Data processing method and system based on AI question and answer robot

Abstract

The application provides a data processing method and system based on an AI question and answer robot, belongs to the technical field of data processing, and specifically comprises the following steps: based on the analysis deviation of the screening question type, determining the switching processing data of the user between different reply processing modes due to the existence of the analysis deviation, determining the switching influence type of the screening question type based on the switching processing data, and determining the data processing method of the user interaction data of the screening question type based on the switching influence type of the different screening question types and the update data of the screening question type, thereby improving the operation stability of the question and answer robot.

Description

A data processing method and system based on an AI question-answering robot Technical Field

[0001] This invention belongs to the field of data processing technology, and in particular relates to a data processing method and system based on an AI question-answering robot. Background Technology

[0002] Different companies have different rules and regulations, and different office issues. The original technical solutions often rely on manual experience to handle problems, which is not only inefficient, but also often results in technical problems where the data processing results fail to meet the requirements due to a lack of comprehensive understanding of the rules and regulations.

[0003] To address the aforementioned technical issues, existing solutions utilize question-answering robots based on local knowledge bases to meet users' personalized answering needs. However, as regulations and systems are updated, the existing knowledge base needs to be updated accordingly. If the AI ​​question-answering robot is frequently updated along with the knowledge base, it will inevitably lead to poor operational stability of the robot.

[0004] Therefore, there is an urgent need for a data processing method and system based on AI question-answering robots. Summary of the Invention

[0005] To achieve the objectives of this invention, the following technical solution is adopted:

[0006] Specifically, this application provides a data processing method based on an AI question-answering robot, which includes:

[0007] S1 uses the question answering data of the question answering robot to determine that the question answering robot cannot perform real-time update processing of the robot's question recognition model. Based on the updated data of the knowledge base, it determines the question type that uses the knowledge base for question reply processing and uses it as the screening question type.

[0008] S2 determines the switching processing data between different response processing methods due to the existence of parsing deviation in the filtering question type based on the parsing deviation. Based on the switching processing data, it determines the switching impact type of the filtering question type. Based on the switching impact type of different filtering question types and the updated data of the filtering question type, it determines the data processing method for user interaction data of filtering question types.

[0009] The beneficial effects of this invention are as follows:

[0010] Based on the switching processing data, the impact type of switching on the problem type is determined. This enables a comprehensive consideration of the problem that has a significant impact on the overall problem processing efficiency, from the perspective of frequent switching between multiple response processing methods due to identification bias. The targeted determination of the switching impact type lays the foundation for further determining the data processing method based on the overall impact.

[0011] By considering the impact of switching different question types and the updated data of question types, a data processing method for user interaction data of question types was determined. This method not only takes into account the impact of the differences in the impact of switching on the overall question response efficiency, but also, by combining the daily updated data of question types, assesses the changing state of the impact of question types on response efficiency. Based on the impact on response efficiency, targeted data processing is carried out, laying the foundation for timely and effective updating of the question recognition system of the question-answering robot's AI.

[0012] Furthermore, the question-answering data is determined by including the number of questions answered by the question-answering robot in history.

[0013] Furthermore, it is determined that the question-answering robot cannot perform real-time updates to the robot's question recognition model, specifically including:

[0014] Based on the question-answering data of the question-answering robot, determine the question data answered by the question-answering robot on different dates;

[0015] Based on the question data answered on different dates in history, identify the peak periods for question processing on different dates;

[0016] Based on the peak time for question processing on different dates, it is determined whether the question-answering robot can perform real-time updates of the robot's question recognition model.

[0017] Furthermore, the method for determining the data processing method for filtering user interaction data of question types is as follows:

[0018] By switching the impact type of different screening question types, the constituent data of the serious impact questions in the screening question types are determined;

[0019] Based on the updated data of the filter question types, determine the number of updates for the filter question types on different dates within the most recent preset time period;

[0020] A data processing method for determining user interaction data for the filtering question type based on the number of updates to the filtering question type on different dates within a recent preset time period and the constituent data of the question that seriously affects the problem.

[0021] In a second aspect, the present invention provides a computer system comprising: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the aforementioned data processing method based on an AI question-answering robot when running the computer program.

[0022] Other features and advantages will be set forth in the following description, and the objects and other advantages of the invention are realized and obtained through the structures particularly pointed out in the description and the drawings.

