CN112116187A - Method for dynamically optimizing expression evaluation questions - Google Patents
Method for dynamically optimizing expression evaluation questions Download PDFInfo
- Publication number
- CN112116187A CN112116187A CN202010254255.6A CN202010254255A CN112116187A CN 112116187 A CN112116187 A CN 112116187A CN 202010254255 A CN202010254255 A CN 202010254255A CN 112116187 A CN112116187 A CN 112116187A
- Authority
- CN
- China
- Prior art keywords
- avg
- question
- parameter
- optimizing
- samples
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 27
- 238000011156 evaluation Methods 0.000 title claims abstract description 26
- 230000006870 function Effects 0.000 claims description 15
- 238000005457 optimization Methods 0.000 claims description 8
- 238000011478 gradient descent method Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 description 2
- 239000004744 fabric Substances 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
Images
Classifications
-
- 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
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
- G06Q10/06393—Score-carding, benchmarking or key performance indicator [KPI] analysis
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Strategic Management (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Educational Administration (AREA)
- Operations Research (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention belongs to the technical field of statistical learning, and relates to a method for dynamically adjusting and automatically optimizing subject parameters in expression evaluation, in particular to a method for dynamically optimizing expression evaluation subjects, which improves the accuracy of evaluation and appraisal and dynamically adjusts the difficulty of evaluation while increasing the number of samples by applying a statistical learning method, and comprises the following specific steps: 1) for a question qjD represents the difficulty of the question, b represents a word stock of the question, w represents a weight parameter of the word stock, and l represents the discrimination of the question; 2) optimizing the parameter d; 3) and optimizing the parameter w. According to the method, a statistical learning method is used, useful information is extracted from the existing sample data which is continuously added to optimize the topic parameters, so that the topic model is more robust, and sensitive feedback is provided for different expressive force evaluation inputs; by using the unsupervised learning model, extra marking work is not needed in the learning process, the workload of experts is reduced, and the labor cost is reduced.
Description
Technical Field
The invention belongs to the technical field of statistical learning, and particularly relates to a method for dynamically optimizing expression evaluation questions.
Background
The rapid expression test is a novel spoken language test mode, and compared with the traditional Chinese language test, the rapid expression test has the advantages of wide application range, various question sources, rapid test and objective evaluation. However, each attribute of the current theme is specified by experts according to own experience during initialization, and the judgment may be inaccurate. Aiming at the problem, the invention provides a method which can adjust the subject parameters unsupervised through the continuous increase of the number of samples and can improve the accuracy of the judgment of the subjects to a greater extent.
Disclosure of Invention
The invention aims to provide a method for dynamically optimizing an expressive force evaluation topic so as to solve the problems in the background technology.
In order to achieve the purpose, the invention provides the following technical scheme:
a method for dynamically optimizing an expression evaluation question applies a statistical learning method to increase the number of samples and simultaneously improve the accuracy of evaluation and appraisal and dynamically adjust the difficulty of the evaluation and appraisal, and comprises the following specific steps:
1) for a question qjD represents the difficulty of the question, b represents the word stock of the question, w represents the weight parameter of the word stock, l represents the distinguishing degree of the question, and all samples of the word stock form a sample set S ═ S1,s2,s3...,sn]Wherein s isi={hi,sciDenotes the ith sample, hiIs its hit vector, sciRepresents its score;
2) optimizing the parameter d;
3) and optimizing the parameter w.
As a preferred aspect of the present invention, the specific steps of the optimization of the parameter d are as follows:
a) extracting all question sets of the same type Q ═ Q1,q2,q3...qnExtracting an average set avgs ═ avg of the questions1,avg2,avg3...,avgn};
b) According to the question qjAverage score avg ofjThe difficulty can be adjusted at the position in the set avgs, if the average score is higher, the difficulty is lower, otherwise, the difficulty is higher.
As a preferred aspect of the present invention, the specific steps of the optimization of the parameter w are as follows:
a) sorting the sample set S according to the sample score sc, wherein the sorted sample set is S';
b) generating a random normal distribution achievement set sc' by taking the average avg and the standard deviation sd of the achievement sets sc of all samples as parameters;
c) by adjusting the weight parameter w, the sc and sc' are fitted by a gradient descent method.
As a preferred aspect of the present invention, sc is obtained in step 1)iAnd sc are respectively:
sc=w·hT。
as a preferred aspect of the present invention, the topic q in step 2)jThe difficulty adjustment formula is as follows:
the near function maps the values to three numbers of-1, 0 and 1 nearby, the avg function calculates the average value of a set, and the sd function calculates the standard deviation of the set.
