CN112785001A - Artificial intelligence education counter-province robot capable of overcoming discrimination and prejudice - Google Patents
Artificial intelligence education counter-province robot capable of overcoming discrimination and prejudice Download PDFInfo
- Publication number
- CN112785001A CN112785001A CN202110228016.8A CN202110228016A CN112785001A CN 112785001 A CN112785001 A CN 112785001A CN 202110228016 A CN202110228016 A CN 202110228016A CN 112785001 A CN112785001 A CN 112785001A
- Authority
- CN
- China
- Prior art keywords
- artificial intelligence
- data
- model
- user
- school
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 323
- 238000000034 method Methods 0.000 claims abstract description 60
- 238000012552 review Methods 0.000 claims abstract description 15
- 239000013589 supplement Substances 0.000 claims abstract description 7
- 238000012986 modification Methods 0.000 claims description 14
- 230000004048 modification Effects 0.000 claims description 14
- 238000012360 testing method Methods 0.000 claims description 14
- 238000004364 calculation method Methods 0.000 claims description 9
- 238000013136 deep learning model Methods 0.000 claims description 9
- 238000012549 training Methods 0.000 claims description 9
- 238000012827 research and development Methods 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 238000010801 machine learning Methods 0.000 claims description 5
- 235000020803 food preference Nutrition 0.000 claims description 4
- 230000001502 supplementing effect Effects 0.000 claims description 4
- 238000004590 computer program Methods 0.000 claims description 3
- 230000007547 defect Effects 0.000 abstract description 11
- 230000000694 effects Effects 0.000 abstract description 11
- 238000004422 calculation algorithm Methods 0.000 abstract description 8
- 238000003062 neural network model Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/20—Education
- G06Q50/205—Education administration or guidance
Abstract
An artificial intelligence education counter-province robot for overcoming discrimination and prejudice, comprising: a data generation supplement step; retraining the model; and (5) model prediction. According to the method, the system and the robot, the problem of data unbalance of different types of users is solved by generating more data based on big data and an artificial intelligence model, so that the defects of artificial intelligence bias, discrimination and the like caused by data unbalance can be avoided; meanwhile, the user data is predicted through a plurality of artificial intelligence models, the problems are found earlier by comparing the predicted user data with the user data, and the problems existing in artificial intelligence application can be found earlier and faster by manual review, and can be corrected and remedied as soon as possible; the two aspects enable the artificial intelligence to be capable of performing inverse province on two levels of data and algorithm models, and the artificial intelligence application method is significant in many fields of artificial intelligence application, and can greatly reduce or even avoid teaching accidents and negative effects of the artificial intelligence caused by discrimination, prejudice and the like when being applied to the field of education.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to an artificial intelligence education counter-province robot capable of overcoming discrimination and prejudice.
Background
Under the prior art, the teacher can be helped to better help students to learn when artificial intelligence is used for education through earlier-stage research and discovery.
In the process of implementing the invention, the inventor finds that at least the following problems exist in the prior art: in the prior art, the algorithm adopted by artificial intelligence in the process of helping students to learn is based on big data, and the proportion of different types of samples in the big data is different, so that the students are easy to feel the sky and feel blindly, and further, learning content recommendation bringing negative effects to the learning of the students is caused, such as the students have problems of deviation, limitation to local knowledge, dependence on artificial intelligence, discrimination by artificial intelligence and the like, after the problems occur, the artificial intelligence still has nothing to know, and only the bias and discrimination are increased seriously and continuously, because the previous error or even wrong data becomes a later big data source, so that vicious circle of bias and discrimination of the artificial intelligence can be formed, and in this situation, once the algorithm used for education by the artificial intelligence has problems, the students can fall into badness and badness without self-knowing, bringing about serious teaching accidents.
Accordingly, the prior art is yet to be improved and developed.
Disclosure of Invention
Therefore, it is necessary to provide an artificial intelligence education counter-province robot for overcoming the differences and prejudices in view of the defects or shortcomings in the prior art, the artificial intelligence can perform self counter-province, further can continuously check possible errors of the robot while providing education services, and can continuously self-adjust an algorithm, so that the current defects of the artificial intelligence education robot can be overcome while addressing both symptoms and root causes.