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0024] The above and other features and advantages of the present invention will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0025] Figure 1 is a flowchart of a data processing method based on an AI question-answering robot;

[0026] Figure 2 is a flowchart illustrating the process of determining when a question-answering robot cannot perform real-time updates to its question recognition model;

[0027] Figure 3 is a flowchart of the method for determining the screening question type;

[0028] Figure 4 is a flowchart of the method for determining the impact of switching the question type on the screening;

[0029] Figure 5 is a framework diagram of a computer system. Detailed Implementation

[0030] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this specification, and not all embodiments. Based on the embodiments of this specification, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of this specification.

[0031] In this application, based on the update status of the knowledge base and the identification deviation of question types using the knowledge base for question processing, a data processing scheme for the interactive data of question types using the knowledge base for question processing is determined in advance. Then, the interactive data after data processing can be used to train and update the question-answering robot, thereby avoiding the technical problem of poor operational stability of the question-answering robot caused by training and updating the question-answering robot when the knowledge base is frequently updated and the amount of interactive data of the question types associated with the knowledge base is small.

[0032] Example 1

[0033] As shown in Figure 1, this application provides a data processing method based on an AI question-answering robot, specifically including:

[0034] S1 uses the question answering data of the question answering robot to determine that the question answering robot cannot perform real-time update processing of the robot's question recognition model. Based on the updated data of the knowledge base, it determines the question type that uses the knowledge base for question reply processing and uses it as the screening question type.

[0035] Furthermore, the question-answering data is determined by including the number of questions answered by the question-answering robot in history.

[0036] Specifically, as shown in Figure 2, determining that the question-answering robot cannot perform real-time updates to the robot's question recognition model includes:

[0037] Based on the question-answering data of the question-answering robot, determine the question data answered by the question-answering robot on different dates;

[0038] Based on the question data answered on different dates in history, identify the peak periods for question processing on different dates;

[0039] Based on the peak time for question processing on different dates, it is determined whether the question-answering robot can perform real-time updates of the robot's question recognition model.

[0040] It is understood that the busy period for problem processing is the period when the number of problems answered exceeds a preset threshold.

[0041] It should be noted that if there are busy periods for problem processing on different dates, then the question-answering robot will not be able to perform real-time updates of the robot's problem recognition model.

[0042] Optionally, determining that the question-answering robot cannot perform real-time updates to the robot's question recognition model specifically includes:

[0043] Based on the question-answering data of the question-answering robot, determine the question data answered by the question-answering robot on different dates;

[0044] Based on the question data answered on different dates in history, determine the total number of questions answered on different dates;

[0045] Based on the total number of questions answered on different dates, it is determined whether the question-answering robot can perform real-time updates of the robot's question recognition model.

[0046] It is understandable that when the total number of questions answered on different dates exceeds a preset threshold, the question-answering robot is not allowed to perform real-time updates of its question recognition model. This is to avoid problems in the question recognition model update process that could significantly affect the accuracy of the answers.

[0047] Specifically, as shown in Figure 3, the method for determining the type of question to be filtered is as follows:

[0048] Based on the updated data of the knowledge base, determine the items in the knowledge base that have been updated;

[0049] The question type corresponding to the project whose knowledge base has been updated is used as the filter question type.

[0050] It should be noted that the user's switching processing data between different response processing methods includes the number of times the user switches between the response processing method using the question recognition model and the response processing method using the knowledge base.

[0051] S2 determines the switching processing data between different response processing methods due to the existence of parsing deviation in the filtering question type based on the parsing deviation. Based on the switching processing data, it determines the switching impact type of the filtering question type. Based on the switching impact type of different filtering question types and the updated data of the filtering question type, it determines the data processing method for user interaction data of filtering question types.

[0052] Specifically, the method for determining the impact type of the switching of the screening question type is as follows:

[0053] Based on the parsing deviation of the question type, determine the number of times the user switches between different response processing methods due to the parsing deviation, and use this as the interference switching number.

[0054] Based on the number of interference switching, users who switch data between different response processing methods due to parsing bias are identified and treated as users affected by interference.

[0055] Based on the number of times different interferences affect users, interference impact factors are determined for different interferences, and the switching impact type of the screening problem type is determined using the interference impact factors.

[0056] Specifically, if there are no users whose interference impact factor does not meet the requirements, that is, if there are no users whose interference impact factor is too large, then the switching impact type of the screening problem type is determined to be an issue with no impact.

[0057] In one possible specific embodiment, the interference impact factor of the interference affecting the user is determined based on the ratio of the number of times the interference affecting the user switches to the number of times the interference affecting the user responds in the question-and-answer robot.

