CN110795997B - Teaching method and device based on long-short-term memory and computer equipment - Google Patents

Teaching method and device based on long-short-term memory and computer equipment Download PDF

Info

Publication number
CN110795997B
CN110795997B CN201910886610.9A CN201910886610A CN110795997B CN 110795997 B CN110795997 B CN 110795997B CN 201910886610 A CN201910886610 A CN 201910886610A CN 110795997 B CN110795997 B CN 110795997B
Authority
CN
China
Prior art keywords
value
answer sheet
teaching
preset
long
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.)
Active
Application number
CN201910886610.9A
Other languages
Chinese (zh)
Other versions
CN110795997A (en
Inventor
张奇
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ping An Technology Shenzhen Co Ltd
Original Assignee
Ping An Technology Shenzhen Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Ping An Technology Shenzhen Co Ltd filed Critical Ping An Technology Shenzhen Co Ltd
Priority to CN201910886610.9A priority Critical patent/CN110795997B/en
Priority to PCT/CN2019/116338 priority patent/WO2021051502A1/en
Publication of CN110795997A publication Critical patent/CN110795997A/en
Application granted granted Critical
Publication of CN110795997B publication Critical patent/CN110795997B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/40—Document-oriented image-based pattern recognition
    • G06V30/41—Analysis of document content
    • G06V30/413—Classification of content, e.g. text, photographs or tables
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
    • G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/049—Temporal neural networks, e.g. delay elements, oscillating neurons or pulsed inputs
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • 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—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
    • G06Q50/10—Services
    • G06Q50/20—Education
    • G06Q50/205—Education administration or guidance
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/20—Image preprocessing
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/56—Extraction of image or video features relating to colour
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/14—Image acquisition
    • G06V30/148—Segmentation of character regions
    • G06V30/153—Segmentation of character regions using recognition of characters or words
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • G06V30/32—Digital ink
    • G06V30/36—Matching; Classification
    • G—PHYSICS
    • G10—MUSICAL INSTRUMENTS; ACOUSTICS
    • G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
    • G10L15/00—Speech recognition
    • G10L15/26—Speech to text systems
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
    • G06V30/10—Character recognition
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
    • Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Health & Medical Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Business, Economics & Management (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Software Systems (AREA)
  • Biomedical Technology (AREA)
  • Educational Technology (AREA)
  • Tourism & Hospitality (AREA)
  • Educational Administration (AREA)
  • Strategic Management (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • Acoustics & Sound (AREA)
  • Human Resources & Organizations (AREA)
  • Marketing (AREA)
  • Primary Health Care (AREA)
  • General Business, Economics & Management (AREA)
  • Human Computer Interaction (AREA)
  • Economics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Image Analysis (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Electrically Operated Instructional Devices (AREA)

Abstract

The application discloses a teaching method, a teaching device, computer equipment and a storage medium based on long-term and short-term memory, wherein the teaching method comprises the following steps: acquiring a designated answer sheet picture, and performing text recognition processing on the designated answer sheet picture to obtain an answer sheet text; receiving scoring results of teacher end on answer volume text; obtaining learning characteristic data of students corresponding to the answer sheet text, and obtaining a predicted result output by a scoring predicted model; calculating the difference degree value of the estimated result and the grading result; if the difference degree value is larger than a preset error threshold value, a deduction knowledge point is generated; acquiring a teaching time period corresponding to each deduction knowledge point; calculating an association index between the teaching time periods; and acquiring a specified association index and a specified time period with the rank larger than a preset ranking threshold value, and sending reminding information for improving teaching quality to the teacher end, wherein the reminding information is attached with the specified time period. Thereby effectively improving the teaching quality.