As a preferred aspect of the present invention, the loss function in step 3) is formulated as follows:
J(w)=L2(w·hT-sc′)+w2,
wherein L is2The function represents the L2 distance of a vector to the origin.
As a preferred aspect of the present invention, the specific operation steps of the gradient descent method in step 3) are as follows:
a) selecting one distribution curve of the weight parameters w;
b) obtaining the achievement set sc of all samples in the distribution curve according to the selected distribution curve1;
c) Gathering sc with the achievements of all samples1Average avg of1And standard deviation sd1As parameters, a set of random normal distribution results sc 'is generated'1;
d) Selecting another distribution curve of the weight parameter w;
e) obtaining the achievement set sc of all samples in the distribution curve according to the selected distribution curve2;
f) Gathering sc with the achievements of all samples2Average avg of2And standard deviation sd2As parameters, a set of random normal distribution results sc 'is generated'2;
g) Repeating the above six steps to obtain the sample score sc1,sc2,sc3,...scnAnd average number avg1,avg2,avg3...avgnAnd standard deviation sd1,sd2,sd3,...sdnAnd generating a random normal distribution score set sc 'from the obtained data'n。
The invention has the advantages that:
1. according to the method, a statistical learning method is used, useful information is extracted from the existing sample data which is continuously added, and the topic parameters are optimized, so that the topic model is more robust, and sensitive feedback is provided for different expressive force evaluation inputs;
2. by using the unsupervised learning model, the invention does not need additional labeling work in the learning process, reduces the workload of experts and lowers the labor cost.
Drawings
FIG. 1 is a schematic diagram of difficulty parameter optimization in the present invention;
FIG. 2 is a diagram illustrating the optimization of thesaurus weight vector parameters in the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be described clearly and completely with reference to the following embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The technical scheme provided by the invention is as follows:
a method for dynamically optimizing an expression evaluation question applies a statistical learning method to increase the number of samples and simultaneously improve the accuracy of evaluation and appraisal and dynamically adjust the difficulty of the evaluation and appraisal, and comprises the following specific steps:
1) for a question qjD represents the difficulty of the question, b represents the word stock of the question, w represents the weight parameter of the word stock, l represents the distinguishing degree of the question, and all samples of the word stock form a sample set S ═ S1,s2,s3...,sn]Wherein s isi={hi,sciDenotes the ith sample, hiIs its hit vector, sciRepresents its score;
2) optimizing the parameter d;
3) and optimizing the parameter w.
As a preferred aspect of the present invention, the specific steps of the optimization of the parameter d are as follows:
a) extracting all question sets of the same type Q ═ Q1,q2,q3...qnExtracting an average set avgs ═ avg of the questions1,avg2,avg3...,avgn};
b) According to the question qjAverage score avg ofjThe position in the set avgs can be adjusted in difficulty, and if the average score is higher, the difficulty is lowerOtherwise, the difficulty is high.
As a preferred aspect of the present invention, the specific steps of the optimization of the parameter w are as follows:
a) sorting the sample set S according to the sample score sc, wherein the sorted sample set is S';
b) generating a random normal distribution achievement set sc' by taking the average avg and the standard deviation sd of the achievement sets sc of all samples as parameters;
c) by adjusting the weight parameter w, the sc and sc' are fitted by a gradient descent method.
As a preferred aspect of the present invention, sc is obtained in step 1)iAnd sc are respectively:
sc=w·hT。
as a preferred aspect of the present invention, the topic q in step 2)jThe difficulty adjustment formula is as follows:
the near function maps the values to three numbers of-1, 0 and 1 nearby, the avg function calculates the average value of a set, and the sd function calculates the standard deviation of the set.
As a preferred aspect of the present invention, the loss function in step 3) is formulated as follows:
wherein L is2The function represents the L2 distance of a vector to the origin.