In a first aspect, an embodiment of the present invention provides an artificial intelligence method, where the method includes:
a data generation supplement step: acquiring each user data, acquiring the proportion of various user data volumes, judging whether the proportion is unbalanced according to the proportion of each user data volume, if so, generating new each user data, and supplementing the generated new each user data into each user data;
model retraining: retraining the artificial intelligence model of each user data by using each user data;
model prediction step: respectively inputting the user data into the artificial intelligence model and a preset artificial intelligence model corresponding to the artificial intelligence model for prediction to obtain two prediction results, judging whether the difference between the two prediction results is within a preset range, and if so, returning at least one prediction result to the user; if not, requesting manual rechecking and modification of the prediction result, and requesting manual rechecking of the artificial intelligence model and another artificial intelligence model corresponding to the artificial intelligence model;
preferably, the method further comprises:
the method comprises the following steps of obtaining the data volume proportion of various users: the method comprises the steps that the type of users in big data based on artificial intelligence used for education is obtained, wherein the users comprise students or/and teachers, the proportion of the data volume of each different type of user needing to be input into each artificial intelligence connection school model (including a deep learning model, a neural network model, a quantum computing model and the like) or each different type of user meeting preset conditions to the total data volume of the users input into each artificial intelligence connection school model is counted and used as the data volume proportion of each different type of user under each artificial intelligence connection school model;
judging whether the proportion is unbalanced or not: judging whether the absolute value of the difference value between the data volume proportion of each different type of user under each artificial intelligence connection school model and the maximum value of the data volume proportion of each different type of user under each artificial intelligence connection school model is lower than a preset proportion difference threshold value which can be tolerated by each artificial intelligence connection school model or not, if so, the data volume proportion of each different type of user under each artificial intelligence connection school model meets the requirement; if not, the data volume proportion of each different type user under each artificial intelligence connection school model does not meet the requirement, and the data of each different type user is generated through a preset data first generation step or a preset data second generation step until the absolute value of the difference value between the data volume proportion of each different type user under each artificial intelligence connection school model and the maximum value of the data volume proportion of each type user under each artificial intelligence connection school model is not lower than a preset proportion difference value threshold value which can be tolerated by each artificial intelligence connection school model;
model retraining: and retraining and testing each artificial intelligence connection school model according to the data set of each different type of user under each artificial intelligence connection school model.
Preferably, the method further comprises:
a first data generation step: acquiring a user type which is closest to the type of each different type user under each artificial intelligence connection school model and has a data volume proportion meeting requirements, acquiring data of the user type which is closest to the type of each different type user under each artificial intelligence connection school model and has a data volume proportion meeting requirements, taking each data of each different type user as each first data, obtaining data which is most similar to each first data as each second data by fuzzy matching from the data of the user type which is closest to the type of the user and has a data volume proportion meeting requirements, and taking each data except the second data in the data of the user type which is closest to the type of the user and has a data volume proportion meeting requirements as each third data; initializing a deep learning model or a neural network or a machine learning model or a quantum computation model as a data prediction model; taking each second data and each corresponding first data in the training and testing samples as the input and expected output of the code prediction model, and training and testing the data prediction model; taking each third data as the input of the code prediction model, and taking the calculated output as each fourth data; adding all fourth data into the data set of each different type of user under each artificial intelligence connection school model;
a second generation step of data: the method comprises the steps of obtaining an artificial intelligence symbolic school model (such as an inference machine, a rule base, a knowledge base, an expert system and the like) capable of generating data required by each artificial intelligence connection school model, obtaining a value range of the data required by each artificial intelligence connection school model, generating a plurality of fifth data in the value range through the artificial intelligence symbolic school model, and adding the fifth data into a data set of each different type of user under each artificial intelligence connection school model.
Preferably, the method further comprises:
and (3) connecting the school model prediction step: acquiring input data to be predicted of each artificial intelligence connection school model, inputting the input data to be predicted into each artificial intelligence connection school model for calculation, and taking the obtained output as a first prediction result;
and (3) symbol school model prediction step: if the user to which the input data to be predicted belongs meets a preset condition, acquiring the input data to be predicted of each artificial intelligence connection school model, inputting the input data to be predicted into the artificial intelligence symbolic school model corresponding to each artificial intelligence connection school model for calculation, and taking the obtained output as a second prediction result;
and a result judgment step: if the user to which the input data to be predicted belongs meets a preset condition, judging whether the difference between the first prediction result and the second prediction result is within a preset credible range, and if so, taking the first prediction result as the prediction result corresponding to the input data to be predicted (mainly, the prediction result of the artificial intelligence connection school model is preferentially selected as the prediction result of the artificial intelligence connection school model is finer than the prediction result of the artificial intelligence connection school model); if not, the first prediction result and the second prediction result are sent to the teacher, information is sent to remind the teacher to refer and review, and receiving the prediction result after review or/and modification by the teacher, taking the prediction result after review or/and modification by the teacher as the prediction result corresponding to the input data to be predicted, simultaneously sending the first prediction result and the second prediction result as well as each artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model to the artificial intelligence research and development related personnel to recheck each artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model, and receiving the artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model which are rechecked or/and modified by the artificial intelligence research and development related personnel to replace each original artificial intelligence connection school model and each corresponding artificial intelligence symbolic school model.
Preferably, the method further comprises:
a symbolic school model obtaining step: acquiring artificial intelligence symbolic school models (such as inference machines, rule bases, knowledge bases, expert systems and the like) which are the same as the input items of each artificial intelligence connection school model and are matched with the output items, and taking the artificial intelligence symbolic school models as artificial intelligence symbolic school models corresponding to each artificial intelligence connection school model; the output item matching comprises the steps that the types of the output items are the same but the required precision is not necessarily the same, or the output items of the artificial intelligence connection school model represent specific numerical values but the output items of the artificial intelligence symbolic school model represent value ranges; or the output items of the artificial intelligence joint school model represent quantitative data while the output items of the artificial intelligence symbolic school model represent qualitative data.
Preferably, the method further comprises:
a preset condition setting step: the preset conditions comprise one or more preset conditions of poor learning achievement, the problem of learning 'food preference', dissatisfaction with artificial intelligence service (explaining that the artificial intelligence service has defects), discrimination of feedback or complaint of the artificial intelligence service, too frequent use of the artificial intelligence service (explaining that the artificial intelligence service may depend on the artificial intelligence service and can be judged by using a preset threshold of frequency), and very few use of the artificial intelligence service (explaining that the artificial intelligence service may have defects and can be judged by using a preset threshold of frequency).