Description

Teaching method and device based on long-short-term memory and computer equipment
Technical Field
The present disclosure relates to the field of computers, and in particular, to a teaching method, apparatus, computer device, and storage medium based on long-short-term memory.
Background
The online intelligent examination and approval examination paper simplifies a plurality of links such as examination paper storage, distribution, transportation, recovery, core and the like into only one examination paper evaluation flow under the traditional examination paper mode, and the other links are uniformly completed by a computer, so that manpower and material resources are greatly saved, and the whole examination paper time is shortened. However, the intelligent examination and approval paper is too much in the result, only the knowledge of which part of the student is not enough can be known, but the reason why the knowledge of which part is not enough (i.e. the teaching quality is to be improved) can not be known. Therefore, the conventional technology cannot know which parts of the teaching quality need to be improved, and therefore a technical scheme capable of accurately obtaining which parts of the teaching quality need to be improved is needed.
Disclosure of Invention
The main purpose of the application is to provide a teaching method, a teaching device, computer equipment and a storage medium based on long-term and short-term memory, and aims to improve teaching quality.
In order to achieve the above purpose, the present application proposes a teaching method based on long-short-term memory, comprising the following steps:
acquiring a designated answer sheet picture, performing word recognition processing on the designated answer sheet picture to obtain an answer sheet text, wherein the designated answer sheet picture refers to a picture obtained by performing image acquisition on a paper test sheet after answering;
Sending the answer sheet text to a teacher end, and receiving a grading result of the teacher end on the answer sheet text;
obtaining learning characteristic data of students corresponding to the answer sheet text, and inputting the learning characteristic data into a pre-set score estimation model after training, so as to obtain an estimation result output by the score estimation model, wherein the score estimation model is trained based on a long-and-short-term memory model;
calculating a difference degree value of the estimated result and the grading result according to a preset difference degree value calculation method, and judging whether the difference degree value is larger than a preset error threshold value, wherein the error threshold value is larger than or equal to 0;
if the difference degree value is larger than a preset error threshold value, generating a withholding knowledge point according to the withholding position in the answer sheet text;
a preset knowledge point teaching time table is called, and a teaching time period corresponding to each deduction knowledge point is obtained according to the time table;
calculating to obtain association indexes among teaching time periods according to a preset time period association index calculation method, and arranging the association indexes in descending order according to the numerical value to obtain an association index table;
Acquiring a designated association index with the rank larger than a preset ranking threshold value in the association index table, acquiring a designated time period corresponding to the designated association index, and sending reminding information for improving teaching quality to the teacher end, wherein the reminding information is attached with the designated time period.
Further, the paper test paper after answering includes handwritten characters and printed characters, and the step of performing character recognition processing on the specified answer picture to obtain an answer text includes:
collecting the value of an R color channel, the value of a G color channel and the value of a B color channel in an RGB color model of a pixel point in the appointed answer sheet picture, and setting the RGB color of the pixel point in the appointed answer sheet picture to be (0, 0), (255 ) or (Q, Q, Q) according to a preset color setting method, wherein Q is a preset value which is more than 0 and less than 255, so that a temporary picture formed by three colors is obtained;
calculating the occupied areas of the three colors in the temporary picture, and respectively carrying out word segmentation processing on the occupied areas of the two colors with smaller areas, so as to obtain segmented first font words and segmented second font words;
Extracting the characteristics of the first font characters and the characteristics of the second font characters, inputting the characteristics into a preset character classification model based on a support vector machine for classification, and classifying the first font characters as handwritten characters or classifying the second font characters as handwritten characters;
and combining all the divided handwritten characters into handwritten character texts, and marking the handwritten character texts as answer sheet texts.
Further, the step of collecting the value of the R color channel, the value of the G color channel, and the value of the B color channel in the RGB color model of the pixel point in the specified answer sheet picture, and setting the RGB color of the pixel point in the specified answer sheet picture to be (0, 0), (255 ) or (Q, Q) according to a preset color setting method, where Q is a preset value greater than 0 and less than 255, includes:
collecting the values of an R color channel, a G color channel and a B color channel in an RGB color model of a pixel point in the appointed answer sheet picture, and according to the formula: f1 Obtaining a color influence value F1 by =min { ROUND [ (a1r+a2g+a3b)/L, 0], a }, wherein MIN is a minimum function, ROUND is a rounding function, a1, a2, a3 are positive numbers greater than 0 and less than L, L is an integer greater than 0, a is a preset first threshold parameter with a value within a range (0, 255), R, G, B is a value of an R color channel, a value of a G color channel, and a value of a B color channel in an RGB color model of a specified pixel point in the specified picture, respectively;
Judging whether the value of the color influence numerical value F1 is equal to A;
if the value of the color affecting value F1 is equal to a, then according to the formula: f2 Obtaining a color impact value F2, wherein MAX is a maximum function, B is a preset second threshold parameter with a value within a range (0, 255), and B is greater than a;
judging whether the value of the color influence numerical value F2 is equal to B;
if the value of the color impact value F2 is not equal to B, the RGB color of the specified pixel point is set to (255, 255, 255).
Further, the scoring prediction model includes a long-short-period memory network for encoding and a long-short-period memory network for decoding, which are sequentially connected, and the learning feature data is input into a pre-set scoring prediction model after training, so as to obtain a prediction result output by the scoring prediction model, where the scoring prediction model is trained based on the long-short-period memory model, and the method includes the steps of:
inputting the learning characteristic data into the long-term and short-term memory network for coding for processing to obtain a hidden state vector sequence in the long-term and short-term memory network for coding;
inputting the hidden state vector sequence into the long-short-term memory network for decoding to process, so as to obtain predicted knowledge points and corresponding mastery degree values output by the long-short-term memory network for decoding;
And taking the knowledge points with the grasping degree value larger than a preset grasping degree threshold value as the estimated result, and outputting the estimated result.
Further, the step of inputting the learning feature data into the long-term and short-term memory network for encoding to obtain a hidden state vector sequence in the long-term and short-term memory network for encoding includes:
according to the formula: h is a t ＝LSTM enc (x t ，h t-1 ) Obtaining a hidden state vector h in the long-short-term memory network for encoding t Wherein t is the t time period, h t For a hidden state vector corresponding to the t-th time period, h t-1 To a hidden state vector corresponding to the t-1 th time period, X t For learning feature data of the t-th time period, LSTM enc The method refers to performing coding operation by using a long-period memory network for coding;
according to the formula:e ij ＝score(s i ，h j ) Obtaining a final hidden state vector c in the long-period memory network for encoding i ，a ij Is a weight parameter, wherein n time periods are total, s i For the i-th hidden state vector in the long-short-term memory network for encoding, score (s i ，h j ) Refers to the use of a preset score function according to s i And h j The calculated score;
the final hidden state vectors corresponding to a plurality of preset time periods form a hidden state vector sequence c 1 、c 2 ...、c n 。
Further, each teaching time period has m labels, the labels record label values, and the step of calculating the association index between the teaching time periods according to a preset time period association index calculation method includes:
mapping the teaching time period into a high-dimensional vector of a high-dimensional virtual space according to the label value, wherein the dimension of the high-dimensional vector is m;
according to the formula:
calculating to obtain a correlation index DIS between two teaching time periods, wherein C is a high-dimensional vector corresponding to one teaching time period, ci is an ith component vector of the high-dimensional vector C, the high-dimensional vector C shares m component vectors, D is a high-dimensional vector corresponding to the other teaching time period, di is an ith component vector of the high-dimensional vector D, and the high-dimensional vector D shares m component vectors.
Further, the teacher end is provided with a voice input device, and the method sends the teacher end a reminding message with improved teaching quality, wherein after the step of attaching the reminding message to the designated time period, the method comprises the following steps:
acquiring voice data acquired by the teacher end by utilizing the voice input device;
according to a preset voice recognition technology, recognizing the voice data into voice texts;
Judging whether a specified keyword exists in the voice text;
if the appointed keyword exists in the voice text, acquiring the appointed knowledge point corresponding to the appointed keyword according to the corresponding relation between the preset keyword and the knowledge point;
and attaching the voice data to a specified position in the scoring result, wherein the specified position is a position corresponding to the specified knowledge point.
The application provides a teaching device based on long-term memory includes:
the answer sheet text acquisition unit is used for acquiring specified answer sheet pictures, performing text recognition processing on the specified answer sheet pictures to obtain answer sheet texts, wherein the specified answer sheet pictures refer to pictures obtained by performing image acquisition on paper test sheets after answering;
the answer sheet text sending unit is used for sending the answer sheet text to a teacher end and receiving a grading result of the teacher end on the answer sheet text;
the pre-estimation result acquisition unit is used for acquiring learning characteristic data of students corresponding to the answer sheet text, inputting the learning characteristic data into a pre-set score pre-estimation model after training, and obtaining a pre-estimation result output by the score pre-estimation model, wherein the score pre-estimation model is trained based on a long-period memory model;
An error threshold value judging unit, configured to calculate a difference degree value between the estimated result and the scoring result according to a preset difference degree value calculating method, and judge whether the difference degree value is greater than a preset error threshold value, where the error threshold value is greater than or equal to 0;