[0058] Specifically, if there are users whose interference impact factor does not meet the requirements, the number of users whose interference impact factor does not meet the requirements is obtained. If the number of users whose interference impact factor does not meet the requirements is greater than the preset threshold for the number of users whose interference impact factor does not meet the requirements, the degree of interference impact on users is high, and the switching impact type of the screened problem type is determined to be a serious impact problem.

[0059] It is understandable that if the number of users whose interference impact factor does not meet the requirements is not greater than the preset threshold for the number of users whose interference impact factor does not meet the requirements, the interference impact value is determined based on the sum of the interference impact factors of different users whose interference impact factor does not meet the requirements. When the interference impact value is greater than the preset impact threshold, the switching impact type of the problem type is determined to be a serious impact problem.

[0060] Furthermore, if the interference impact value is not greater than the preset impact threshold, then the switching impact type of the filtered problem type is determined to be a problem with no impact.

[0061] Specifically, the method for determining the data processing method for filtering user interaction data of question types is as follows:

[0062] By switching the impact type of different screening question types, the constituent data of the serious impact questions in the screening question types are determined, and the comprehensive interference impact value is determined based on the number of serious impact questions and the interference impact value.

[0063] Based on the updated data of the filter question types, determine the number of updates for the filter question types on different dates within the most recent preset time period;

[0064] A data processing method for determining user interaction data of the filtering question type based on the number of updates to the filtering question type on different dates within a recent preset time period and the overall interference impact value.

[0065] Specifically, when the comprehensive interference impact value is less than the preset impact threshold, only the user interaction data of the seriously affecting problem needs to be processed to obtain a training set that can be used for the problem identification model. When the amount of user interaction data of the seriously affecting problem meets the requirements, the user interaction data of all the filtered problem types can be processed to realize the training and update of the problem identification model.

[0066] It should be noted that when the number of interactions of the user interaction data for the selected question type, i.e., the number of interactions of user interaction data using different keywords, is greater than the preset interaction number threshold, it is determined that the amount of user interaction data for the selected question type meets the requirements. When the amount of user interaction data that seriously affects the question meets the requirements, the training and updating process of the question identification model is performed.

[0067] Furthermore, when the comprehensive interference impact value is not less than the preset impact threshold, the number of updates for the screening question type in different dates within the most recent preset time period is determined. If the average number of updates for the screening question type in different dates within the most recent preset time period is greater than the preset update number threshold, then data processing is performed on the user interaction data of all screening question types. Thus, when the user interaction data meets the requirements, the training and update processing of the question identification model is realized.

[0068] Furthermore, if the average number of updates for different question types within the recent preset time period is not greater than the preset update number threshold, only the user interaction data that seriously affects the question needs to be processed to obtain a training set that can be used for the question identification model. When the amount of user interaction data that seriously affects the question meets the requirements, all user interaction data for different question types can be processed to achieve the training and update of the question identification model.

[0069] Example 2

[0070] Secondly, as shown in Figure 5, the present invention provides a computer system, including: a memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the above-described data processing method based on an AI question-answering robot when running the computer program.

[0071] In one possible specific embodiment:

[0072] The period during which at least 50 questions are answered is defined as the busy period for question processing. If a busy period for question processing exists on different dates, then the question-answering robot cannot perform real-time updates of the robot's question recognition model.

[0073] For example, if there are no users with interference impact factors greater than 0.05, the switching impact type of the selected question type is determined to be an "no impact" question. If the number of users with interference impact factors that do not meet the requirements is greater than 10, the degree of interference impact on users is relatively high, and the switching impact type of the selected question type is determined to be a "serious impact" question. The interference impact value is determined based on the sum of the interference impact factors of different users. When the interference impact value is greater than a preset impact threshold, for example, greater than 0.6, the switching impact type of the selected question type is determined to be a "serious impact" question. If the interference impact value is not greater than the preset impact threshold, the switching impact type of the selected question type is determined to be an "no impact" question.

[0074] When the comprehensive interference impact value is less than 1.5, only the user interaction data of the seriously affecting problem needs to be processed. If the number of interactions of the user interaction data with different keywords corresponding to the problem type being filtered is greater than 50, then the amount of user interaction data of the problem type being filtered is determined to meet the requirements. When the amount of user interaction data of any seriously affecting problem meets the requirements, then the problem identification model is trained and updated.

[0075] If the average number of updates for different filter question types on different dates within the recent preset time period is greater than 2, then data processing is performed on all user interaction data for all filter question types. Once the user interaction data meets the requirements, the training and update processing of the question identification model can be achieved. If the average number of updates for different filter question types on different dates within the recent preset time period is not greater than the preset update number threshold, only the user interaction data that seriously affects the questions needs to be processed to obtain a training set that can be used for the question identification model. Once the amount of user interaction data that seriously affects the questions meets the requirements, data processing can be performed on the user interaction data for all filter question types to achieve the training and update processing of the question identification model.

[0076] Specifically, as shown in Figure 4, the method for determining the impact type of the switching of the screening question type is as follows:

[0077] Based on the parsing deviation of the question type, determine the number of times the user switches between different response processing methods due to the parsing deviation, and use this as the interference switching number.

[0078] Based on the number of interference switching, users who switch data between different response processing methods due to parsing bias are identified and treated as users affected by interference.

[0079] Based on the number of times different interferences affect users, the switching impact type of the filtering problem type is determined.

[0080] It is understandable that when there is no interference affecting users, there is no error due to parsing deviation when switching between different response processing methods, which slows down the response processing efficiency. Therefore, based on this, the switching of the filtering question type is determined to be an issue with no impact.

[0081] Furthermore, when interference affects users, the number of interference switching times within the most recent preset time period is obtained. If either the number of users affected by interference or the number of interference switching times within the most recent preset time period does not meet the requirements, that is, if the number of users affected by interference is too large or the number of interference switching times is too large, then the switching impact type of the filtered problem type is determined to be a serious impact problem.

[0082] Additionally, it should be noted that when the number of users affected by the interference and the number of interference switching times both meet the requirements within the most recent preset time period, the interference impact factor for different users is determined based on the number of interference switching times for different users. If there are no users whose interference impact factor does not meet the requirements, i.e., there are no users whose interference impact factor is too large, then the switching impact type of the selected problem type is determined to be a problem with no impact.

[0083] In one possible specific embodiment, the interference impact factor of the interference affecting the user is determined based on the ratio of the number of times the interference affecting the user switches to the number of times the interference affecting the user responds in the question-and-answer robot.

[0084] Specifically, if there are users whose interference impact factor does not meet the requirements, the number of users whose interference impact factor does not meet the requirements is obtained. If the number of users whose interference impact factor does not meet the requirements is greater than a preset threshold for the number of users whose interference impact factor does not meet the requirements, then the switching impact type of the problem type is determined to be a serious impact problem.

[0085] It is understandable that if the number of users whose interference impact factor does not meet the requirements is not greater than the preset threshold for the number of users whose interference impact factor does not meet the requirements, the interference impact value is determined based on the sum of the interference impact factors of different users whose interference impact factor does not meet the requirements. When the interference impact value is greater than the preset impact threshold, the switching impact type of the problem type is determined to be a serious impact problem.

[0086] Furthermore, if the interference impact value is not greater than the preset impact threshold, then the switching impact type of the filtered problem type is determined to be a problem with no impact.

[0087] Example 3

[0088] Specifically, the method for determining the data processing method for filtering user interaction data of question types is as follows:

[0089] By switching the impact type of different screening question types, the constituent data of the serious impact questions in the screening question types are determined;

[0090] Based on the updated data of the filter question types, determine the number of updates for the filter question types on different dates within the most recent preset time period;

[0091] A data processing method for determining user interaction data for the filtering question type based on the number of updates to the filtering question type on different dates within a recent preset time period and the constituent data of the question that seriously affects the problem.

[0092] Furthermore, when the number of question types to be screened is within a preset range, that is, when the number of question types to be screened is small, only the user interaction data that seriously affects the question needs to be processed to obtain a training set that can be used for the question identification model. When the amount of user interaction data that seriously affects the question meets the requirements, the user interaction data of all question types to be screened can be processed to achieve the training and update of the question identification model.

[0093] Additionally, it can be understood that when the number of the selected question types is not within a preset range, specifically when the constituent data of the seriously impactful questions in the selected question types does not meet the requirements, that is, when the proportion of seriously impactful questions in the selected question types is greater than a preset constituent proportion threshold, then data processing is performed on the user interaction data of all selected question types, so that when the user interaction data meets the requirements, the training and update processing of the question identification model can be realized.

[0094] Furthermore, when the constituent data of the critically impacting issues in the selected question types meet the requirements, the comprehensive interference impact value is determined based on the number of critically impacting issues and the interference impact value. When the comprehensive interference impact value is less than the preset impact threshold, only the user interaction data of the critically impacting issues needs to be processed to obtain a training set that can be used for the question identification model. When the amount of user interaction data of the critically impacting issues meets the requirements, the user interaction data of all selected question types can be processed to achieve the training and update of the question identification model.