In a second aspect, an embodiment of the present invention provides an artificial intelligence system, where the system includes:
a data generation supplement module: acquiring each user data, acquiring the proportion of various user data volumes, judging whether the proportion is unbalanced according to the proportion of each user data volume, if so, generating new each user data, and supplementing the generated new each user data into each user data;
a model retraining module: retraining the artificial intelligence model of each user data by using each user data;
a model prediction module: respectively inputting the user data into the artificial intelligence model and a preset artificial intelligence model corresponding to the artificial intelligence model for prediction to obtain two prediction results, judging whether the difference between the two prediction results is within a preset range, and if so, returning at least one prediction result to the user; and if not, requesting manual rechecking and modification of the prediction result, and requesting manual rechecking of the artificial intelligence model and another artificial intelligence model corresponding to the artificial intelligence model.
Preferably, the system further comprises:
the data volume proportion acquisition module for various users: the method comprises the steps that the type of users in big data based on artificial intelligence used for education is obtained, wherein the users comprise students or/and teachers, the proportion of the data volume of each different type of user needing to be input into each artificial intelligence connection school model (including a deep learning model, a neural network model, a quantum computing model and the like) or each different type of user meeting preset conditions to the total data volume of the users input into each artificial intelligence connection school model is counted and used as the data volume proportion of each different type of user under each artificial intelligence connection school model;
the proportion unbalance judging and processing module: judging whether the absolute value of the difference value between the data volume proportion of each different type of user under each artificial intelligence connection school model and the maximum value of the data volume proportion of each different type of user under each artificial intelligence connection school model is lower than a preset proportion difference threshold value which can be tolerated by each artificial intelligence connection school model or not, if so, the data volume proportion of each different type of user under each artificial intelligence connection school model meets the requirement; if not, the data volume proportion of each different type of user under each artificial intelligence connection school model does not meet the requirement, and data of each different type of user is generated through a preset data first generation module or a preset data second generation module until the absolute value of the difference value between the data volume proportion of each different type of user under each artificial intelligence connection school model and the maximum value of the data volume proportion of each type of user under each artificial intelligence connection school model is not lower than a preset proportion difference value threshold value which can be tolerated by each artificial intelligence connection school model;
a model retraining module: and retraining and testing each artificial intelligence connection school model according to the data set of each different type of user under each artificial intelligence connection school model.
Preferably, the system further comprises:
a first data generation module: acquiring a user type which is closest to the type of each different type user under each artificial intelligence connection school model and has a data volume proportion meeting requirements, acquiring data of the user type which is closest to the type of each different type user under each artificial intelligence connection school model and has a data volume proportion meeting requirements, taking each data of each different type user as each first data, obtaining data which is most similar to each first data as each second data by fuzzy matching from the data of the user type which is closest to the type of the user and has a data volume proportion meeting requirements, and taking each data except the second data in the data of the user type which is closest to the type of the user and has a data volume proportion meeting requirements as each third data; initializing a deep learning model or a neural network or a machine learning model or a quantum computation model as a data prediction model; taking each second data and each corresponding first data in the training and testing samples as the input and expected output of the code prediction model, and training and testing the data prediction model; taking each third data as the input of the code prediction model, and taking the calculated output as each fourth data; adding all fourth data into the data set of each different type of user under each artificial intelligence connection school model;
a second generation module of data: the method comprises the steps of obtaining an artificial intelligence symbolic school model (such as an inference machine, a rule base, a knowledge base, an expert system and the like) capable of generating data required by each artificial intelligence connection school model, obtaining a value range of the data required by each artificial intelligence connection school model, generating a plurality of fifth data in the value range through the artificial intelligence symbolic school model, and adding the fifth data into a data set of each different type of user under each artificial intelligence connection school model.
Preferably, the system further comprises:
connecting the school model prediction module: acquiring input data to be predicted of each artificial intelligence connection school model, inputting the input data to be predicted into each artificial intelligence connection school model for calculation, and taking the obtained output as a first prediction result;
the symbolic school model prediction module: if the user to which the input data to be predicted belongs meets a preset condition, acquiring the input data to be predicted of each artificial intelligence connection school model, inputting the input data to be predicted into the artificial intelligence symbolic school model corresponding to each artificial intelligence connection school model for calculation, and taking the obtained output as a second prediction result;
and a result judgment module: if the user to which the input data to be predicted belongs meets a preset condition, judging whether the difference between the first prediction result and the second prediction result is within a preset credible range, and if so, taking the first prediction result as the prediction result corresponding to the input data to be predicted (mainly, the prediction result of the artificial intelligence connection school model is preferentially selected as the prediction result of the artificial intelligence connection school model is finer than the prediction result of the artificial intelligence connection school model); if not, the first prediction result and the second prediction result are sent to the teacher, information is sent to remind the teacher to refer and review, and receiving the prediction result after review or/and modification by the teacher, taking the prediction result after review or/and modification by the teacher as the prediction result corresponding to the input data to be predicted, simultaneously sending the first prediction result and the second prediction result as well as each artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model to the artificial intelligence research and development related personnel to recheck each artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model, and receiving the artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model which are rechecked or/and modified by the artificial intelligence research and development related personnel to replace each original artificial intelligence connection school model and each corresponding artificial intelligence symbolic school model.