the withhold knowledge point generating unit is used for generating withhold knowledge points according to withhold positions in the answer sheet text if the difference degree value is larger than a preset error threshold value;
the teaching time period acquisition unit is used for calling a preset knowledge point teaching time table and acquiring a teaching time period corresponding to each deduction knowledge point according to the time table;
the association index calculation unit is used for calculating association indexes among teaching time periods according to a preset time period association index calculation method, and arranging the association indexes in descending order according to the numerical value to obtain an association index table;
the reminding information sending unit is used for obtaining the appointed association index with the rank larger than the preset ranking threshold value in the association index table, obtaining the appointed time period corresponding to the appointed association index, and sending the reminding information with the improved teaching quality to the teacher side, wherein the reminding information is attached with the appointed time period.
The present application provides a computer device comprising a memory storing a computer program and a processor implementing the steps of any of the methods described above when the computer program is executed by the processor.
The present application provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps of the method of any of the above.
According to the teaching method, the teaching device, the computer equipment and the storage medium based on long-short-term memory, the appointed answer sheet picture is obtained, and the appointed answer sheet picture is subjected to word recognition processing to obtain an answer sheet text; sending the answer sheet text to a teacher end, and receiving a grading result of the teacher end on the answer sheet text; obtaining learning characteristic data of students corresponding to the answer sheet text, and obtaining an estimated result output by the scoring estimated model; calculating the difference degree value of the estimated result and the grading result; if the difference degree value is larger than a preset error threshold value, a deduction knowledge point is generated; acquiring a teaching time period corresponding to each deduction knowledge point; calculating to obtain an association index between the teaching time periods; acquiring a designated association index with the rank larger than a preset ranking threshold value in the association index table, acquiring a designated time period corresponding to the designated association index, and sending reminding information for improving teaching quality to the teacher end, wherein the reminding information is attached with the designated time period. Thereby effectively improving the teaching quality.
Drawings
FIG. 1 is a flow chart of a teaching method based on long-term and short-term memory according to an embodiment of the present application;
FIG. 2 is a schematic block diagram of a teaching device based on long-short term memory according to an embodiment of the present application;
fig. 3 is a block diagram schematically illustrating a structure of a computer device according to an embodiment of the present application.
The realization, functional characteristics and advantages of the present application will be further described with reference to the embodiments, referring to the attached drawings.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.
Referring to fig. 1, an embodiment of the present application provides a teaching method based on long-short-term memory, including the following steps:
s1, acquiring a specified answer sheet picture, and performing text recognition processing on the specified answer sheet picture to obtain an answer sheet text, wherein the specified answer sheet picture refers to a picture obtained by performing image acquisition on a paper test sheet after answering;
s2, sending the answer sheet text to a teacher end, and receiving a grading result of the teacher end on the answer sheet text;
S3, learning feature data of students corresponding to the answer sheet text are obtained, the learning feature data are input into a pre-set score estimation model after training is completed, and accordingly an estimation result output by the score estimation model is obtained, wherein the score estimation model is trained based on a long-period and short-period memory model;
s4, calculating a difference degree value of the estimated result and the grading result according to a preset difference degree value calculation method, and judging whether the difference degree value is larger than a preset error threshold value, wherein the error threshold value is larger than or equal to 0;
s5, if the difference degree value is larger than a preset error threshold value, generating a withholding knowledge point according to the withholding position in the answer sheet text;
s6, a preset knowledge point teaching time table is called, and a teaching time period corresponding to each deduction knowledge point is obtained according to the time table;
s7, calculating the association indexes among the teaching time periods according to a preset time period association index calculation method, and arranging the association indexes in descending order according to the numerical value to obtain an association index table;
s8, acquiring a specified association index with the rank larger than a preset ranking threshold value in the association index table, acquiring a specified time period corresponding to the specified association index, and sending reminding information for improving teaching quality to the teacher side, wherein the reminding information is attached with the specified time period.
And step S1, obtaining a designated answer sheet picture, and performing text recognition processing on the designated answer sheet picture to obtain an answer sheet text, wherein the designated answer sheet picture refers to a picture obtained by performing image acquisition on a paper test sheet after answering. The character recognition processing refers to recognizing characters in the picture as character texts. The text recognition process may be performed by any method, such as OCR (Optical Character Recognition ) recognition. Further, the text recognition processing of the specified answer sheet picture includes: and recognizing the handwritten text from the appointed answer sheet picture, and taking the handwritten text as the answer sheet text. Thereby reducing network overhead and improving information transmission efficiency.
And step S2, the answer sheet text is sent to a teacher end, and a grading result of the teacher end on the answer sheet text is received. The scoring result may be any form of scoring result, for example, including one or more of sub-scores of each question (or knowledge point), total scores of the whole answer text, comments corresponding to the sub-scores, and total comments corresponding to the total scores.
And step S3, learning feature data of students corresponding to the answer sheet text are obtained, and the learning feature data are input into a pre-set score estimation model after training, so that an estimation result output by the score estimation model is obtained, wherein the score estimation model is trained based on a long-period and short-period memory model. The long-term memory model is a model using a long-term memory network, wherein the long-term memory network is a time recurrent neural network and is suitable for processing and predicting important events with relatively long intervals and delays in a time sequence, and compared with a common recurrent neural network, the long-term memory model is added with a processor for judging whether information is useful or not, only information conforming to algorithm authentication can be left, and inconsistent information is forgotten through a forgetting gate, so that the long-term dependency problem is solved. The estimated result may be any form of estimated result, for example, an overall score, or a knowledge point of grasp, etc. Further, the scoring prediction model includes a long-period memory network for encoding and a long-period memory network for decoding, which are sequentially connected, and the processing procedure of the scoring prediction model is as follows: inputting the learning characteristic data into the long-term and short-term memory network for coding for processing to obtain a hidden state vector sequence in the long-term and short-term memory network for coding; inputting the hidden state vector sequence into the long-short-term memory network for decoding to process, so as to obtain predicted knowledge points and corresponding mastery degree values output by the long-short-term memory network for decoding; and taking the knowledge points with the grasping degree value larger than a preset grasping degree threshold value as the estimated result, and outputting the estimated result.
And (4) according to the preset difference degree value calculating method, calculating the difference degree value between the estimated result and the grading result, and judging whether the difference degree value is larger than a preset error threshold value, wherein the error threshold value is larger than or equal to 0. The difference degree value calculating method can be any method (related to the estimated result and the scoring result), for example, a difference value between the estimated result and the scoring result is calculated by adopting a difference method (at this time, the estimated score of the estimated result is taken, and the corresponding scoring result is the score total score); or the number of the same knowledge points (the same knowledge points refer to the knowledge points with the same scores in the scoring as the estimated knowledge points which are mastered) is used as the difference degree value. The present application prefers the number of identical knowledge points as the degree of differentiation value.
And (5) if the difference degree value is greater than the preset error threshold, generating a withholding knowledge point according to the withholding position in the answer sheet text. If the difference degree value is larger than a preset error threshold value, the teaching quality is not expected, and therefore the teaching quality of which parts need to be analyzed needs to be improved. Therefore, according to the withheld positions in the answer sheet text, withholding knowledge points are generated for subsequent analysis.
As described in the above step S6, a preset teaching time schedule of knowledge points is called, and a teaching time period corresponding to each deduction knowledge point is obtained according to the time schedule. Accordingly, the obtained teaching time periods are all time periods in which the suspicious teaching quality is to be improved. However, since knowledge point misclassification is difficult to avoid, further analysis is required for which of these time periods is a major problem for teaching quality.
And as described in the step S7, according to a preset time period association index calculation method, calculating association indexes between teaching time periods, and arranging the association indexes in descending order according to the numerical values to obtain an association index table. The preset time period association index calculating method includes the following steps: mapping the teaching time period into a high-dimensional vector of a high-dimensional virtual space according to the label value, wherein the dimension of the high-dimensional vector is m; according to the formula:
calculating to obtain a correlation index DIS between two teaching time periods, wherein C is a high-dimensional vector corresponding to one teaching time period, ci is an ith sub-vector of the high-dimensional vector C, the high-dimensional vector C shares m sub-vectors, D is a high-dimensional vector corresponding to the other teaching time period, ci is an ith sub-vector of the high-dimensional vector D, and the high-dimensional vector D shares m sub-vectors. Wherein, the label refers to factors that have influence on teaching quality, for example, are: whether the knowledge point teaches after a physical lesson; the degree of association of the knowledge point in the whole knowledge point network; the ease of learning of the knowledge point; the importance of the knowledge point, etc. Resulting in an index of association between the teaching time periods.