As a preferred aspect of the present invention, the specific operation steps of the gradient descent method in step 3) are as follows:
a) selecting one distribution curve of the weight parameters w;
b) obtaining the score according to the selected distribution curveScore set sc for all samples within a cloth Curve1;
c) Gathering sc with the achievements of all samples1Average avg of1And standard deviation sd1As parameters, a set of random normal distribution results sc 'is generated'1;
d) Selecting another distribution curve of the weight parameter w;
e) obtaining the achievement set sc of all samples in the distribution curve according to the selected distribution curve2;
f) Gathering sc with the achievements of all samples2Average avg of2And standard deviation sd2As parameters, a set of random normal distribution results sc 'is generated'2;
g) Repeating the above six steps to obtain the sample score sc1,sc2,sc3,...scnAnd average number avg1,avg2,avg3...avgnSdn, and a standard deviation sd1, sd2, sd3,. and generating a random normal distribution achievement set sc' n from the resulting data.
The foregoing shows and describes the general principles, essential features, and advantages of the invention. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, and the preferred embodiments of the present invention are described in the above embodiments and the description, and are not intended to limit the present invention. The scope of the invention is defined by the appended claims and equivalents thereof.
Claims (7)
1. A method for dynamically optimizing an expression evaluation question is characterized in that: by applying the statistical learning method, the accuracy of evaluation and appraisal is improved and the difficulty of evaluation and appraisal is dynamically adjusted while the number of samples is increased, and the method specifically comprises the following steps:
1) for a question qjD represents the difficulty of the question, b represents the thesaurus of the question, w represents the weighting parameter of the thesaurus, and l represents the weight of the questionDiscrimination, all samples of which constitute a set of samples S ═ S1,s2,s3...,sn]Wherein s isi={hi,sciDenotes the ith sample, hiIs its hit vector, sciRepresents its score;
2) optimizing the parameter d;
3) and optimizing the parameter w.
2. The method for dynamically optimizing an expressive force evaluation topic according to claim 1, wherein: the specific steps of the optimization of the parameter d are as follows:
a) extracting all question sets of the same type Q ═ Q1,q2,q3...qnExtracting an average set avgs ═ avg of the questions1,avg2,avg3...,avgn};
b) According to the question qjAverage score avg ofjThe difficulty can be adjusted at the position in the set avgs, if the average score is higher, the difficulty is lower, otherwise, the difficulty is higher.
3. The method for dynamically optimizing an expressive force evaluation topic according to claim 1, wherein: the specific steps of the optimization of the parameter w are as follows:
a) sorting the sample set S according to the sample score sc, wherein the sorted sample set is S';
b) generating a random normal distribution achievement set sc' by taking the average avg and the standard deviation sd of the achievement sets sc of all samples as parameters;
c) by adjusting the weight parameter w, the sc and sc' are fitted by a gradient descent method.
5. the method for dynamically optimizing an expressive force evaluation topic according to claim 1, wherein: topic q in step 2)jThe difficulty adjustment formula is as follows:
the near function maps the values to three numbers of-1, 0 and 1 nearby, the avg function calculates the average value of a set, and the sd function calculates the standard deviation of the set.
6. The method for dynamically optimizing an expressive force evaluation topic according to claim 1, wherein: the loss function formula in step 3) is as follows:
J(w)=L2(w·hT-sc′)+w2,
wherein L is2The function represents the L2 distance of a vector to the origin.
7. The method for dynamically optimizing an expressive force evaluation topic according to claim 1, wherein: the specific operation steps of the gradient descent method in the step 3) are as follows:
a) selecting one distribution curve of the weight parameters w;
b) obtaining the achievement set sc of all samples in the distribution curve according to the selected distribution curve1;
c) Gathering sc with the achievements of all samples1Average avg of1And standard deviation sd1As parameters, a set of random normal distribution results sc 'is generated'1;
d) Selecting another distribution curve of the weight parameter w;
e) obtaining the achievement set sc of all samples in the distribution curve according to the selected distribution curve2;
f) Set of achievements with all samplessc2Average avg of2And standard deviation sd2As parameters, a set of random normal distribution results sc 'is generated'2;
g) Repeating the above six steps to obtain the sample score sc1,sc2,sc3,...scnAnd average number avg1,avg2,avg3...avgnAnd standard deviation sd1,sd2,sd3,...sdnAnd generating a random normal distribution score set sc 'from the obtained data'n。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010254255.6A CN112116187A (en) | 2020-04-02 | 2020-04-02 | Method for dynamically optimizing expression evaluation questions |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010254255.6A CN112116187A (en) | 2020-04-02 | 2020-04-02 | Method for dynamically optimizing expression evaluation questions |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112116187A true CN112116187A (en) | 2020-12-22 |
Family
ID=73798938
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010254255.6A Pending CN112116187A (en) | 2020-04-02 | 2020-04-02 | Method for dynamically optimizing expression evaluation questions |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112116187A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114037172A (en) * | 2021-11-18 | 2022-02-11 | 国网江苏省电力有限公司电力科学研究院 | Push optimization method and device for conversation ability evaluation test questions |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105590283A (en) * | 2016-03-03 | 2016-05-18 | 云南电网有限责任公司教育培训评价中心 | Examination data analysis method on the basis of fuzzy synthetic evaluation model |
CN107845047A (en) * | 2017-11-07 | 2018-03-27 | 语联网(武汉)信息技术有限公司 | A kind of dynamic grading system, method and computer-readable recording medium |
CN110442828A (en) * | 2019-06-21 | 2019-11-12 | 杜向然 | A kind of examination question optimization algorithm based on dynamic particles group |
-
2020
- 2020-04-02 CN CN202010254255.6A patent/CN112116187A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105590283A (en) * | 2016-03-03 | 2016-05-18 | 云南电网有限责任公司教育培训评价中心 | Examination data analysis method on the basis of fuzzy synthetic evaluation model |
CN107845047A (en) * | 2017-11-07 | 2018-03-27 | 语联网(武汉)信息技术有限公司 | A kind of dynamic grading system, method and computer-readable recording medium |
CN110442828A (en) * | 2019-06-21 | 2019-11-12 | 杜向然 | A kind of examination question optimization algorithm based on dynamic particles group |
Non-Patent Citations (5)
Title |
---|
方德坚: "主观题自动评分算法模型研究", 《电子世界》 * |
潘刚: "智能组卷系统中试题难度正态分布算法的研究", 《云南民族大学学报(自然科学版)》 * |
赵志弘等: "基于NLP-TF技术的主观题自动评测系统研究", 《软件导刊》 * |
赵明月: "基于词性和关键词的短文本相似度计算方法", 《计算机时代》 * |
陈明晶: "潜在语义分析方法在主观题评判中的应用", 《浙江科技学院学报》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN114037172A (en) * | 2021-11-18 | 2022-02-11 | 国网江苏省电力有限公司电力科学研究院 | Push optimization method and device for conversation ability evaluation test questions |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN106709754A (en) | Power user grouping method based on text mining | |
CN106503805A (en) | A kind of bimodal based on machine learning everybody talk with sentiment analysis system and method | |
CN109871851B (en) | Chinese character writing normalization judging method based on convolutional neural network algorithm | |
CN103605990B (en) | Integrated multi-classifier fusion classification method and integrated multi-classifier fusion classification system based on graph clustering label propagation | |
CN108899049A (en) | A kind of speech-emotion recognition method and system based on convolutional neural networks | |
CN111933127A (en) | Intention recognition method and intention recognition system with self-learning capability | |
CN107492103A (en) | Gray threshold acquisition methods, image partition method based on APSO algorithm | |
CN111128128B (en) | Voice keyword detection method based on complementary model scoring fusion | |
US6112021A (en) | Markov model discriminator using negative examples | |
CN110084266A (en) | A kind of dynamic emotion identification method based on audiovisual features depth integration | |
CN112233655B (en) | Neural network training method for improving recognition performance of voice command words | |
CN110825850A (en) | Natural language theme classification method and device | |
CN112116187A (en) | Method for dynamically optimizing expression evaluation questions | |
CN114882531A (en) | Cross-domain pedestrian re-identification method based on deep learning | |
CN109409231A (en) | Multiple features fusion sign Language Recognition Method based on adaptive hidden Markov | |
CN111104515A (en) | Emotional word text information classification method | |
CN114384999A (en) | User irrelevant myoelectricity gesture recognition system based on self-adaptive learning | |
WO2022146333A1 (en) | A text summarization performance evaluation method sensitive to text categorization and a summarization system using the said method | |
CN103871413A (en) | Men and women speaking voice classification method based on SVM and HMM mixing model | |
CN117350287A (en) | Text emotion analysis method based on public opinion big data | |
CN116050419B (en) | Unsupervised identification method and system oriented to scientific literature knowledge entity | |
CN116561692A (en) | Dynamic update real-time measurement data detection method | |
CN109325241B (en) | Translation robot optimization method based on consistency calculation and computer system thereof | |
CN115862639A (en) | Artificial intelligence voice analysis method based on K-means clustering analysis | |
CN111191510A (en) | Relation network-based remote sensing image small sample target identification method in complex scene |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20201222 |
|
RJ01 | Rejection of invention patent application after publication |