Preferably, the system further comprises:
the symbolic school model acquisition module: acquiring artificial intelligence symbolic school models (such as inference machines, rule bases, knowledge bases, expert systems and the like) which are the same as the input items of each artificial intelligence connection school model and are matched with the output items, and taking the artificial intelligence symbolic school models as artificial intelligence symbolic school models corresponding to each artificial intelligence connection school model; the output item matching comprises the steps that the types of the output items are the same but the required precision is not necessarily the same, or the output items of the artificial intelligence connection school model represent specific numerical values but the output items of the artificial intelligence symbolic school model represent value ranges; or the output items of the artificial intelligence joint school model represent quantitative data while the output items of the artificial intelligence symbolic school model represent qualitative data.
Preferably, the system further comprises:
the preset condition setting module: the preset conditions comprise one or more preset conditions of poor learning achievement, the problem of learning 'food preference', dissatisfaction with artificial intelligence service (explaining that the artificial intelligence service has defects), discrimination of feedback or complaint of the artificial intelligence service, too frequent use of the artificial intelligence service (explaining that the artificial intelligence service may depend on the artificial intelligence service and can be judged by using a preset threshold of frequency), and very few use of the artificial intelligence service (explaining that the artificial intelligence service may have defects and can be judged by using a preset threshold of frequency).
In a third aspect, an embodiment of the present invention provides an artificial intelligence apparatus, where the apparatus includes the modules of the system in any one of the embodiments of the second aspect.
In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, implements the steps of the method according to any one of the embodiments of the first aspect.
In a fifth aspect, an embodiment of the present invention provides a robot, including a memory, a processor, and an artificial intelligence robot program stored in the memory and executable on the processor, where the processor executes the program to implement the steps of the method according to any one of the embodiments of the first aspect.
The artificial intelligence education counter-province robot for overcoming the discrimination and prejudice provided by the embodiment comprises: a data generation supplement step; retraining the model; and (5) model prediction. According to the method, the system and the robot, the problem of data unbalance of different types of users is solved by generating more data based on big data and an artificial intelligence model, so that the defects of artificial intelligence bias, discrimination and the like caused by data unbalance can be avoided; meanwhile, the user data is predicted through a plurality of artificial intelligence models, the problems are found earlier by comparing the predicted user data with the user data, and the problems existing in artificial intelligence application can be found earlier and faster by manual review, and can be corrected and remedied as soon as possible; the two aspects enable the artificial intelligence to be capable of performing inverse province on two levels of data and algorithm models, and the artificial intelligence application method is significant in many fields of artificial intelligence application, and can greatly reduce or even avoid teaching accidents and negative effects of the artificial intelligence caused by discrimination, prejudice and the like when being applied to the field of education.
Drawings
FIG. 1 is a flow diagram of an artificial intelligence method provided by one embodiment of the invention;
FIG. 2 is a flow diagram of an artificial intelligence method comprising according to an embodiment of the invention;
FIG. 3 is a flow chart of an artificial intelligence method comprising according to an embodiment of the invention;
FIG. 4 is a flow chart of an artificial intelligence method according to an embodiment of the present invention.
Detailed Description
The technical solutions in the examples of the present invention are described in detail below with reference to the embodiments of the present invention.
Basic embodiment of the invention
One embodiment of the present invention provides an artificial intelligence method, as shown in fig. 1, the method including: a data generation supplement step; retraining the model; and (5) model prediction. The technical effects are as follows: the method compensates the data unbalance problem of different types of users by generating more data based on big data and an artificial intelligence model, thereby avoiding the defects of artificial intelligence prejudice, discrimination and the like caused by data unbalance; meanwhile, the user data is predicted through a plurality of artificial intelligence models, the problems are found earlier by comparing the predicted user data with the user data, and the problems existing in artificial intelligence application can be found earlier and faster by manual review, and can be corrected and remedied as soon as possible; the two aspects enable the artificial intelligence to be capable of performing inverse province on two levels of data and algorithm models, and the artificial intelligence application method is significant in many fields of artificial intelligence application, and can greatly reduce or even avoid teaching accidents and negative effects of the artificial intelligence caused by discrimination, prejudice and the like when being applied to the field of education.
In a preferred embodiment, as shown in fig. 2, the method further comprises: acquiring the data volume proportion of various users; judging whether the proportion is unbalanced; and (5) retraining the model. The technical effects are as follows: the method is based on big data and a connection school artificial intelligence model, the user type with insufficient data volume is found in time through comparison of data volume proportions of different types of users, the connection school artificial intelligence model is retrained after data is supplemented, therefore, the problem that the connection school artificial intelligence model, particularly the currently most popular deep learning model, is judged by mistake due to different data volumes of different types of users can be greatly reduced, the existing models such as a human face recognition model and a crime prediction model based on deep learning have prejudice and discrimination on certain ethnicities, certain sexes, certain professions and the like, and if the prejudice and discrimination exist when artificial intelligence is applied to education, great hidden dangers are brought to teaching.