[0095] In one possible embodiment, the overall interference impact value is determined based on the sum of the interference impact values ​​of severely impacting problems among the selected problem types.

[0096] Furthermore, when the comprehensive interference impact value is not less than the preset impact threshold, the number of updates for the screening question type in different dates within the most recent preset time period is determined. If the average number of updates for the screening question type in different dates within the most recent preset time period is greater than the preset update number threshold, then data processing is performed on the user interaction data of all screening question types. Thus, when the user interaction data meets the requirements, the training and update processing of the question identification model is realized.

[0097] Furthermore, if the average number of updates for different question types within the recent preset time period is not greater than the preset update number threshold, only the user interaction data that seriously affects the question needs to be processed to obtain a training set that can be used for the question identification model. When the amount of user interaction data that seriously affects the question meets the requirements, all user interaction data for different question types can be processed to achieve the training and update of the question identification model.

[0098] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0099] The above description is merely one or more embodiments of this specification and is not intended to limit this specification. Various modifications and variations can be made to the one or more embodiments of this specification by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of one or more embodiments of this specification should be included within the scope of the claims of this specification.

Claims

1. A data processing method based on an AI question-answering robot, characterized in that, Specifically, it includes: When it is determined that the question-answering robot cannot perform real-time update processing of the robot's question recognition model based on the question answering data of the question-answering robot, the question type that uses the knowledge base for question reply processing is determined based on the updated data of the knowledge base, and this is used as the screening question type. Based on the parsing deviation of the question type selection, the switching processing data of users between different response processing methods is determined. Based on the switching processing data, the switching impact type of the question type selection is determined. Based on the switching impact types of different question types selection and the update data of the question types selection, data processing of user interaction data of the question types selection is performed. Based on the switching impact types of different question types selection, the constituent data of seriously impacting questions in the question types selection are determined. Based on the number of seriously impacting questions and the interference impact value, the comprehensive interference impact value is determined. When the comprehensive interference impact value is less than the preset impact threshold, only the user interaction data of seriously impacting questions needs to be processed to obtain a training set that can be used for the question recognition model. The switching processing data of users between different response processing methods includes the number of switching processing between the response processing method using the question recognition model and the response processing method using the knowledge base. Based on the busy question processing period on different dates, it is determined whether the question answering robot can perform real-time update processing of the robot's question recognition model. The switching impact type includes seriously impacting questions and no-impact questions.

2. The data processing method based on an AI question-answering robot as described in claim 1, characterized in that, The question-answering data is determined based on the number of questions the question-answering robot has answered in the past.

3. The data processing method based on an AI question-answering robot as described in claim 1, characterized in that, Determining that the question-answering robot cannot perform real-time updates of the robot's question recognition model specifically includes: based on the question-answering robot's question-answering data, determining the question data answered by the question-answering robot on different dates; based on the question data answered on different dates in history, determining the busy periods for question processing on different dates; and determining whether the question-answering robot can perform real-time updates of the robot's question recognition model based on the busy periods for question processing on different dates.

4. The data processing method based on an AI question-answering robot as described in claim 1, characterized in that, The busy period for problem processing is the period when the number of problems answered exceeds a preset threshold.

5. The data processing method based on an AI question-answering robot as described in claim 1, characterized in that, If there are busy periods for problem processing on different dates, then the question-answering robot cannot perform real-time updates of the robot's problem recognition model.

6. The data processing method based on an AI question-answering robot as described in claim 1, characterized in that, The method for determining the filtering question type is as follows: based on the updated data of the knowledge base, determine the items in the knowledge base that have been updated; and use the question type corresponding to the items in the knowledge base that have been updated as the filtering question type.

7. The data processing method based on an AI question-answering robot as described in claim 1, characterized in that, By switching the impact type of different screening question types, the constituent data of the serious impact questions in the screening question types are determined; Based on the updated data of the filter question types, determine the number of updates for the filter question types on different dates within the most recent preset time period; The user interaction data for the filter question types is processed based on the number of updates to the filter question types on different dates within the most recent preset time period and the composition data of the questions that seriously affect the filter question types.

8. The data processing method based on an AI question-answering robot as described in claim 7, characterized in that, When the number of the selected problem types is within a preset range, only the user interaction data that seriously affects the problem needs to be processed.

9. A computer system, comprising: A memory and a processor connected in communication, and a computer program stored in the memory and capable of running on the processor, characterized in that, when the processor runs the computer program, it executes a data processing method based on an AI question-answering robot as described in any one of claims 1-8.

Images