Intelligence question and answers system and method

Abstract

This invention provides an intelligent question-answering system and method. The method includes the following steps: Receiving question information; Collecting multiple pieces of knowledge related to the question information; Based on the multiple pieces of knowledge, executing a large language model to generate preliminary analysis results according to the question information. The preliminary analysis results include multiple steps required to provide answer information. Based on the question information and the preliminary analysis results, recalling multiple candidate APIs related to the multiple steps; Selecting a selected API corresponding to each step from the multiple candidate APIs to generate a problem-solving plan; Based on the multiple pieces of knowledge, executing the problem-solving plan with the large language model and dynamically adjusting the problem-solving plan according to the execution results; Finally, outputting answer information with the large language model based on the final execution results, thereby improving the accuracy of the answer.

Claims

1. An intelligent question-answering method, comprising: Receive the question information entered by the user; Collect multiple pieces of knowledge related to the information in question; Based on the aforementioned knowledge, a large language model is executed to generate preliminary analysis results according to the problem information, wherein the preliminary analysis results include multiple steps required to solve the problem; based on the problem information and the preliminary analysis results, multiple candidate application program interfaces (APIs) associated with the multiple steps are recalled; the multiple candidate APIs are integrated into a candidate API list; a selected API corresponding to each step is selected from the multiple candidate APIs to generate a problem-solving plan, wherein the step of selecting a selected API corresponding to each step to generate the problem-solving plan includes: based on the problem information, the preliminary analysis results, and the candidate API list, determining the parameters of each candidate API corresponding to each step to generate a first determination result; Based on the first judgment result, determine whether each candidate API corresponding to each step is applicable to generate a second judgment result; and based on the second judgment result, select the selected API corresponding to each step to generate a preliminary version of the problem-solving plan; and based on the multiple knowledge, execute the problem-solving plan and the large language model to output the solution information.

2. The intelligent question answering method as described in request item 1, wherein the step of recalling the plurality of candidate APIs associated with the plurality of steps based on the question information and the preliminary analysis results includes: Based on the problem information and the preliminary analysis results, the description of each step is expanded to include multiple scenario descriptions; Execute the Retrieval Augmented Generation (RAG) module to retrieve the multiple candidate APIs associated with each of the multiple context descriptions in each of the multiple steps; And integrate the multiple candidate APIs from each of the aforementioned steps to generate the candidate API list.

3. The intelligent question answering method as described in request item 1, wherein the preliminary analysis result includes the plurality of steps ordered according to the problem-solving approach, and the candidate API list includes the plurality of candidate APIs ordered according to the problem-solving approach.

4. The intelligent question answering method as described in request item 1, wherein a RAG module simultaneously retrieves the plurality of candidate APIs in the plurality of steps.

5. The intelligent question-answering method as described in claim 1, wherein the step of selecting the selected API corresponding to each of the plurality of candidate APIs to generate the problem-solving plan further includes: Based on the problem information, the correctness of the initial version of the solution plan is determined to generate a third judgment result; And based on the third judgment result, the parameters of the multiple selected APIs are modified to generate the problem-solving plan.

6. The intelligent question-answering method as described in claim 1, wherein the step of executing the problem-solving plan and the large language model based on the plurality of knowledge to output the answer information includes: The large language model is executed so that, if the execution progress of the problem-solving plan has not yet been completed, an API execution instruction for the selected API corresponding to the next step is constructed based on the execution progress. Based on the multiple pieces of knowledge, it is determined whether the API execution instruction requires authorization, so as to directly execute the API execution instruction to generate a call result, or to determine whether to continue executing the API execution instruction based on the authorization information provided by the user.

7. The intelligent question-answering method as described in claim 6, wherein the step of executing the problem-solving plan and the large language model based on the plurality of knowledge to output the answer information further includes: Determine whether the number of times the current step has encountered an anomaly exceeds a threshold, and then regenerate the problem-solving plan based on the execution progress and the intermediate execution results of the execution progress, or ignore the anomaly that occurred in the current step and continue to execute the problem-solving plan.

8. The intelligent question-answering method as described in claim 7, wherein the step of executing the problem-solving plan and the large language model based on the plurality of knowledge to output the answer information further includes: Execute the large language model to update the execution result in the problem-solving plan based on the invocation result; And execute the large language model to output the execution result as the intermediate execution result.

9. The intelligent question-answering method as described in claim 1, wherein the step of collecting the plurality of knowledge related to the question information includes: Execute a RAG module to retrieve domain knowledge associated with the problem information; The large language model is executed to retrieve internal knowledge associated with the question information; Utilize online tools to retrieve external knowledge related to the problem information; And to integrate the domain knowledge, the internal knowledge, and the external knowledge to form the multiple knowledge sets.

10. The intelligent question-answering method as described in claim 9, wherein the step of collecting the plurality of knowledge related to the question information further includes: Multiple preference scores are set for the domain knowledge, the internal knowledge, and the external knowledge, respectively; When knowledge conflict occurs, one of the domain knowledge, the internal knowledge, and the external knowledge is selected based on the total score of the multiple preference scores, or one of the domain knowledge, the internal knowledge, and the external knowledge is selected based on the user's selection instruction.

11. The intelligent question-answering method as described in claim 10 further includes: Based on the aforementioned knowledge, the large language model is executed to identify key entities associated with the question information; And based on the aforementioned knowledge, the large language model is executed to embed the key entities in each of the aforementioned steps.

12. An intelligent question-answering system, comprising: Storage device for storing large language models; And a processor, coupled to the storage device, for executing the intelligent question-answering method as described in claim 1.

Images