In a preferred embodiment, as shown in fig. 3, the method further comprises: a first data generation step; and a second generation step of data. The technical effects are as follows: the method generates more user data based on the connection school artificial intelligence model or the symbolic school artificial intelligence model to make up for the problem of insufficient data quantity of part of users, greatly reduces the cost of data acquisition by automatically generating data, provides a feasible solution for the data lacking acquisition sources, and has great significance for eliminating artificial intelligence algorithm and application risk caused by insufficient data quantity or unbalanced data.
In a preferred embodiment, as shown in fig. 4, the method further comprises: connecting the school model prediction step; predicting a symbolic school model; and judging a result. The technical effects are as follows: the method is based on the connection school and the symbolic school artificial intelligence, whether the prediction result is credible or not is judged by comparing the prediction results of two different school models, namely two experts can consult, if the diagnosis results of the two experts are basically consistent and are necessarily credible, otherwise, the result of artificial intelligence processing is not necessarily credible, and manual recheck is needed, so that the problems existing in the artificial intelligence algorithm and the processing result can be found in time, and the credibility of the artificial intelligence is greatly improved.
In a preferred embodiment, the method further comprises: and (5) obtaining a symbolic school model. The technical effects are as follows: the method is based on the symbolic school artificial intelligence model, the school artificial intelligence model and the symbolic school artificial intelligence model can be matched through comparison and fuzzy matching of input items and output items, if data in the same format is input, output data with compatible formats can be obtained, two different output formats, namely quantitative output format and qualitative output format, can be compatible, and therefore prediction through the two different school models becomes possible.
In a preferred embodiment, the method further comprises: and setting a preset condition. The technical effects are as follows: the method reduces the calculation amount through the preset conditions based on the big data, only detects whether the data of the user type which is possibly unbalanced is unbalanced, and only predicts and compares the user type which is possibly mispredicted for 2 times, so that the method not only can greatly reduce the effect of computing resources and improve the computing efficiency and speed, but also basically covers all the user types which are possibly problematic by the preset conditions, and can more quickly confirm and screen the user type with the problems and timely make up and correct the corresponding data and the artificial intelligence model.
PREFERRED EMBODIMENTS OF THE PRESENT INVENTION
A data generation supplement step: acquiring each user data, acquiring the proportion of various user data volumes, judging whether the proportion is unbalanced according to the proportion of each user data volume, if so, generating new each user data, and supplementing the generated new each user data into each user data;
model retraining: retraining the artificial intelligence model of each user data by using each user data;
model prediction step: respectively inputting the user data into the artificial intelligence model and a preset artificial intelligence model corresponding to the artificial intelligence model for prediction to obtain two prediction results, judging whether the difference between the two prediction results is within a preset range, and if so, returning at least one prediction result to the user; if not, requesting manual rechecking and modification of the prediction result, and requesting manual rechecking of the artificial intelligence model and another artificial intelligence model corresponding to the artificial intelligence model;
the method comprises the following steps of obtaining the data volume proportion of various users: the method comprises the steps of obtaining the type of users in big data based on artificial intelligence used for education, wherein the users comprise students or/and teachers, and counting the proportion of the data volume of each different type of user needing to be input into each artificial intelligence connection school model (comprising a deep learning model, a neural network model, a quantum computing model and the like) or each different type of user meeting preset conditions to the total data volume of the users input into each artificial intelligence connection school model, wherein the data volume is used as the data volume proportion of each different type of user under each artificial intelligence connection school model.
A preset condition setting step: the preset conditions comprise one or more preset conditions of poor learning achievement, the problem of learning 'food preference', dissatisfaction with artificial intelligence service (explaining that the artificial intelligence service has defects), discrimination of feedback or complaint of the artificial intelligence service, too frequent use of the artificial intelligence service (explaining that the artificial intelligence service may depend on the artificial intelligence service and can be judged by using a preset threshold of frequency), and very few use of the artificial intelligence service (explaining that the artificial intelligence service may have defects and can be judged by using a preset threshold of frequency).
