Method for constructing intelligent question-answering service and related device

Abstract

A method and related equipment for constructing an intelligent question-and-answer service are disclosed. The method includes: obtaining a first database corresponding to the data domain requirements based on the data domain requirements corresponding to the question-and-answer scenario; constructing a question-and-answer knowledge model based on a preset large-scale language model and the first database; obtaining the application context of the question-and-answer scenario; decoupling a question-and-answer service clone corresponding to the application context from the question-and-answer knowledge model, wherein the application context includes at least one of single-context application, multi-context application, local context application, and network context application, and the data on which the question-and-answer service clone depends is stored in a local database; and outputting the question-and-answer result corresponding to the question-and-answer content based on the response of the question-and-answer service clone to the received question-and-answer content.

Claims

1. A method for constructing an intelligent question-answering service, applied to computer devices, wherein, The construction method includes: the computer device acquiring a first database corresponding to the data domain requirements of the question-and-answer scenario; the computer device constructing a question-and-answer knowledge model based on a preset large-scale language model and the first database; the computer device acquiring the application context of the question-and-answer scenario; the computer device decoupling a question-and-answer service clone corresponding to the application context from the question-and-answer knowledge model, wherein the application context includes at least one of single-context application, multi-context application, local context application, and network context application, and the data on which the question-and-answer service clone depends is stored in a local database; and the computer device responding to the received question-and-answer content based on the question-and-answer service clone and outputting the question-and-answer result corresponding to the question-and-answer content.

2. The method for constructing the intelligent question-answering service as described in request item 1, wherein, The question-and-answer service clone can be any one of the following: service clone architecture, application clone architecture, or local clone architecture. The service clone architecture is used for multi-context question-and-answer, the application clone architecture is used for single-context question-and-answer under the service clone architecture, and the local clone architecture is used for local single-context question-and-answer under the service clone architecture.

3. The method for constructing the intelligent question-answering service as described in claim 2, wherein, The decoupling steps of the service avatar architecture include: the computer device constructs a virtual machine service and a service avatar database on which the virtual machine service depends, the virtual machine service being obtained by virtualizing a preset artificial intelligence search engine and a preset large language model; the computer device copies the data content in the first database and the data content in the second database on which the preset large language model depends to the service avatar database; the computer device constructs an application manager for managing tools and agents; the computer device deploys the tools required by the service avatar architecture to the virtual machine service based on the application manager, and establishes an interaction channel between the agents required by the service avatar architecture and the virtual machine service.

4. The method for constructing the intelligent question-answering service as described in claim 2, wherein, The first database includes data content for various question-and-answer scenarios. The decoupling steps of the application avatar architecture include: the computer device dividing the data content of the first database into multiple data sets, each data set corresponding to a question-and-answer scenario, and managing access permissions for each data set separately; the computer device setting the tools and intelligent agents required for the application avatar architecture based on the application manager; and the computer device constructing an application avatar database corresponding to the application avatar architecture, the application avatar database including data content from the second database on which the preset large language model depends and data content from the data sets corresponding to the application avatar architecture.

5. The method for constructing the intelligent question-answering service as described in claim 2, wherein, The first database includes data content for various question-and-answer scenarios. The decoupling steps of the local clone architecture include: the computer device dividing the data content of the first database into multiple datasets, each dataset corresponding to a question-and-answer scenario, and each dataset having its own access control; the computer device setting up the tools and agents required for the local clone architecture based on the application manager; the computer device constructing a local clone database corresponding to the local clone architecture, the local clone database including vectors obtained by transforming target data content based on a preset embedding model, the target data content including data content in the second database on which the preset large language model depends and data content in the dataset corresponding to the local clone architecture; and the computer device establishing an interaction channel between the preset embedding model and the tools and agents required for the local clone architecture.

6. The method for constructing the intelligent question-answering service as described in claim 1, wherein, The step of responding to the received question-and-answer content based on the question-and-answer service clone and outputting the question-and-answer result corresponding to the question-and-answer content includes: the question-and-answer service clone generating a query vector for the question-and-answer content based on the question-and-answer content; the question-and-answer service clone retrieving a database corresponding to the question-and-answer service clone based on the query vector to determine the target question-and-answer data; and the question-and-answer service clone reconstructing the target question-and-answer data based on preset grammar rules and outputting the question-and-answer result that conforms to natural language rules.

7. The method for constructing the intelligent question-answering service as described in claim 6, wherein, The process of reconstructing the target question-and-answer data based on preset grammatical rules to output question-and-answer results that conform to natural language rules includes: the question-and-answer service clone performing syntactic correction on the target question-and-answer data based on the preset grammatical rules; the question-and-answer service clone identifying professional information in the target question-and-answer data and replacing the professional information with words using a preset specialized terminology mapping table; and the question-and-answer service clone performing text coherence detection and logical relationship enhancement on the target question-and-answer data after word replacement to obtain the question-and-answer results that conform to natural language rules.

8. A device for constructing an intelligent question-and-answer service, wherein, The construction device includes: a first acquisition module, used to acquire a first database corresponding to the data domain requirements based on the data domain requirements corresponding to the question-and-answer scenario; a construction module, used to construct a question-and-answer knowledge model based on a preset large-scale language model and the first database; a second acquisition module, used to acquire the application context of the question-and-answer scenario; a decoupling module, used to decouple a question-and-answer service clone corresponding to the application context from the question-and-answer knowledge model, wherein the application context includes at least one of single-context application, multi-context application, local context application, and network context application, and the data on which the question-and-answer service clone depends is stored in a local database; and a response module, used to respond to the received question-and-answer content based on the question-and-answer service clone and output the question-and-answer result corresponding to the question-and-answer content.

9. A computer device, said computer device comprising a processor and memory, wherein, The memory is used to store instructions, and the processor is used to call the instructions in the memory to cause the computer device to execute the method for constructing a smart question-and-answer service as described in any one of requests 1 to 7.

10. A computer-readable storage medium, wherein, The computer-readable storage medium stores computer instructions that, when executed on a computer device, cause the computer device to perform a method for constructing a smart question-and-answer service as described in any one of requests 1 to 7.

Images