And step S8, acquiring a specified association index with the rank larger than a preset ranking threshold value in the association index table, acquiring a specified time period corresponding to the specified association index, and sending reminding information for improving teaching quality to the teacher side, wherein the reminding information is attached with the specified time period. The specified time periods corresponding to the specified association indexes indicate that the specified time periods are time periods with great influence on teaching quality, and if teaching quality is improved for the time periods, teaching quality can be improved more effectively.
Further, after the step of sending the reminding information for improving the teaching quality to the teacher end, the method further includes: acquiring voice data acquired by the teacher end by utilizing the voice input device; according to a preset voice recognition technology, recognizing the voice data into voice texts; judging whether a specified keyword exists in the voice text; if the appointed keyword exists in the voice text, acquiring the appointed knowledge point corresponding to the appointed keyword according to the corresponding relation between the preset keyword and the knowledge point; and attaching the voice data to a specified position in the scoring result, wherein the specified position is a position corresponding to the specified knowledge point.
In one embodiment, the paper test paper after answering includes handwritten characters and printed characters, and the step S1 of performing text recognition processing on the specified answer sheet picture to obtain an answer sheet text includes:
s101, acquiring a value of an R color channel, a value of a G color channel and a value of a B color channel in an RGB color model of a pixel point in the appointed answer sheet picture, and setting the RGB color of the pixel point in the appointed answer sheet picture to be (0, 0), (255 ) or (Q, Q, Q) according to a preset color setting method, wherein Q is a preset value which is more than 0 and less than 255, so that a temporary picture formed by three colors is obtained;
s102, calculating the occupied areas of the three colors in the temporary picture, and respectively carrying out character segmentation processing on the occupied areas of the two colors with smaller areas, so as to obtain segmented characters of a first font and segmented characters of a second font;
s103, extracting the characteristics of the first font characters and the characteristics of the second font characters, and inputting the characteristics into a preset character classification model based on a support vector machine for classification, so that the first font characters are classified as handwritten characters, or the second font characters are classified as handwritten characters;
S104, combining all the divided handwritten characters into handwritten character texts, and marking the handwritten character texts as answer sheet texts.
As described above, the handwritten text and the printed text obtained by the recognition using the color setting method are realized. The method and the device make the distinction between the handwritten text and the printed text more obvious, specifically, the RGB colors of the pixel points in the appointed answer sheet picture are set to be (0, 0), (255 ) or (Q, Q, Q), wherein Q is a preset value which is larger than 0 and smaller than 255, so that a temporary picture formed by three colors is obtained, the occupied areas of the three colors are calculated, text segmentation processing (the color area with the largest area is used as the background) is respectively carried out on the occupied areas of the two colors with the smaller area, and therefore the divided first font text and the divided second font text (which is temporarily unknown) are obtained. The support vector machine is a generalized linear classifier for binary classification of data according to a supervised learning mode, and is suitable for comparing characters to be identified with pre-stored characters so as to output the most similar characters. And extracting the characteristics of the characters of the first font and the characteristics of the characters of the second font, and inputting the characteristics of the characters of the first font and the characteristics of the characters of the second font into a preset character classification model based on a support vector machine for classification, so as to obtain which character is the handwritten character. And finally, combining all the divided handwritten characters into handwritten character texts, and marking the handwritten character texts as answer sheet texts. When the teacher end reviews, only the answer sheet content of the student is needed, so that the application only takes the answer sheet content of the student as an answer sheet text, and network expenditure is reduced. In addition, since the RGB colors of the pixel points are set to be (0, 0), (255 ) or (Q, Q), the recognition of the background color is more accurate (due to the influence of light rays when taking pictures, the RGB values of the background color are not pure white, and the recognition method of the conventional scheme can cause inaccuracy of the recognition of the background area, thereby affecting the extraction of the handwritten characters). The characteristics of the first font character and the characteristics of the second font character are, for example, special points in pixel points corresponding to the characters: such as extreme points or isolated points, etc.
In one embodiment, the step S101 of collecting the R color channel value, the G color channel value, and the B color channel value in the RGB color model of the pixel point in the specified answer sheet picture, and setting the RGB color of the pixel point in the specified answer sheet picture to be (0, 0), (255 ) or (Q, Q) according to a preset color setting method, where Q is a preset value greater than 0 and less than 255 includes:
s1011, collecting the values of the R color channel, the G color channel and the B color channel in the RGB color model of the pixel point in the appointed answer sheet picture, and according to the formula: f1 Obtaining a color influence value F1 by =min { ROUND [ (a1r+a2g+a3b)/L, 0], a }, wherein MIN is a minimum function, ROUND is a rounding function, a1, a2, a3 are positive numbers greater than 0 and less than L, L is an integer greater than 0, a is a preset first threshold parameter with a value within a range (0, 255), R, G, B is a value of an R color channel, a value of a G color channel, and a value of a B color channel in an RGB color model of a specified pixel point in the specified picture, respectively;
s1012, judging whether the value of the color influence value F1 is equal to A;
S1013, if the value of the color affecting value F1 is equal to a, according to the formula: f2 Obtaining a color impact value F2, MAX { ROUND [ (a1r+a2g+a3b)/L, 0, B }, wherein MAX is a maximum function, B is a preset second threshold parameter with a value within a range (0, 255), and B is greater than a;
s1014, judging whether the value of the color influence value F2 is equal to B;
s1015, if the value of the color impact value F2 is not equal to B, setting the RGB color of the designated pixel point to be (255 ).
As described above, the acquisition of the R color channel value, the G color channel value, and the B color channel value in the RGB color model of the pixel point in the specified answer sheet picture is realized, and the RGB color of the pixel point in the specified answer sheet picture is set to (0, 0), (255, 255, 255), or (Q, Q) according to the preset color setting method. Specifically, two formulas are employed: f1 =min { ROUND [ (a1r+a2g+a3b)/L, 0], a }, f2=max { ROUND [ (a1r+a2g+a3b)/L, 0], B }, to set the specified pixel point to (0, 0), (255, 255, 255) or (Q, Q). Further, if the value of the color influence value F1 is not equal to a, the RGB color of the specified pixel point is set to (0, 0). Further, if the value of the color influence value F2 is equal to B, the RGB color of the specified pixel point is set to (Q, Q). Therefore, the three-valued processing is realized, so that the background, the printed text and the handwritten text are completely distinguished, and the text recognition is more accurate. The ROUND function is a rounding function, and ROUND (M, s) refers to a rounding operation for a real number M by a decimal number s, where s is an integer greater than or equal to 0, for example, ROUND (8.3,0) =8.
In one embodiment, the scoring prediction model includes a long-period memory network for encoding and a long-period memory network for decoding, which are sequentially connected, and the learning feature data is input into a pre-set scoring prediction model after training, so as to obtain a prediction result output by the scoring prediction model, where the scoring prediction model is trained based on the long-period memory model, and the method includes the following step S3:
s301, inputting the learning characteristic data into the long-period and short-period memory network for coding for processing to obtain a hidden state vector sequence in the long-period and short-period memory network for coding;
s302, inputting the hidden state vector sequence into the long-period and short-period memory network for decoding for processing to obtain predicted knowledge points and corresponding mastery degree values output by the long-period and short-period memory network for decoding;
s303, taking knowledge points with the grasping degree value larger than a preset grasping degree threshold value as a pre-estimated result, and outputting the pre-estimated result.
As described above, the estimation result output by the scoring estimation model is obtained. The encoding finger in the long-period memory network for encoding converts the input information into a vector sequence with a specified length, and the decoding finger in the long-period memory network for decoding converts the input vector sequence into a predicted vector sequence. The long-period and short-period memory network for decoding can be operated by adopting any method, for example, adopting the formula: e ij ＝score(s i ，h j )，/>Wherein c i Long-short-period memory net for codingIn-complex final hidden state vector c i ，a ij Is a weight parameter in which n time periods (because knowledge points are changed with time, for example, a knowledge point is forgotten without using and reviewing for a long period of time, and therefore, n time periods are set by using the time characteristics of a long-short-period memory network), si is the i-th hidden state vector in the long-short-period memory network for decoding, score(s) i ，h j ) Refers to the score, W, calculated from si and hj using a predetermined score function C The weight, p, yt, and x are the output probability, the output, and the input (directly related to learning feature data) of the decoding long-short-term memory network corresponding to the t-th time period. And taking the knowledge points with the mastery degree value larger than a preset mastery degree threshold value as the estimated result, and outputting the estimated result, so that the knowledge points with the high mastery degree value are taken as the estimated result.
In one embodiment, the step S301 of inputting the learning feature data into the long-short-term memory network for encoding to obtain the hidden state vector sequence in the long-short-term memory network for encoding includes:
S3011, according to the formula: h is a t ＝LSTM enc (x t ，h t-1 ) Obtaining a hidden state vector h in the long-short-term memory network for encoding t Wherein t is the t time period, h t For a hidden state vector corresponding to the t-th time period, h t-1 To a hidden state vector corresponding to the t-1 th time period, X t For learning feature data of the t-th time period, LSTM enc The method refers to performing coding operation by using a long-period memory network for coding;
s3012, according to the formula:e ij ＝score(s i ，h j ) Obtaining a final hidden state vector c in the long-period memory network for encoding i ，a ij Is a weight parameter, wherein n time periods are total, s i Long and short term memory network for said encodingI-th hidden state vector in (a), score(s) i ，h j ) Refers to the use of a preset score function according to s i And h j The calculated score;