Judging whether the proportion is unbalanced or not: judging whether the absolute value of the difference value between the data volume proportion of each different type of user under each artificial intelligence connection school model and the maximum value of the data volume proportion of each different type of user under each artificial intelligence connection school model is lower than a preset proportion difference threshold value which can be tolerated by each artificial intelligence connection school model or not, if so, the data volume proportion of each different type of user under each artificial intelligence connection school model meets the requirement; if not, the data volume proportion of each different type user under each artificial intelligence connection school model does not meet the requirement, and the data of each different type user is generated through a preset data first generation step or a preset data second generation step until the absolute value of the difference value between the data volume proportion of each different type user under each artificial intelligence connection school model and the maximum value of the data volume proportion of each type user under each artificial intelligence connection school model is not lower than a preset proportion difference value threshold value which can be tolerated by each artificial intelligence connection school model;
a first data generation step: acquiring a user type which is closest to the type of each different type user under each artificial intelligence connection school model and has a data volume proportion meeting requirements, acquiring data of the user type which is closest to the type of each different type user under each artificial intelligence connection school model and has a data volume proportion meeting requirements, taking each data of each different type user as each first data, obtaining data which is most similar to each first data as each second data by fuzzy matching from the data of the user type which is closest to the type of the user and has a data volume proportion meeting requirements, and taking each data except the second data in the data of the user type which is closest to the type of the user and has a data volume proportion meeting requirements as each third data; initializing a deep learning model or a neural network or a machine learning model or a quantum computation model as a data prediction model; taking each second data and each corresponding first data in the training and testing samples as the input and expected output of the code prediction model, and training and testing the data prediction model; taking each third data as the input of the code prediction model, and taking the calculated output as each fourth data; adding all fourth data into the data set of each different type of user under each artificial intelligence connection school model;
a second generation step of data: acquiring an artificial intelligence symbolic school model (such as an inference machine, a rule base, a knowledge base, an expert system and the like) capable of generating data required by each artificial intelligence connection school model, acquiring a value range of the data required by each artificial intelligence connection school model, generating a plurality of fifth data in the value range through the artificial intelligence symbolic school model, and adding the fifth data into a data set of each different type of user under each artificial intelligence connection school model;
model retraining: retraining and testing each artificial intelligence connection school model according to the data set of each different type of user under each artificial intelligence connection school model;
and (3) connecting the school model prediction step: acquiring input data to be predicted of each artificial intelligence connection school model, inputting the input data to be predicted into each artificial intelligence connection school model for calculation, and taking the obtained output as a first prediction result;
a symbolic school model obtaining step: acquiring artificial intelligence symbolic school models (such as inference machines, rule bases, knowledge bases, expert systems and the like) which are the same as the input items of each artificial intelligence connection school model and are matched with the output items, and taking the artificial intelligence symbolic school models as artificial intelligence symbolic school models corresponding to each artificial intelligence connection school model; the output item matching comprises the steps that the types of the output items are the same but the required precision is not necessarily the same, or the output items of the artificial intelligence connection school model represent specific numerical values but the output items of the artificial intelligence symbolic school model represent value ranges; or the output item of the artificial intelligence connection school model represents quantitative data, but the output item of the artificial intelligence symbolic school model represents qualitative data;
and (3) symbol school model prediction step: if the user to which the input data to be predicted belongs meets a preset condition, acquiring the input data to be predicted of each artificial intelligence connection school model, inputting the input data to be predicted into the artificial intelligence symbolic school model corresponding to each artificial intelligence connection school model for calculation, and taking the obtained output as a second prediction result;
and a result judgment step: if the user to which the input data to be predicted belongs meets a preset condition, judging whether the difference between the first prediction result and the second prediction result is within a preset credible range, and if so, taking the first prediction result as the prediction result corresponding to the input data to be predicted (mainly, the prediction result of the artificial intelligence connection school model is preferentially selected as the prediction result of the artificial intelligence connection school model is finer than the prediction result of the artificial intelligence connection school model); if not, the first prediction result and the second prediction result are sent to the teacher, information is sent to remind the teacher to refer and review, and receiving the prediction result after review or/and modification by the teacher, taking the prediction result after review or/and modification by the teacher as the prediction result corresponding to the input data to be predicted, simultaneously sending the first prediction result and the second prediction result as well as each artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model to the artificial intelligence research and development related personnel to recheck each artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model, and receiving the artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model which are rechecked or/and modified by the artificial intelligence research and development related personnel to replace each original artificial intelligence connection school model and each corresponding artificial intelligence symbolic school model.
A trusted judgment model construction step: initializing a deep learning model or a neural network or a machine learning model or a quantum computing model as a credible judgment model; obtaining a judgment result of whether a first prediction result and a second prediction result in a training and testing sample and a difference between the first prediction result and the second prediction result are within a preset credibility range, taking the first prediction result and the second prediction result as input, and taking a judgment result of whether the difference between the first prediction result and the second prediction result is within the preset credibility range as output to train and test a credibility judgment model;
the credible judgment model using step: and acquiring a first prediction result and a second prediction result to be judged, inputting the first prediction result and the second prediction result into the credibility judgment model, and calculating the obtained output to be used as a result for judging whether the difference between the first prediction result and the second prediction result is within a preset credibility range.
The above-mentioned embodiments only express several embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present invention. It should be noted that, for those skilled in the art, various changes and modifications can be made without departing from the spirit of the present invention, and these changes and modifications are within the scope of the present invention. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (10)
1. An artificial intelligence method, the method comprising:
a data generation supplement step: acquiring each user data, acquiring the proportion of various user data volumes, judging whether the proportion is unbalanced according to the proportion of each user data volume, if so, generating new each user data, and supplementing the generated new each user data into each user data;
model retraining: retraining the artificial intelligence model of each user data by using each user data;
model prediction step: respectively inputting the user data into the artificial intelligence model and a preset artificial intelligence model corresponding to the artificial intelligence model for prediction to obtain two prediction results, judging whether the difference between the two prediction results is within a preset range, and if so, returning at least one prediction result to the user; and if not, requesting manual rechecking and modification of the prediction result, and requesting manual rechecking of the artificial intelligence model and another artificial intelligence model corresponding to the artificial intelligence model.
