efficiently handle common problems and provide users with faster answers without further processing by the domain expert system, thereby providing a quick response and saving system resources. If the business intelligence system cannot provide an answer, there are three solutions to obtain the answer: the first is to first parse it with a large language model, then input it into the domain expert system for answering, and after parsing with the large language model, the domain expert system can better understand and answer it; the second is to first parse it with a large language model, and then the distribution intelligent system selects a domain expert system suitable for answering to answer it. The distribution intelligent system can efficiently determine the field to which the user input information belongs, and the corresponding domain expert system will answer it, saving time and computing resources; the third is that the business intelligence system has the function of parsing and locating the field. The business intelligence system can directly parse the user input information and select a domain expert system suitable for answering to answer it, omitting the intermediate steps involving the large speech model, reducing delays, and being more efficient and responsive when solving problems in specific fields. The present invention uses the business intelligence system to answer, reducing the frequent calls to the domain expert system, saving computing resources and time. Its advantage is that it can efficiently handle various types of problems, provide accurate and professional answers, and save computing resources at the same time. The large language model, business intelligence system, domain expert system, and distribution intelligent system work together to provide users with comprehensive and efficient services. The flexibility and scalability of this system enables it to adapt to queries of different domains and complexity to provide a high-quality user experience.

[0069] In a preferred embodiment, before inputting the user input information into the large language model to obtain the parsed input information, it also includes: training to obtain a number of domain expert systems of different domain types; and training to obtain a business intelligence system.

[0070] In a specific embodiment, several domain expert systems of different domain types are trained, including: using several professional knowledge bases of different domain types to pre-train several domain expert systems of different domain types, wherein the professional knowledge bases and domain expert systems are pre-trained one by one. The purpose of this step is to allow the domain expert system to master the basic professional knowledge of the corresponding domain type on the basis of the prompt voice ability. Through the knowledge distillation method, the professional knowledge of different domain types is understood and translated by the large language model, and the output content understanding and the expression of generated content are respectively transferred to the domain expert system of the corresponding domain type. Specifically, the large language model is bridged with the artificial input of the specific domain professional knowledge, and the logical content is generated after the language is organized. Then, the generated content is used as the target to train the corresponding domain expert system to simulate the output process. During the training process, the domain expert system can learn the artificial "memory" and the understanding and generation capabilities of the large language model, and also has the professional knowledge of the professional knowledge base in the specific field.

[0071] In a specific embodiment, training a business intelligence system includes: pre-training the business intelligence system using several professional knowledge bases of different domain types. The purpose of this step is to allow the business intelligence system to master basic professional knowledge of different domain types on the basis of voice prompting capabilities. Through knowledge distillation, the professional knowledge of different domain types is understood and translated by a large language model, and the output content understanding and generated content performance are migrated to the business intelligence system. Specifically, the large language model is bridged with several professional knowledge of different domain types input manually, and logical content is generated after the language is organized. Then, the generated content is used as the target to train the business intelligence system to simulate the output process. During the training process, the business intelligence system can learn the artificial "memory" and the understanding and generation capabilities of the large language model, and also has the professional knowledge of professional knowledge bases of different domain types.

[0072] In a specific embodiment, inputting user input information into a large language model to obtain parsed input information includes: inputting user input information into the large language model, the large language model extracting key information from the user input information, and parsing it into a form suitable for understanding by a domain expert system, to obtain the parsed input information. In this embodiment, the large language model completes this step by ignoring common sense in its own existing knowledge base and only simulating the semantic annotation method of language experts.

[0073] In a possible embodiment, after the answer information is input into the large language model and the translated answer information is obtained for feedback to the user, it also includes: using the user input information and the translated answer information to train the large language model, the business intelligence system, the distribution intelligence system and the domain expert system, wherein the training results of the domain expert system include the boundaries and scope of domain knowledge, domain knowledge items and interactive prompt modes.

[0074] The following is a specific explanation of the AI ​​content generation method of the hybrid AI model in the second embodiment based on the AI ​​hybrid model of GPT-4 and knowledge graph, wherein GPT-4 (Generative Pre-trained Transformer 4) is used as a large language model, the first knowledge graph model is used as a domain expert system, and the second knowledge graph model is used as a business intelligence system.