s3013, constructing a hidden state vector sequence c from the final hidden state vectors corresponding to a plurality of preset time periods 1 、c 2 ...、c n 。
As described above, the learning feature data is inputted into the long-short-term memory network for encoding and processed, and the hidden state vector sequence in the long-short-term memory network for encoding is obtained. The application adopts the formula: h is a t ＝LSTM enc (x t ，h t-1 ) Obtaining a hidden state vector h in the long-short-term memory network for encoding t And then according to the formula:e ij ＝score(s i ，h j ) Obtaining a final hidden state vector c in the long-period memory network for encoding i Namely, attention mechanisms are introduced to automatically capture information important to the ending, so that the final hidden state vector sequence is used as a decoding basis of the long-period and short-period memory network for decoding. Because of adopting the attention mechanism, the weight distribution is more accurate, and the prediction accuracy is improved. Accordingly, the final hidden state vectors corresponding to a plurality of preset time periods are formed into a hidden state vector sequence c 1 、c 2 ...、c n Thereby being used as the decoding basis of the long-period memory network for decoding.
In one embodiment, each teaching time period has m labels, the labels record label values, and the step S7 of calculating the association index between the teaching time periods according to a preset time period association index calculation method includes:
s701, mapping the teaching time period into a high-dimensional vector of a high-dimensional virtual space according to the label value, wherein the dimension of the high-dimensional vector is m;
s702, according to the formula:
calculating to obtain a correlation index DIS between two teaching time periods, wherein C is a high-dimensional vector corresponding to one teaching time period, ci is an ith component vector of the high-dimensional vector C, the high-dimensional vector C shares m component vectors, D is a high-dimensional vector corresponding to the other teaching time period, di is an ith component vector of the high-dimensional vector D, and the high-dimensional vector D shares m component vectors.
As described above, the association index between the teaching time periods is calculated according to the preset time period association index calculation method. Wherein, the label refers to factors that have influence on teaching quality, for example, are: whether the knowledge point teaches after a physical lesson; the degree of association of the knowledge point in the whole knowledge point network; the ease of learning of the knowledge point; the importance of the knowledge point, etc. The teaching time periods are mapped into high-dimensional vectors of a high-dimensional virtual space according to the label values, the dimension of the high-dimensional vectors is m, and factors affecting the time periods are accurately mapped into the high-dimensional vectors in a numerical mode (namely, the label values are used as the numerical values of the components of the high-dimensional vectors), so that the calculation of the association degree between the time periods is possible. And then according to the formula:
and calculating a correlation index DIS between the two teaching time periods, so as to know the degree of correlation between the influence factors of the two teaching time periods, and taking the degree of correlation as a basis for whether the teaching quality needs to be improved.
In one embodiment, the teacher end is provided with a voice input device, and the sending, to the teacher end, the reminding information for improving the teaching quality, where after step S8 of the reminding information attached with the specified time period, includes:
S81, acquiring voice data acquired by the teacher end by utilizing the voice input device;
s82, recognizing the voice data into voice texts according to a preset voice recognition technology;
s83, judging whether a specified keyword exists in the voice text;
s84, if a specified keyword exists in the voice text, acquiring a specified knowledge point corresponding to the specified keyword according to a corresponding relation between the preset keyword and the knowledge point;
s85, attaching the voice data to a designated position in the scoring result, wherein the designated position is a position corresponding to the designated knowledge point.
As described above, it is achieved that the voice data is attached to a specified position in the scoring result, wherein the specified position is a position corresponding to the specified knowledge point. For example a microphone array. The voice recognition technology is used for recognizing voice into text, so that data processing is more convenient. The keywords may be set as knowledge points themselves, or words related to knowledge points themselves. Accordingly, the voice data is attached to the specified position in the scoring result. Because the speech comments are more concise and easier for the students to understand, the speech comments are easier for the students to recognize the mistakes of the offenders, thereby re-mastering the withhold knowledge points. In addition, due to the adoption of the keyword judgment mode, the teacher end can realize targeted voice input without searching topics corresponding to knowledge points one by one, and the method is more efficient and quicker.
According to the teaching method based on long-short-term memory, a designated answer sheet picture is obtained, and text recognition processing is carried out on the designated answer sheet picture to obtain an answer sheet text; sending the answer sheet text to a teacher end, and receiving a grading result of the teacher end on the answer sheet text; obtaining learning characteristic data of students corresponding to the answer sheet text, and obtaining an estimated result output by the scoring estimated model; calculating the difference degree value of the estimated result and the grading result; if the difference degree value is larger than a preset error threshold value, a deduction knowledge point is generated; acquiring a teaching time period corresponding to each deduction knowledge point; calculating to obtain an association index between the teaching time periods; acquiring a designated association index with the rank larger than a preset ranking threshold value in the association index table, acquiring a designated time period corresponding to the designated association index, and sending reminding information for improving teaching quality to the teacher end, wherein the reminding information is attached with the designated time period. Thereby effectively improving the teaching quality.
Referring to fig. 2, an embodiment of the present application provides a teaching device based on long-short term memory, including:
the answer sheet text obtaining unit 10 is configured to obtain an appointed answer sheet picture, and perform text recognition processing on the appointed answer sheet picture to obtain an answer sheet text, where the appointed answer sheet picture refers to a picture obtained by performing image acquisition on a paper test sheet after answering;
The answer sheet text sending unit 20 is configured to send the answer sheet text to a teacher end, and receive a scoring result of the teacher end on the answer sheet text;
the estimated result obtaining unit 30 is configured to obtain learning feature data of a student corresponding to the answer sheet text, and input the learning feature data into a pre-set score estimated model after training, so as to obtain an estimated result output by the score estimated model, where the score estimated model is trained based on a long-short-term memory model;
an error threshold value judging unit 40, configured to calculate a difference degree value between the estimated result and the scoring result according to a preset difference degree value calculating method, and judge whether the difference degree value is greater than a preset error threshold value, where the error threshold value is greater than or equal to 0;
the withhold knowledge point generating unit 50 is configured to generate withhold knowledge points according to withhold positions in the answer sheet text if the difference degree value is greater than a preset error threshold;
a teaching time period obtaining unit 60, configured to retrieve a preset knowledge point teaching time schedule, and obtain a teaching time period corresponding to each deduction knowledge point according to the time schedule;
The association index calculation unit 70 is configured to calculate association indexes between the teaching time periods according to a preset time period association index calculation method, and arrange the association indexes in descending order according to the numerical values to obtain an association index table;
the reminding information sending unit 80 is configured to obtain a specified association index in the association index table, where the rank is greater than a preset ranking threshold, obtain a specified time period corresponding to the specified association index, and send reminding information for improving teaching quality to the teacher side, where the reminding information is attached with the specified time period.
The operations performed by the units are respectively corresponding to the steps of the teaching method based on long-short-term memory in the foregoing embodiment one by one, and are not described herein again.
In one embodiment, the paper test paper after answering the questions includes handwritten characters and printed characters, and the answer sheet text obtaining unit 10 includes:
a temporary picture obtaining subunit, configured to collect a value of an R color channel, a value of a G color channel, and a value of a B color channel in an RGB color model of a pixel point in the specified answer sheet picture, and set an RGB color of the pixel point in the specified answer sheet picture to be (0, 0), (255 ) or (Q, Q), where Q is a preset value greater than 0 and less than 255, according to a preset color setting method, so as to obtain a temporary picture composed of three colors;
An area calculating subunit, configured to calculate an area occupied by three colors in the temporary picture, and perform text segmentation processing on occupied areas of two colors with smaller areas, so as to obtain segmented first font characters and segmented second font characters;
the handwriting word classifying subunit is used for extracting the features of the first type of word and the features of the second type of word, inputting the features into a preset word classifying model based on a support vector machine for classification, and classifying the first type of word as handwriting word or the second type of word as handwriting word;
and the answer sheet text acquisition subunit is used for combining all the split handwritten characters into handwritten character texts and recording the handwritten character texts as answer sheet texts.
The operations that the subunits are respectively used for executing are in one-to-one correspondence with the steps of the teaching method based on long-short-term memory in the foregoing embodiment, and are not described herein again.
In one embodiment, the temporary picture acquisition subunit comprises:
the color influence value F1 acquisition module is used for acquiring the value of the R color channel, the value of the G color channel and the value of the B color channel in the RGB color model of the pixel point in the appointed answer sheet picture, and according to the formula: f1 Obtaining a color influence value F1 by =min { ROUND [ (a1r+a2g+a3b)/L, 0], a }, wherein MIN is a minimum function, ROUND is a rounding function, a1, a2, a3 are positive numbers greater than 0 and less than L, L is an integer greater than 0, a is a preset first threshold parameter with a value within a range (0, 255), R, G, B is a value of an R color channel, a value of a G color channel, and a value of a B color channel in an RGB color model of a specified pixel point in the specified picture, respectively;
The color influence value F1 judging module is used for judging whether the value of the color influence value F1 is equal to A;
the color influence value F2 obtaining module is configured to, if the value of the color influence value F1 is equal to a, according to the formula: f2 Obtaining a color impact value F2, MAX { ROUND [ (a1r+a2g+a3b)/L, 0, B }, wherein MAX is a maximum function, B is a preset second threshold parameter with a value within a range (0, 255), and B is greater than a;
the color influence value F2 judging module is used for judging whether the value of the color influence value F2 is equal to B or not;
and a color setting module, configured to set (255, 255, 255) the RGB colors of the specified pixel point if the value of the color impact value F2 is not equal to B.
The operations performed by the modules are respectively corresponding to the steps of the teaching method based on long-short-term memory in the foregoing embodiment one by one, and are not described herein again.
In one embodiment, the scoring prediction model includes a long-term memory network for encoding and a long-term memory network for decoding, which are sequentially connected, and the prediction result obtaining unit 30 includes:
the coding subunit is used for inputting the learning characteristic data into the long-period and short-period memory network for coding to process so as to obtain a hidden state vector sequence in the long-period and short-period memory network for coding;
A predicted knowledge point obtaining subunit, configured to input the hidden state vector sequence into the long-short-term memory network for decoding to perform processing, so as to obtain a predicted knowledge point and a corresponding mastery degree value output by the long-short-term memory network for decoding;
and the estimated result output subunit is used for taking the knowledge points with the grasping degree value larger than the preset grasping degree threshold value as the estimated result and outputting the estimated result.
The operations that the subunits are respectively used for executing are in one-to-one correspondence with the steps of the teaching method based on long-short-term memory in the foregoing embodiment, and are not described herein again.
In one embodiment, the coding subunit comprises:
the hidden state vector acquisition module is used for acquiring the hidden state vector according to the formula: h is a t ＝LSTM enc (x t ，h t-1 ) Obtaining a hidden state vector h in the long-short-term memory network for encoding t Wherein t is the t time period, h t For a hidden state vector corresponding to the t-th time period, h t-1 To a hidden state vector corresponding to the t-1 th time period, X t For learning feature data of the t-th time period, LSTM enc The method refers to performing coding operation by using a long-period memory network for coding;
the final hidden state vector acquisition module is used for obtaining the hidden state vector according to the formula: e ij ＝score(s i ，h j ) Obtaining the final hidden shape in the long-period memory network for encodingState vector c i ，a ij Is a weight parameter, wherein n time periods are total, s i For the i-th hidden state vector in the long-short-term memory network for encoding, score (s i ，h j ) Refers to the use of a preset score function according to s i And h j The calculated score;
a hidden state vector sequence obtaining module, configured to form a hidden state vector sequence c from final hidden state vectors corresponding to a plurality of preset time periods 1 、c 2 ...、c n 。
The operations performed by the modules are respectively corresponding to the steps of the teaching method based on long-short-term memory in the foregoing embodiment one by one, and are not described herein again.
In one embodiment, each teaching time period has m tags, the tags record tag values, and the association index calculating unit 70 includes:
a high-dimensional vector mapping subunit, configured to map the teaching time period into a high-dimensional vector of a high-dimensional virtual space according to the tag value, where a dimension of the high-dimensional vector is m;
a correlation index calculation subunit configured to, according to the formula:
calculating to obtain a correlation index DIS between two teaching time periods, wherein C is a high-dimensional vector corresponding to one teaching time period, ci is an ith component vector of the high-dimensional vector C, the high-dimensional vector C shares m component vectors, D is a high-dimensional vector corresponding to the other teaching time period, di is an ith component vector of the high-dimensional vector D, and the high-dimensional vector D shares m component vectors.
The operations that the subunits are respectively used for executing are in one-to-one correspondence with the steps of the teaching method based on long-short-term memory in the foregoing embodiment, and are not described herein again.
In one embodiment, the teacher end is provided with a voice input device, and the device comprises:
the voice data acquisition unit is used for acquiring voice data acquired by the teacher side by utilizing the voice input device;
the voice text acquisition unit is used for recognizing the voice data into voice texts according to a preset voice recognition technology;
a specified keyword judging unit for judging whether a specified keyword exists in the voice text;
a specified knowledge point obtaining unit, configured to obtain a specified knowledge point corresponding to a specified keyword according to a preset correspondence between the keyword and the knowledge point if the specified keyword exists in the voice text;
and a voice data adding unit for adding the voice data to a specified position in the scoring result, wherein the specified position is a position corresponding to the specified knowledge point.
The operations performed by the units are respectively corresponding to the steps of the teaching method based on long-short-term memory in the foregoing embodiment one by one, and are not described herein again.
According to the teaching device based on long-short-term memory, a designated answer sheet picture is obtained, and text recognition processing is carried out on the designated answer sheet picture to obtain an answer sheet text; sending the answer sheet text to a teacher end, and receiving a grading result of the teacher end on the answer sheet text; obtaining learning characteristic data of students corresponding to the answer sheet text, and obtaining an estimated result output by the scoring estimated model; calculating the difference degree value of the estimated result and the grading result; if the difference degree value is larger than a preset error threshold value, a deduction knowledge point is generated; acquiring a teaching time period corresponding to each deduction knowledge point; calculating to obtain an association index between the teaching time periods; acquiring a designated association index with the rank larger than a preset ranking threshold value in the association index table, acquiring a designated time period corresponding to the designated association index, and sending reminding information for improving teaching quality to the teacher end, wherein the reminding information is attached with the designated time period. Thereby effectively improving the teaching quality.
Referring to fig. 3, in an embodiment of the present invention, there is further provided a computer device, which may be a server, and the internal structure of which may be as shown in the drawing. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the computer is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer equipment is used for storing data used by the teaching method based on long-term and short-term memory. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program, when executed by a processor, implements a teaching method based on long and short term memory.
The processor executes the teaching method based on long-short-term memory, wherein the steps included in the method correspond to the steps of executing the teaching method based on long-short-term memory in the foregoing embodiment one by one, and are not described herein.
It will be appreciated by persons skilled in the art that the structures shown in the drawings are only block diagrams of some of the structures that may be associated with the aspects of the present application and are not intended to limit the scope of the computer apparatus to which the aspects of the present application may be applied.
The computer equipment acquires a specified answer sheet picture, and performs text recognition processing on the specified answer sheet picture to obtain an answer sheet text; sending the answer sheet text to a teacher end, and receiving a grading result of the teacher end on the answer sheet text; obtaining learning characteristic data of students corresponding to the answer sheet text, and obtaining an estimated result output by the scoring estimated model; calculating the difference degree value of the estimated result and the grading result; if the difference degree value is larger than a preset error threshold value, a deduction knowledge point is generated; acquiring a teaching time period corresponding to each deduction knowledge point; calculating to obtain an association index between the teaching time periods; acquiring a designated association index with the rank larger than a preset ranking threshold value in the association index table, acquiring a designated time period corresponding to the designated association index, and sending reminding information for improving teaching quality to the teacher end, wherein the reminding information is attached with the designated time period. Thereby effectively improving the teaching quality.
An embodiment of the present application further provides a computer readable storage medium, on which a computer program is stored, where the computer program when executed by a processor implements a long-short term memory-based teaching method, and the steps included in the method are respectively corresponding to the steps of executing the long-short term memory-based teaching method in the foregoing embodiment one by one, which is not described herein again.
The method comprises the steps of obtaining a specified answer sheet picture, and performing text recognition processing on the specified answer sheet picture to obtain an answer sheet text; sending the answer sheet text to a teacher end, and receiving a grading result of the teacher end on the answer sheet text; obtaining learning characteristic data of students corresponding to the answer sheet text, and obtaining an estimated result output by the scoring estimated model; calculating the difference degree value of the estimated result and the grading result; if the difference degree value is larger than a preset error threshold value, a deduction knowledge point is generated; acquiring a teaching time period corresponding to each deduction knowledge point; calculating to obtain an association index between the teaching time periods; acquiring a designated association index with the rank larger than a preset ranking threshold value in the association index table, acquiring a designated time period corresponding to the designated association index, and sending reminding information for improving teaching quality to the teacher end, wherein the reminding information is attached with the designated time period. Thereby effectively improving the teaching quality.
Those skilled in the art will appreciate that implementing all or part of the above described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed, may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium provided herein and used in embodiments may include non-volatile and/or volatile memory. The nonvolatile memory can include Read Only Memory (ROM), programmable ROM (PROM), electrically Programmable ROM (EPROM), electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual speed data rate SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous Link DRAM (SLDRAM), memory bus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), among others.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, apparatus, article, or method. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, apparatus, article or method that comprises the element.
The foregoing description is only of the preferred embodiments of the present application, and is not intended to limit the scope of the claims, and all equivalent structures or equivalent processes using the descriptions and drawings of the present application, or direct or indirect application in other related technical fields are included in the scope of the claims of the present application.

Claims (10)

1. A teaching method based on long-term and short-term memory is characterized by comprising the following steps:
acquiring a designated answer sheet picture, performing word recognition processing on the designated answer sheet picture to obtain an answer sheet text, wherein the designated answer sheet picture refers to a picture obtained by performing image acquisition on a paper test sheet after answering;
Sending the answer sheet text to a teacher end, and receiving a grading result of the teacher end on the answer sheet text;
obtaining learning characteristic data of students corresponding to the answer sheet text, and inputting the learning characteristic data into a pre-set score estimation model after training, so as to obtain an estimation result output by the score estimation model, wherein the score estimation model is trained based on a long-and-short-term memory model;
calculating a difference degree value of the estimated result and the grading result according to a preset difference degree value calculation method, and judging whether the difference degree value is larger than a preset error threshold value, wherein the error threshold value is larger than or equal to 0;
if the difference degree value is larger than a preset error threshold value, generating a withholding knowledge point according to the withholding position in the answer sheet text;
a preset knowledge point teaching time table is called, and a teaching time period corresponding to each deduction knowledge point is obtained according to the time table;
calculating to obtain association indexes among teaching time periods according to a preset time period association index calculation method, and arranging the association indexes in descending order according to the numerical value to obtain an association index table;
Acquiring a designated association index with the rank larger than a preset ranking threshold value in the association index table, acquiring a designated time period corresponding to the designated association index, and sending reminding information for improving teaching quality to the teacher end, wherein the reminding information is attached with the designated time period.
2. The teaching method based on long-short term memory according to claim 1, wherein the paper test paper after answering includes handwritten characters and printed text, and the step of performing text recognition processing on the specified answer picture to obtain an answer text comprises the following steps:
collecting the value of an R color channel, the value of a G color channel and the value of a B color channel in an RGB color model of a pixel point in the appointed answer sheet picture, and setting the RGB color of the pixel point in the appointed answer sheet picture to be (0, 0), (255 ) or (Q, Q, Q) according to a preset color setting method, wherein Q is a preset value which is more than 0 and less than 255, so that a temporary picture formed by three colors is obtained;
calculating the occupied areas of the three colors in the temporary picture, and respectively carrying out word segmentation processing on the occupied areas of the two colors with smaller areas, so as to obtain segmented first font words and segmented second font words;
Extracting the characteristics of the first font characters and the characteristics of the second font characters, inputting the characteristics into a preset character classification model based on a support vector machine for classification, and classifying the first font characters as handwritten characters or classifying the second font characters as handwritten characters;
and combining all the divided handwritten characters into handwritten character texts, and marking the handwritten character texts as answer sheet texts.
3. The long-short term memory-based teaching method according to claim 2, wherein the step of collecting the R color channel value, the G color channel value, and the B color channel value in the RGB color model of the pixel point in the specified answer sheet picture, and setting the RGB color of the pixel point in the specified answer sheet picture to (0, 0), (255 ) or (Q, Q), wherein Q is a preset value greater than 0 and less than 255, according to a preset color setting method, comprises:
collecting the values of an R color channel, a G color channel and a B color channel in an RGB color model of a pixel point in the appointed answer sheet picture, and according to the formula: f1 Obtaining a color influence value F1 by =min { ROUND [ (a1r+a2g+a3b)/L, 0, a }, wherein MIN is a minimum function, ROUND is a rounding function, a1, a2, a3 are positive numbers greater than 0 and less than L, L is an integer greater than 0, a is a preset first threshold parameter with a value within a range (0, 255), and R, G, B is a value of an R color channel, a value of a G color channel, and a value of a B color channel in an RGB color model of a specified pixel point in the specified answer sheet picture, respectively;
Judging whether the value of the color influence numerical value F1 is equal to A;
if the value of the color affecting value F1 is equal to a, then according to the formula: f2 Obtaining a color impact value F2, MAX { ROUND [ (a1r+a2g+a3b)/L, 0, B }, wherein MAX is a maximum function, B is a preset second threshold parameter with a value within a range (0, 255), and B is greater than a;
judging whether the value of the color influence numerical value F2 is equal to B;
if the value of the color impact value F2 is not equal to B, the RGB color of the specified pixel point is set to (255 ).
4. The long-short term memory-based teaching method according to claim 1, wherein the scoring prediction model comprises a long-short term memory network for encoding and a long-short term memory network for decoding, which are sequentially connected, and the learning feature data is input into a pre-set scoring prediction model after training, so as to obtain a prediction result output by the scoring prediction model, wherein the scoring prediction model is trained based on the long-short term memory model, and the method comprises the following steps:
inputting the learning characteristic data into the long-term and short-term memory network for coding for processing to obtain a hidden state vector sequence in the long-term and short-term memory network for coding;
Inputting the hidden state vector sequence into the long-short-term memory network for decoding to process, so as to obtain predicted knowledge points and corresponding mastery degree values output by the long-short-term memory network for decoding;
and taking the knowledge points with the grasping degree value larger than a preset grasping degree threshold value as the estimated result, and outputting the estimated result.
5. The long-short term memory-based teaching method according to claim 4, wherein the step of inputting the learning feature data into the long-short term memory network for encoding to be processed, to obtain a hidden state vector sequence in the long-short term memory network for encoding, comprises:
according to the formula: h is a t ＝LSTM enc (x t ,h t-1 ) Obtaining a hidden state vector h in the long-short-term memory network for encoding t Wherein t is the t time period, h t For a hidden state vector corresponding to the t-th time period, h t-1 To a hidden state vector corresponding to the t-1 th time period, X t For learning feature data of the t-th time period, LSTM enc The method refers to performing coding operation by using a long-period memory network for coding;
according to the formula:e ij ＝score(s i ,h j ) Obtaining a final hidden state vector c in the long-period memory network for encoding i ，a ij Is a weight parameter, wherein n time periods are total, s i For the i-th hidden state vector in the long-short-term memory network for encoding, score (s i ,h j ) Refers to the use of a preset score function according to s i And h j The calculated score;
the final hidden state vectors corresponding to a plurality of preset time periods form a hidden state vector sequence c 1 、c 2 …、c n 。
6. The long and short term memory-based teaching method according to claim 1, wherein each teaching time period has m labels, the labels are recorded with label values, and the step of calculating the association index between the teaching time periods according to a preset time period association index calculation method comprises the steps of:
mapping the teaching time period into a high-dimensional vector of a high-dimensional virtual space according to the label value, wherein the dimension of the high-dimensional vector is m;
according to the formula:
calculating an association index DIS between the two teaching time periods, wherein C is a high dimension corresponding to the one teaching time periodThe vector Ci is the ith component vector of the high-dimensional vector C, the high-dimensional vector C is provided with m component vectors in total, D is the high-dimensional vector corresponding to another teaching time period, di is the ith component vector of the high-dimensional vector D, and the high-dimensional vector D is provided with m component vectors in total.
7. The teaching method based on long-short term memory according to claim 1, wherein the teacher side is provided with a voice input device, and the step of sending the teacher side a reminder for improving teaching quality, wherein the reminder is accompanied by the specified time period comprises:
acquiring voice data acquired by the teacher end by utilizing the voice input device;
according to a preset voice recognition technology, recognizing the voice data into voice texts;
judging whether a specified keyword exists in the voice text;
if the appointed keyword exists in the voice text, acquiring the appointed knowledge point corresponding to the appointed keyword according to the corresponding relation between the preset keyword and the knowledge point;
and attaching the voice data to a specified position in the scoring result, wherein the specified position is a position corresponding to the specified knowledge point.
8. Teaching device based on long-short term memory, characterized by comprising:
the answer sheet text acquisition unit is used for acquiring specified answer sheet pictures, performing text recognition processing on the specified answer sheet pictures to obtain answer sheet texts, wherein the specified answer sheet pictures refer to pictures obtained by performing image acquisition on paper test sheets after answering;
The answer sheet text sending unit is used for sending the answer sheet text to a teacher end and receiving a grading result of the teacher end on the answer sheet text;
the pre-estimation result acquisition unit is used for acquiring learning characteristic data of students corresponding to the answer sheet text, inputting the learning characteristic data into a pre-set score pre-estimation model after training, and obtaining a pre-estimation result output by the score pre-estimation model, wherein the score pre-estimation model is trained based on a long-period memory model;
an error threshold value judging unit, configured to calculate a difference degree value between the estimated result and the scoring result according to a preset difference degree value calculating method, and judge whether the difference degree value is greater than a preset error threshold value, where the error threshold value is greater than or equal to 0;
the withhold knowledge point generating unit is used for generating withhold knowledge points according to withhold positions in the answer sheet text if the difference degree value is larger than a preset error threshold value;
the teaching time period acquisition unit is used for calling a preset knowledge point teaching time table and acquiring a teaching time period corresponding to each deduction knowledge point according to the time table;
The association index calculation unit is used for calculating association indexes among teaching time periods according to a preset time period association index calculation method, and arranging the association indexes in descending order according to the numerical value to obtain an association index table;
the reminding information sending unit is used for obtaining the appointed association index with the rank larger than the preset ranking threshold value in the association index table, obtaining the appointed time period corresponding to the appointed association index, and sending the reminding information with the improved teaching quality to the teacher side, wherein the reminding information is attached with the appointed time period.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the steps of the method of any of claims 1 to 7 when the computer program is executed.
10. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the method of any of claims 1 to 7.
CN201910886610.9A 2019-09-19 2019-09-19 Teaching method and device based on long-short-term memory and computer equipment Active CN110795997B (en)

Priority Applications (2)

Application Number Priority Date Filing Date Title
CN201910886610.9A CN110795997B (en) 2019-09-19 2019-09-19 Teaching method and device based on long-short-term memory and computer equipment
PCT/CN2019/116338 WO2021051502A1 (en) 2019-09-19 2019-11-07 Long short-term memory-based teaching method and apparatus, and computer device

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910886610.9A CN110795997B (en) 2019-09-19 2019-09-19 Teaching method and device based on long-short-term memory and computer equipment

Publications (2)

Publication Number Publication Date
CN110795997A CN110795997A (en) 2020-02-14
CN110795997B true CN110795997B (en) 2023-07-28

Family

ID=69427350

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910886610.9A Active CN110795997B (en) 2019-09-19 2019-09-19 Teaching method and device based on long-short-term memory and computer equipment

Country Status (2)

Country Link
CN (1) CN110795997B (en)
WO (1) WO2021051502A1 (en)

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111091733B (en) * 2020-03-19 2020-06-30 浙江正元智慧科技股份有限公司 Auxiliary detection system for real-time teaching achievements of teachers
CN114283434B (en) * 2021-11-17 2024-05-28 淮阴工学院 Answer sheet identification method based on machine vision
CN114419636A (en) * 2022-01-10 2022-04-29 北京百度网讯科技有限公司 Text recognition method, device, equipment and storage medium

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109389091A (en) * 2018-10-22 2019-02-26 重庆邮电大学 The character identification system and method combined based on neural network and attention mechanism
WO2019075820A1 (en) * 2017-10-20 2019-04-25 深圳市鹰硕技术有限公司 Test paper reviewing system

Family Cites Families (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2011119484A1 (en) * 2010-03-23 2011-09-29 Iogenetics, Llc Bioinformatic processes for determination of peptide binding
CN103164994B (en) * 2013-03-15 2015-01-07 南京信息工程大学 Operation exercise correction and feedback method
KR20180071029A (en) * 2016-12-19 2018-06-27 삼성전자주식회사 Method and apparatus for speech recognition
CN106981227A (en) * 2017-05-24 2017-07-25 亳州学院 A kind of Mathematical Teaching System
CN109117738A (en) * 2018-07-19 2019-01-01 江苏黄金屋教育发展股份有限公司 Method to go over files based on artificial intelligence
CN109284355B (en) * 2018-09-26 2020-09-22 杭州大拿科技股份有限公司 Method and device for correcting oral arithmetic questions in test paper
CN109902683A (en) * 2019-03-07 2019-06-18 大连大学 Paper core point and Data Input method and system
CN110110585B (en) * 2019-03-15 2023-05-30 西安电子科技大学 Intelligent paper reading implementation method and system based on deep learning and computer program

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2019075820A1 (en) * 2017-10-20 2019-04-25 深圳市鹰硕技术有限公司 Test paper reviewing system
CN109389091A (en) * 2018-10-22 2019-02-26 重庆邮电大学 The character identification system and method combined based on neural network and attention mechanism

Also Published As

Publication number Publication date
WO2021051502A1 (en) 2021-03-25
CN110795997A (en) 2020-02-14

Similar Documents

Publication Publication Date Title
WO2020207431A1 (en) Document classification method, apparatus and device, and storage medium
CN111191078B (en) Video information processing method and device based on video information processing model
US11151406B2 (en) Method, apparatus, device and readable storage medium for image-based data processing
CN110795997B (en) Teaching method and device based on long-short-term memory and computer equipment
CN113656570B (en) Visual question-answering method and device based on deep learning model, medium and equipment
CN110569356B (en) Interviewing method and device based on intelligent interviewing interaction system and computer equipment
CN113094578B (en) Deep learning-based content recommendation method, device, equipment and storage medium
WO2021218028A1 (en) Artificial intelligence-based interview content refining method, apparatus and device, and medium
CN113344206A (en) Knowledge distillation method, device and equipment integrating channel and relation feature learning
CN110728182B (en) Interview method and device based on AI interview system and computer equipment
CN110929524A (en) Data screening method, device, equipment and computer readable storage medium
CN114298035A (en) Text recognition desensitization method and system thereof
CN112989117B (en) Video classification method and device, electronic equipment and computer storage medium
CN111932418B (en) Student learning condition identification method and system, teaching terminal and storage medium
CN116049367A (en) Visual-language pre-training method and device based on non-supervision knowledge enhancement
CN109408175B (en) Real-time interaction method and system in general high-performance deep learning calculation engine
CN113505786A (en) Test question photographing and judging method and device and electronic equipment
CN113453065A (en) Video segmentation method, system, terminal and medium based on deep learning
US20200294410A1 (en) Methods, systems, apparatuses and devices for facilitating grading of handwritten sheets
CN116244474A (en) Learner learning state acquisition method based on multi-mode emotion feature fusion
CN114282549B (en) Method and device for identifying root cause relationship between information, electronic equipment and storage medium
CN115759262A (en) Visual common sense reasoning method and system based on knowledge perception attention network
CN115358300A (en) Student cognitive recognition method, device and equipment based on voice and text classification
CN114842982A (en) Knowledge expression method, device and system for medical information system
CN114330943A (en) Interview assisting method and device

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