2. The artificial intelligence method of claim 1, wherein the method further comprises:
the method comprises the following steps of obtaining the data volume proportion of various users: the method comprises the steps that the type of users in big data based on artificial intelligence used for education is obtained, the users comprise students or/and teachers, the proportion of the data volume of each different type of user needing to be input into each artificial intelligence connection school model or each different type of user meeting preset conditions to the total data volume of the users input into each artificial intelligence connection school model is counted, and the data volume proportion of each different type of user under each artificial intelligence connection school model is used as the data volume proportion of each different type of user;
judging whether the proportion is unbalanced or not: judging whether the absolute value of the difference value between the data volume proportion of each different type of user under each artificial intelligence connection school model and the maximum value of the data volume proportion of each different type of user under each artificial intelligence connection school model is lower than a preset proportion difference threshold value which can be tolerated by each artificial intelligence connection school model or not, if so, the data volume proportion of each different type of user under each artificial intelligence connection school model meets the requirement; if not, the data volume proportion of each different type user under each artificial intelligence connection school model does not meet the requirement, and the data of each different type user is generated through a preset data first generation step or a preset data second generation step until the absolute value of the difference value between the data volume proportion of each different type user under each artificial intelligence connection school model and the maximum value of the data volume proportion of each type user under each artificial intelligence connection school model is not lower than a preset proportion difference value threshold value which can be tolerated by each artificial intelligence connection school model;
model retraining: and retraining and testing each artificial intelligence connection school model according to the data set of each different type of user under each artificial intelligence connection school model.
3. The artificial intelligence method of claim 2, wherein the method further comprises:
a first data generation step: acquiring a user type which is closest to the type of each different type user under each artificial intelligence connection school model and has a data volume proportion meeting requirements, acquiring data of the user type which is closest to the type of each different type user under each artificial intelligence connection school model and has a data volume proportion meeting requirements, taking each data of each different type user as each first data, obtaining data which is most similar to each first data as each second data by fuzzy matching from the data of the user type which is closest to the type of the user and has a data volume proportion meeting requirements, and taking each data except the second data in the data of the user type which is closest to the type of the user and has a data volume proportion meeting requirements as each third data; initializing a deep learning model or a neural network or a machine learning model or a quantum computation model as a data prediction model; taking each second data and each corresponding first data in the training and testing samples as the input and expected output of the code prediction model, and training and testing the data prediction model; taking each third data as the input of the code prediction model, and taking the calculated output as each fourth data; adding all fourth data into the data set of each different type of user under each artificial intelligence connection school model;
a second generation step of data: the method comprises the steps of obtaining an artificial intelligence symbolic school model capable of generating data required by each artificial intelligence connection school model, obtaining a value range of the data required by each artificial intelligence connection school model, generating a plurality of fifth data in the value range through the artificial intelligence symbolic school model, and adding the fifth data into a data set of each different type of user under each artificial intelligence connection school model.
4. The artificial intelligence method of claim 1, wherein the method further comprises:
and (3) connecting the school model prediction step: acquiring input data to be predicted of each artificial intelligence connection school model, inputting the input data to be predicted into each artificial intelligence connection school model for calculation, and taking the obtained output as a first prediction result;
and (3) symbol school model prediction step: if the user to which the input data to be predicted belongs meets a preset condition, acquiring the input data to be predicted of each artificial intelligence connection school model, inputting the input data to be predicted into the artificial intelligence symbolic school model corresponding to each artificial intelligence connection school model for calculation, and taking the obtained output as a second prediction result;
and a result judgment step: if the user to which the input data to be predicted belongs meets a preset condition, judging whether the difference between the first prediction result and the second prediction result is within a preset credible range, and if so, taking the first prediction result as the prediction result corresponding to the input data to be predicted; if not, the first prediction result and the second prediction result are sent to the teacher, information is sent to remind the teacher to refer and review, and receiving the prediction result after review or/and modification by the teacher, taking the prediction result after review or/and modification by the teacher as the prediction result corresponding to the input data to be predicted, simultaneously sending the first prediction result and the second prediction result as well as each artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model to the artificial intelligence research and development related personnel to recheck each artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model, and receiving the artificial intelligence connection school model and the corresponding artificial intelligence symbolic school model which are rechecked or/and modified by the artificial intelligence research and development related personnel to replace each original artificial intelligence connection school model and each corresponding artificial intelligence symbolic school model.
5. The artificial intelligence method of claim 4, wherein the method further comprises:
a symbolic school model obtaining step: acquiring an artificial intelligence symbolic school model which is the same as the input item of each artificial intelligence connection school model and is matched with the output item of each artificial intelligence connection school model, and taking the artificial intelligence symbolic school model as the artificial intelligence symbolic school model corresponding to each artificial intelligence connection school model; the output item matching comprises the steps that the types of the output items are the same but the required precision is not necessarily the same, or the output items of the artificial intelligence connection school model represent specific numerical values but the output items of the artificial intelligence symbolic school model represent value ranges; or the output items of the artificial intelligence joint school model represent quantitative data while the output items of the artificial intelligence symbolic school model represent qualitative data.
6. The artificial intelligence method of claim 2 or 4, wherein the method further comprises:
a preset condition setting step: the preset conditions comprise one or more preset conditions of poor learning achievement, the problem of learning 'food preference', dissatisfaction with artificial intelligence service, discrimination of feedback or complaint artificial intelligence, too frequent use of artificial intelligence service and extremely less use of artificial intelligence service.
7. An artificial intelligence system, characterized in that the system is adapted to implement the steps of the method of any of claims 1-6.
8. An artificial intelligence device, wherein the device is configured to implement the steps of the method of any of claims 1-6.
9. A robot comprising a memory, a processor and an artificial intelligence robot program stored on the memory and executable on the processor, wherein the steps of the method of any one of claims 1 to 6 are carried out when the program is executed by the processor.