[0075] 1. Input user input information (e.g., metaverse query or operation) into the GPT-4 model. The GPT-4 model understands the input natural language and extracts key information.

[0076] 2. The GPT-4 model translates key information into a form that the knowledge graph model can understand. For example, if the user input is "I want to know the climate type of this planet", the GPT-4 model may translate it into a knowledge graph query, for example, (planet, 'climate type', ?).

[0077] 3. Input the translated key information into the second knowledge graph model (i.e., business intelligence system). The second knowledge graph model contains a knowledge base built in the metaverse environment, which contains various rules and information about the metaverse. The second knowledge graph model uses this information to answer queries.

[0078] 4. The answer from the second knowledge graph model is returned to the GPT-4 model, which translates it into natural language output that humans can understand. For example, if the second knowledge graph model returns 'climate type' = 'tropical', the GPT-4 model may translate it into "The climate type of this planet is tropical."

[0079] In addition, the present invention provides an AI content generation device of a hybrid AI model, wherein the hybrid AI model includes a large language model, a distribution intelligence system, and a number of domain expert systems of different domain types, see Figure 5The device includes: a first parsing unit 501, which is used to input the user input information into the large language model to obtain the parsed input information; an indication unit 502, which is used to input the parsed input information into the distribution intelligent system to obtain indication information, wherein the indication information includes the type of domain expert system matching the answer; a first answer unit 503, which is used to input the parsed input information into the domain expert system of the corresponding type according to the indication information to obtain the answer information; a first translation unit 504, which is used to input the answer information into the large language model to obtain the translated answer information to feed back to the user. All relevant contents of each step involved in the first embodiment of the above method can be referred to the functional description of the corresponding functional module, and will not be repeated here.

[0080] In addition, the present invention also provides an AI content generation device of a hybrid AI model, wherein the hybrid AI model includes a large language model, a business intelligence system, and a number of domain expert systems of different domain types, see Figure 6 , the device includes: an input unit 601, which is used to input the user input information into the large language model to obtain the parsed input information; a second answer unit 602, which is used to input the user input information into the large language model to obtain the parsed input information if the business intelligence system cannot answer it, and input the parsed input information into the domain expert system to obtain the answer information; or, input the user input information into the large language model to obtain the parsed input information, input the parsed input information into the distribution intelligent system to obtain the indication information, and according to the indication information, input the parsed input information into the domain expert system of the corresponding type to obtain the answer information; or, the business intelligence system processes the user input information to obtain the parsed input information and the indication information, and according to the indication information, inputs the parsed input information into the domain expert system of the corresponding type to obtain the answer information; wherein the indication information includes the type of domain expert system matching the answer; a second translation unit 603, which is used to input the answer information into the large language model to obtain the translated answer information to feed back to the user. All relevant contents of each step involved in the second embodiment of the above method can be referred to the functional description of the corresponding functional module, which will not be repeated here.

[0081] In some other embodiments of the present invention, an electronic device is disclosed. Figure 7 , the electronic device may include: one or more processors 701; memory 702; display 703; one or more applications (not shown); and one or more computer programs 704, and the above components may be connected via one or more communication buses 705. The one or more computer programs 704 are stored in the above memory 702 and configured to be executed by the one or more processors 701, and the one or more computer programs 704 include instructions, which may be used to execute the following: Figure 1 , Figure 3 , Figure 5 and Figure 6 And each step in the corresponding embodiment.

[0082] Through the description of the above implementation methods, technicians in the relevant field can clearly understand that for the convenience and simplicity of description, only the division of the above functional modules is used as an example. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device is divided into different functional modules to complete all or part of the functions described above. The specific working process of the system, device and unit described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0083] Each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0084] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned storage medium includes: flash memory, mobile hard disk, read-only memory, random access memory, disk or optical disk, etc., various media that can store program codes.

[0085] Although the embodiments of the present invention are described in detail above, it is obvious to those skilled in the art that various modifications and variations can be made to these embodiments. However, it should be understood that such modifications and variations are within the scope and spirit of the present invention as described in the claims. Moreover, the present invention described herein may have other embodiments and may be implemented or realized in a variety of ways.

Claims

1. A method for generating AI content using a hybrid AI model, characterized in that: The hybrid AI model includes a large language model, a distribution intelligence system, and several domain expert systems of different domain types. The method includes: Inputting user input information into the large language model to obtain parsed input information; Inputting the parsed input information into the distribution intelligent system to obtain indication information, wherein the indication information includes the type of domain expert system matching the answer; According to the indication information, the parsed input information is input into the domain expert system of the corresponding type to obtain answer information; The answer information is input into the large language model to obtain translated answer information to be fed back to the user.

2. The method according to claim 1, characterized in that After inputting the answer information into the large language model to obtain the translated answer information and feeding it back to the user, the method further includes: The user input information and the translated answer information are used to train the large language model, the distribution intelligence system and the domain expert system.

3. A method for generating AI content using a hybrid AI model, characterized in that: The hybrid AI model includes a large language model, a business intelligence system, and several domain expert systems of different domain types. The method includes: inputting user input information into the business intelligence system; If the business intelligence system is able to answer the question, the business intelligence system outputs answer information; If the business intelligence system cannot answer the question, inputting the user input information into the large language model to obtain parsed input information, and inputting the parsed input information into the domain expert system to obtain answer information; Alternatively, the user input information is input into the large language model to obtain parsed input information, the parsed input information is input into the distribution intelligent system to obtain indication information, and according to the indication information, the parsed input information is input into the domain expert system of the corresponding type to obtain answer information; Alternatively, the business intelligence system processes the user input information to obtain parsed input information and indication information, and inputs the parsed input information into the domain expert system of the corresponding type according to the indication information to obtain answer information; Wherein, the indication information includes the type of domain expert system matching the solution; The answer information is input into the large language model to obtain translated answer information to be fed back to the user.

4. The method according to claim 3, characterized in that After inputting the answer information into the large language model to obtain the translated answer information and feeding it back to the user, the method further includes: The user input information and the translated answer information are used to train the large language model, the business intelligence system and a plurality of the domain expert systems.

5. An AI content generation device of a hybrid AI model, characterized in that: The hybrid AI model includes a large language model, a distribution intelligence system, and several domain expert systems of different domain types. The device includes: A first parsing unit, configured to input user input information into the large language model to obtain parsed input information; an indication unit, configured to input the parsed input information into the distribution intelligent system to obtain indication information, wherein the indication information includes a domain expert system type matching the answer; A first answering unit, configured to input the parsed input information into the domain expert system of the corresponding type according to the indication information, to obtain answering information; The first translation unit is used to input the answer information into the large language model to obtain translated answer information to feed back to the user.

6. The device according to claim 5, characterized in that Also includes: The first training unit is used to train the large language model, the distribution intelligent system and the domain expert system by using the user input information and the translated answer information after feeding back the translated answer information to the user.

7. An AI content generation device of a hybrid AI model, characterized in that: The hybrid AI model includes a large language model, a business intelligence system, and several domain expert systems of different domain types. The device includes: An input unit, used to input user input information into the business intelligence system; A second answering unit, configured to cause the business intelligence system to output answering information if the business intelligence system is able to answer the question; If the business intelligence system cannot answer the question, inputting the user input information into the large language model to obtain parsed input information, and inputting the parsed input information into the domain expert system to obtain answer information; Alternatively, the user input information is input into the large language model to obtain parsed input information, the parsed input information is input into the distribution intelligent system to obtain indication information, and according to the indication information, the parsed input information is input into the domain expert system of the corresponding type to obtain answer information; Alternatively, the business intelligence system processes the user input information to obtain parsed input information and indication information, and inputs the parsed input information into the domain expert system of the corresponding type according to the indication information to obtain answer information; Wherein, the indication information includes the type of domain expert system matching the solution; The second translation unit is used to input the answer information into the large language model to obtain translated answer information to feed back to the user.

8. The device according to claim 7, characterized in that Also includes: The second training unit is used to train the large language model, the business intelligence system and the plurality of domain expert systems by using the user input information and the translated answer information after feeding back the translated answer information to the user.

9. An electronic device, characterized in that: include: A processor and a memory, wherein the memory is used to store a computer program; The processor is configured to execute the computer program stored in the memory so as to enable the electronic device to perform the method according to any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images