10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110228016.8A CN112785001B (en) | 2021-03-02 | 2021-03-02 | Artificial intelligence educational back-province robot for overcoming discrimination and prejudice |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110228016.8A CN112785001B (en) | 2021-03-02 | 2021-03-02 | Artificial intelligence educational back-province robot for overcoming discrimination and prejudice |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112785001A true CN112785001A (en) | 2021-05-11 |
| CN112785001B CN112785001B (en) | 2024-01-05 |
Family
ID=75762091
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110228016.8A Active CN112785001B (en) | 2021-03-02 | 2021-03-02 | Artificial intelligence educational back-province robot for overcoming discrimination and prejudice |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112785001B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113420878A (en) * | 2021-07-09 | 2021-09-21 | 浙江工业大学 | Deep learning model depolarization method based on bidirectional feature embedding |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10303978B1 (en) * | 2018-03-26 | 2019-05-28 | Clinc, Inc. | Systems and methods for intelligently curating machine learning training data and improving machine learning model performance |
| WO2020140619A1 (en) * | 2019-01-03 | 2020-07-09 | 平安科技（深圳）有限公司 | Machine learning-based system upgrade failure diagnosis method and apparatus, computer device, and computer readable storage medium |
| CN111860577A (en) * | 2020-06-08 | 2020-10-30 | 华南师范大学 | Artificial intelligence ethical method for identifying human being harmless to human being and robot |
| CN111860133A (en) * | 2020-06-08 | 2020-10-30 | 华南师范大学 | Artificial intelligence ethical method and robot for identifying human without ethnic bias |
| CN112149837A (en) * | 2020-08-04 | 2020-12-29 | 华南师范大学 | Artificial intelligence ethical risk identification and prevention method based on algorithm selection and robot |
-
2021
- 2021-03-02 CN CN202110228016.8A patent/CN112785001B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10303978B1 (en) * | 2018-03-26 | 2019-05-28 | Clinc, Inc. | Systems and methods for intelligently curating machine learning training data and improving machine learning model performance |
| WO2020140619A1 (en) * | 2019-01-03 | 2020-07-09 | 平安科技（深圳）有限公司 | Machine learning-based system upgrade failure diagnosis method and apparatus, computer device, and computer readable storage medium |
| CN111860577A (en) * | 2020-06-08 | 2020-10-30 | 华南师范大学 | Artificial intelligence ethical method for identifying human being harmless to human being and robot |
| CN111860133A (en) * | 2020-06-08 | 2020-10-30 | 华南师范大学 | Artificial intelligence ethical method and robot for identifying human without ethnic bias |
| CN112149837A (en) * | 2020-08-04 | 2020-12-29 | 华南师范大学 | Artificial intelligence ethical risk identification and prevention method based on algorithm selection and robot |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113420878A (en) * | 2021-07-09 | 2021-09-21 | 浙江工业大学 | Deep learning model depolarization method based on bidirectional feature embedding |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112785001B (en) | 2024-01-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| De La Torre | An empirically based method of Q‐matrix validation for the DINA model: Development and applications | |
| Thai-Nghe et al. | Factorization Models for Forecasting Student Performance. | |
| CN110516116B (en) | Multi-step hierarchical learner cognitive level mining method and system | |
| Chang et al. | A Bayes net toolkit for student modeling in intelligent tutoring systems | |
| Sao Pedro et al. | Leveraging machine-learned detectors of systematic inquiry behavior to estimate and predict transfer of inquiry skill | |
| US20190026656A1 (en) | System and method for defining and using different levels of ground truth | |
| US10354544B1 (en) | Predicting student proficiencies in knowledge components | |
| CN111274411A (en) | Course recommendation method and device, electronic equipment and readable storage medium | |
| CN113779272B (en) | Knowledge graph-based data processing method, device, equipment and storage medium | |
| WO2023071121A1 (en) | Multi-model fusion-based object detection method and apparatus, device and medium | |
| CN111144079B (en) | Method and device for intelligently acquiring learning resources, printer and storage medium | |
| CN114254208A (en) | Identification method of weak knowledge points and planning method and device of learning path | |
| Chan et al. | Question-answering dialogue system for emergency operations | |
| CN113887930B (en) | Question-answering robot health evaluation method, device, equipment and storage medium | |
| CN114281957A (en) | Natural language data query method and device, electronic equipment and storage medium | |
| CN111126610A (en) | Topic analysis method, topic analysis device, electronic device and storage medium | |
| CN112785001A (en) | Artificial intelligence education counter-province robot capable of overcoming discrimination and prejudice | |
| CN115114974A (en) | Model distillation method, device, computer equipment and storage medium | |
| CN113283488B (en) | Learning behavior-based cognitive diagnosis method and system | |
| Sadanand et al. | An automated essay evaluation system using natural language processing and sentiment analysis | |
| Díaz-Francés et al. | The simplicity of likelihood based inferences for P (X< Y) and for the ratio of means in the exponential model | |
| CN112419812A (en) | Exercise correction method and device | |
| CN112528011A (en) | Open type mathematic operation correction method, system and equipment driven by multiple data sources | |
| Aguagallo et al. | Analysis of student performance applying data mining techniques in a virtual learning environment | |
| CN116431821A (en) | Knowledge graph completion method and question-answering system based on common sense perception |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |