Multivariate evaluation method, system and device based on multi-scenario educational information analysis

Abstract

The present invention discloses a multivariate evaluation method, system, and device based on multi-scenario educational information analysis. The method includes: obtaining classroom interactive audio data, post-class voice comment data, roll call image data, non-teaching scene voice data, and third-party operation data; obtaining classroom analysis data, post-class analysis data, subject information analysis data, and non-teaching scene analysis data through voice preprocessing and image preprocessing; combining the post-class analysis data and subject information analysis data to obtain stage feedback data; performing matching analysis on the classroom interactive audio data, post-class voice comment data, and non-teaching scene voice data, and obtaining behavior analysis data through evaluation analysis; optimizing and correcting the third-party operation behavior data, and then obtaining student multivariate evaluation data through multivariate analysis. The method of the present invention solves the problem of defects in the comprehensiveness and accuracy of the evaluation method, while reducing manual workload through automated evaluation.

Description

Technical Field

[0001] The present invention relates to the field of data analysis technology, and in particular to a multi-evaluation method, system and device based on multi-scenario educational information analysis. Background Art

[0002] In the current school education scenario, students need to be systematically and comprehensively evaluated at important time points in each semester, such as mid-term or final exams.

[0003] Currently, teachers can only evaluate students by reviewing limited and fragmented test scores and homework grading records, combined with their memory of daily classroom impressions. Because this evaluation method is artificial, relying primarily on students' subjective performance and teachers' subjective opinions, it is highly subjective and random, resulting in significant limitations in the content and planning of student evaluations, and failing to provide sound advice for students' comprehensive development. Currently, it is not possible to comprehensively and accurately assess students by combining diverse information such as daily homework, exams, classroom performance, extracurricular activities, and non-classroom performance, as well as related evaluations. Summary of the Invention

[0004] In response to the shortcomings of the prior art, the present invention provides a multi-evaluation method, system and device based on multi-scenario educational information analysis.

[0005] In order to solve the above technical problems, the present invention is solved by the following technical solutions:

[0006] A multi-evaluation method based on multi-scenario educational information analysis includes the following steps:

[0007] Obtain classroom interaction audio data, post-class voice comment data, roll call image data, non-teaching scene voice data, and third-party operation behavior data;

[0008] Perform voice preprocessing on classroom interaction audio data, post-class voice comment data, and non-teaching scene voice data to obtain classroom analysis data, post-class analysis data, and non-teaching scene analysis data;

[0009] Establish an initial behavioral event library, combine classroom analysis data, after-class analysis data, and non-teaching scenario analysis data, match the initial behavioral event library through sentence matching, obtain specific behavioral events, and conduct evaluation and analysis based on specific behavioral events in different scenarios to obtain behavioral analysis data;

[0010] Perform image preprocessing on the roll call image data to obtain subject information analysis data, and combine the post-class analysis data and subject information analysis data to obtain stage feedback data;

[0011] Analyze and process third-party operation behavior data to obtain operation analysis data, and use the operation analysis data to correct and optimize the behavior analysis data and stage feedback data to obtain optimized analysis data;

[0012] Multivariate analysis is performed on the optimized analysis data to obtain students' multivariate evaluation data.

[0013] As an implementable method, performing voice preprocessing on classroom interactive audio data to obtain classroom analysis data includes the following steps:

[0014] Split and value analyze the classroom interaction audio data to obtain the classroom interaction audio data after value analysis;

[0015] Using speech recognition technology, the classroom interaction audio data after value analysis is converted to obtain converted classroom interaction audio data;

[0016] Conduct preliminary analysis and text extraction on the converted classroom interactive audio data to obtain key information data;

[0017] Conduct deep semantic analysis on key information data to obtain semantic positive and negative expression data, behavior description data and corresponding relationship data, and then obtain classroom analysis data through behavior classification label matching.

[0018] As an implementable method, the voice preprocessing of the post-class voice comment data to obtain the post-class evaluation data includes the following steps:

[0019] By using speech recognition technology, the post-class speech comment data is converted into text to obtain post-class comment text data;

[0020] The post-class review text data is subjected to in-depth semantic analysis to obtain positive and negative review data, review behavior data, and review correspondence data, and then the post-class evaluation data is obtained through behavioral classification label matching.

[0021] As an implementable method, performing speech preprocessing on the non-teaching scene speech data to obtain the non-teaching scene parsed data includes the following steps:

[0022] Through data analysis and data matching, the non-teaching scene voice data is extracted to obtain non-teaching voice segment data;

[0023] Deep semantic analysis is performed on non-teaching voice segment data to obtain non-teaching positive and negative data, non-teaching behavior data, and non-teaching correspondence data, and then non-teaching scene analysis data is obtained through behavior classification label matching.

[0024] As an implementable method, obtaining a specific behavior event includes the following steps:

[0025] Based on students' daily behavior events, daily behavior data is obtained, and a basic event database, namely an initial behavior event database, is established through the daily behavior data;

[0026] Through semantic understanding, classroom analysis data, after-class analysis data, and non-teaching scenario analysis data are matched with daily behavior data to obtain specific behavioral events, specifically:

[0027] Based on the classroom analysis data, the after-class analysis data and the non-teaching scenario analysis data, a text library is formed, and text error correction is performed on the text library to obtain a corrected text library;

[0028] Identify the behavioral events and subject names in the corrected text library through text matching and name entity recognition to obtain behavioral events and name entities;

[0029] Through the name entity of the behavior event, the behavior events in the initial behavior event library are extracted and identified, the matching behavior events are obtained and associated with the name entity to obtain the behavior event corresponding to the name entity.

[0030] As an implementable method, performing image preprocessing on the roll call image data to obtain subject information analysis data includes the following steps:

[0031] Performing corner point detection on the point list image data to obtain corner point position information, correcting the point list image data using the corner point position information to obtain corrected image data, and obtaining identification code area data using the relative position of the corner point position information;

[0032] Based on the roll call image data, we obtain row and column information data, and combine the corner point position information and the detection and positioning model to obtain the subject information position data;

[0033] By combining the character recognition model and the subject information location data, the subject information recognition data is obtained, and then character correction and character proofreading are performed on the subject information recognition data. Combined with the identification code area data, the subject information analysis data is obtained.

[0034] As an implementable embodiment, the character recognition model is obtained by:

[0035] Acquire an original data set and preprocess the original data set to obtain a training data set, wherein the training data set is a grayscale image, and the preprocessing includes at least data enhancement and data synthesis;

[0036] Build a character recognition pre-training model and train it using the training dataset to obtain a character recognition model;

[0037] Among them, the character recognition pre-training model is implemented through the Beam Search algorithm. During decoding, the top K characters with the highest probability of the current decoded character are screened and saved. Specifically, the first list is constructed to store the largest K results obtained each time, and the decoding length S is traversed; the first list is traversed to decode a candidate set for each of the largest K results so far; each decoded K result is stored in the second list; the top K are selected as the first list according to the confidence level; and the traversal is continued until the decoding length S is obtained.

[0038] As an implementable method, obtaining the stage feedback data includes the following steps:

[0039] Based on the preset subject understanding threshold, the subject information analysis data is analyzed and processed to obtain understanding data;

[0040] By analyzing and summarizing the post-class analysis data and subject information analysis data, we can obtain evaluation data;

[0041] The evaluation data is adjusted and optimized to obtain stage feedback data.

[0042] As an implementable method, performing multivariate analysis on the optimization analysis data includes the following steps:

[0043] Slicing the behavior analysis data in the optimization analysis data to obtain short word data, removing short words and matching hot words from the short word data to generate a student hot word portrait;

[0044] Summarize the behavior analysis data in the optimization analysis data, analyze the behavior characteristics and student types, and obtain behavior characteristic data and student type data;

[0045] Analyze the behavioral analysis data in the optimization analysis data and combine it with expert advice to obtain targeted guidance and suggestions;

[0046] The stage feedback data in the optimization analysis data is summarized, and the learning behavior characteristics are analyzed to obtain the learning behavior data.

[0047] As an implementable method, it also includes the following steps: displaying students' multi-dimensional evaluation data through an interface display module, wherein the displayed content includes at least students' hot word portraits, behavioral characteristic data, student type data, targeted guidance suggestions and learning behavior data.

[0048] A multivariate evaluation system based on multi-scenario educational information analysis, including a data acquisition module, a voice preprocessing module, an event library establishment module, an image preprocessing module, a correction and optimization module, and a multivariate analysis module;

[0049] The data acquisition module acquires classroom interaction audio data, post-class voice comment data, roll call image data, non-teaching scene voice data, and third-party operation behavior data;

[0050] The speech preprocessing module performs speech preprocessing on the classroom interactive audio data, the post-class speech commentary data, and the non-teaching scene speech data to obtain classroom analysis data, post-class analysis data, and non-teaching scene analysis data;

[0051] The event library establishment module establishes an initial behavior event library, combines classroom analysis data, after-class analysis data, and non-teaching scenario analysis data, matches the initial behavior event library through sentence matching, obtains specific behavior events, and performs evaluation and analysis based on the specific behavior events in different scenarios to obtain behavior analysis data;

[0052] The image preprocessing module performs image preprocessing on the roll call image data to obtain subject information analysis data, and combines the after-class analysis data and the subject information analysis data to obtain stage feedback data;

[0053] The correction and optimization module analyzes and processes the third-party operation behavior data to obtain operation analysis data, and corrects and optimizes the behavior analysis data and stage feedback data based on the operation analysis data to obtain optimized analysis data;

[0054] The multivariate analysis module performs multivariate analysis on the optimized analysis data to obtain student multivariate evaluation data.

[0055] A computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the following method:

[0056] Obtain classroom interaction audio data, post-class voice comment data, roll call image data, non-teaching scene voice data, and third-party operation behavior data;

[0057] Perform voice preprocessing on classroom interaction audio data, post-class voice comment data, and non-teaching scene voice data to obtain classroom analysis data, post-class analysis data, and non-teaching scene analysis data;

[0058] Establish an initial behavioral event library, combine classroom analysis data, after-class analysis data, and non-teaching scenario analysis data, match the initial behavioral event library through sentence matching, obtain specific behavioral events, and conduct evaluation and analysis based on specific behavioral events in different scenarios to obtain behavioral analysis data;

[0059] Perform image preprocessing on the roll call image data to obtain subject information analysis data, and combine the post-class analysis data and subject information analysis data to obtain stage feedback data;

[0060] Analyze and process third-party operation behavior data to obtain operation analysis data, and use the operation analysis data to correct and optimize the behavior analysis data and stage feedback data to obtain optimized analysis data;

[0061] Multivariate analysis is performed on the optimized analysis data to obtain students' multivariate evaluation data.

[0062] A multi-evaluation device based on multi-scenario educational information analysis includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, the following method is implemented:

[0063] Obtain classroom interaction audio data, post-class voice comment data, roll call image data, non-teaching scene voice data, and third-party operation behavior data;

[0064] Perform voice preprocessing on classroom interaction audio data, post-class voice comment data, and non-teaching scene voice data to obtain classroom analysis data, post-class analysis data, and non-teaching scene analysis data;

[0065] Establish an initial behavioral event library, combine classroom analysis data, after-class analysis data, and non-teaching scenario analysis data, match the initial behavioral event library through sentence matching, obtain specific behavioral events, and conduct evaluation and analysis based on specific behavioral events in different scenarios to obtain behavioral analysis data;

[0066] Perform image preprocessing on the roll call image data to obtain subject information analysis data, and combine the post-class analysis data and subject information analysis data to obtain stage feedback data;

[0067] Analyze and process third-party operation behavior data to obtain operation analysis data, and use the operation analysis data to correct and optimize the behavior analysis data and stage feedback data to obtain optimized analysis data;

[0068] Multivariate analysis is performed on the optimized analysis data to obtain students' multivariate evaluation data.

[0069] The present invention has significant technical effects due to the adoption of the above technical solutions:

[0070] The method of the present invention realizes the collection of educational data in multiple scenarios, and comprehensively records and analyzes students' performance inside and outside the classroom. At the same time, the operation analysis data obtained after the analysis of third-party operation behavior data is combined to perform secondary data analysis and correction, and comprehensively and scientifically analyze the students' actual learning ability and personal characteristics. At the same time, combined with artificial intelligence algorithms, targeted suggestions are provided for the subsequent development direction and personal improvement of students, providing help and guidance for students' learning career. BRIEF DESCRIPTION OF THE DRAWINGS

[0071] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0072] Figure 1 It is a schematic flow diagram of the method of the present invention;

[0073] Figure 2 It is an overall schematic diagram of the system of the present invention;

[0074] Figure 3 It is a schematic diagram of the student portrait result data of the present invention. DETAILED DESCRIPTION

[0075] The present invention will be further described in detail below with reference to the examples. The following examples are intended to explain the present invention but the present invention is not limited to the following examples.

[0076] Example 1:

[0077] A multi-evaluation method based on multi-scenario educational information analysis, such as Figure 1 As shown, the following steps are included:

[0078] S100: Acquire classroom interaction audio data, post-class voice comment data, roll call image data, non-teaching scene voice data, and third-party operation behavior data;

[0079] S200, performing voice preprocessing on the classroom interaction audio data, the post-class voice comment data, and the non-teaching scene voice data to obtain classroom analysis data, post-class analysis data, and non-teaching scene analysis data;

[0080] S300: Establish an initial behavior event library, combine classroom analysis data, after-class analysis data, and non-teaching scenario analysis data, match the initial behavior event library through sentence matching to obtain specific behavior events, and perform evaluation and analysis based on the specific behavior events in different scenarios to obtain behavior analysis data;

[0081] S400, performing image preprocessing on the roll call image data to obtain subject information analysis data, and combining the after-class analysis data and the subject information analysis data to obtain stage feedback data;

[0082] S500: Analyze and process the third-party operation behavior data to obtain operation analysis data, and correct and optimize the behavior analysis data and stage feedback data based on the operation analysis data to obtain optimized analysis data;

[0083] S600: Perform multivariate analysis on the optimized analysis data to obtain multivariate evaluation data of students.

[0084] The method of the present invention solves the problem that existing student evaluation methods rely solely on students' classroom performance, test records and teachers' subjective evaluations. Data is collected from multiple scenarios such as classroom and after-class, and then the data is analyzed and processed to evaluate and analyze students in multiple scenarios. At the same time, career analysis guidance and suggestions are provided to students. Through comprehensive analysis of multi-scenario data, the problem of insufficient comprehensiveness and accuracy in student evaluation is avoided.

[0085] To ensure data diversity, we collect classroom interaction audio data, post-class voice comment data, roll call image data, non-teaching scene voice data, and third-party operation behavior data. In other words, during class time and teaching activities at school, teachers can use microphones provided by teachers or cameras installed in the classroom to collect data on interactions between teachers and students, including when teachers call on students to answer questions, students' responses, and teacher comments. By collecting complete data for the entire interactive learning time, we obtain the classroom voice data, which can then be uploaded to the server in real time or with a delay via the app to obtain classroom interaction audio data.

[0086] In addition, after-class scenarios, in some teaching environments where real-time voice collection is not suitable, teachers can use computers or mobile phones equipped with audio data collection capabilities to provide comments on subsequent homework, daily behaviors, and activities related to classroom teaching content through WeChat official accounts. This includes correcting paper or online assignments, as well as providing voice evaluations of students' classroom performance, thus obtaining post-class voice comment data.

[0087] The roll call image data is obtained by designing a roll call on the platform. The process is as follows: combining the teacher-student management system to obtain the learning teacher-student relationship and the list of students corresponding to each teacher. The teacher can obtain targeted content to be counted by editing the content of the items to be counted in the table. After completing the roll call design, the upper right corner of the roll call will automatically generate a corresponding identification code containing the teacher and class subject information. The teacher can use it in the teaching process after printing. In daily teaching work, after correcting homework or filling in the roll call content after class, take a photo with a mobile phone and upload the roll call photo based on the WeChat enterprise account to obtain the roll call image data.

[0088] Non-teaching scene voice data can include relevant data provided by students themselves and their peers during the non-teaching process of the school; or in areas such as classrooms, students use fixed classroom hardware recording equipment to input voice-related evaluation content about themselves or their classmates. These contents include peer interaction data, personal psychological change data, and daily behavior performance data. In addition, relevant data in the home environment will also be included. For example, after students return home from school, parents can upload student-related evaluation data to the platform through WeChat voice or taking photos, and the platform will summarize these data.

[0089] Third-party operation behavior data is obtained through products of third-party authoritative institutions. Products of third-party authoritative institutions can be tablet devices or other hardware devices. Students conduct human-computer dialogue through hardware devices. During the human-computer dialogue, a variety of concrete "life scenarios" are set rather than abstract "questions". Through concrete "life scenarios", more emphasis is placed on examining students' life experience, thinking quality, problem-solving ability, rapid learning ability, sociality and emotional expression, etc. Third-party authoritative institutions' products obtain third-party operation behavior data by collecting students' operation data, etc.

[0090] In step S200, the classroom interaction audio data, the post-class voice comment data, and the non-teaching scene voice data are subjected to voice preprocessing to obtain classroom analysis data, post-class analysis data, and non-teaching scene analysis data, including the following steps:

[0091] S210: Performing voice preprocessing on the classroom interaction audio data, including the following steps:

[0092] Step 1: Split the classroom interactive audio data and retain the parts with analytical value. In this embodiment, the meaningless noise parts with low volume and incomprehensible external interference that cannot be accurately judged are eliminated to obtain the classroom interactive audio data after value analysis;

[0093] Step 2: Convert the classroom interactive audio data after value analysis into text to obtain converted classroom interactive audio data. The text conversion can be performed using speech recognition technology, such as speech recognition products provided by Tencent, iFlytek, Huawei, and Alibaba Cloud.

[0094] Step 3: Based on the preliminary analysis of natural semantics and text interception, extract the key text information of the evaluative sentences and the description of student-related content to obtain the key information data;

[0095] Step 4: Conduct in-depth natural semantic analysis on the key information data to obtain the positive and negative expressions in the key information data, as well as the corresponding relationship between the behaviors described by the semantic descriptions and the students. This can be done by using OpenAI or other natural semantic understanding models to accurately identify which student a particular description is directed at.

[0096] Step 5: Match the behavior classification labels on the semantic positive and negative expression data, behavior description data, and corresponding relationship data to obtain classroom analysis data.

[0097] S220: Performing voice preprocessing on the post-class voice comment data, including the following steps:

[0098] S221: Using speech recognition technology, convert the post-class speech comment data into text to obtain post-class comment text data. In this embodiment, the speech recognition technology can be provided by speech recognition products such as Tencent, iFlytek, Huawei, and Alibaba Cloud.

[0099] S222: Performing deep natural semantic analysis on the post-class review text data to analyze the student name analysis, positive and negative expressions of the evaluation, and the corresponding relationship between the behavior and the student in the semantic description, thereby obtaining positive and negative review data, review behavior data, and review correspondence data. The obtained data is accurate enough to determine whether a certain sentence is a description of a specific student;

[0100] S223: Match the positive and negative comment data, comment behavior data, and comment correspondence data with behavior classification labels to obtain post-class evaluation data.

[0101] S230: performing speech preprocessing on the non-teaching scenario speech data, including the following steps:

[0102] S231: extracting and matching the non-teaching scene voice data, extracting the student's name and the corresponding evaluation content, and obtaining the non-teaching voice segment data;

[0103] S232: Perform deep natural semantic analysis on non-teaching speech segments to obtain positive and negative semantic expressions, as well as the corresponding relationships between behaviors and students. This will then yield non-teaching positive and negative data, non-teaching behavior data, and non-teaching corresponding relationship data. This is accurate enough to determine the corresponding description relationship between a particular sentence and a particular student. This can be performed using OpenAI or other natural language understanding models.

[0104] S233: Classify and label match the non-teaching positive and negative data, non-teaching behavior data, and non-teaching corresponding relationship data to obtain non-teaching scenario analysis data.

[0105] In step S300, an initial behavior event library is established. The initial behavior event library is matched with classroom analysis data, after-class analysis data, and non-teaching scenario analysis data through sentence matching to obtain specific behavior events. Based on the specific behavior events, evaluation and analysis are performed in different scenarios to obtain behavior analysis data, including the following steps:

[0106] Obtain all behavioral events that will occur to students during their time at school, and establish an initial behavioral event library based on the behavioral events. In other words, this initial behavioral event library is a summary of all behavioral events. Perform semantic understanding on classroom analysis data, after-class analysis data, and non-teaching scenario analysis data, and achieve matching with behavioral events in the initial behavioral event library through sentence matching to obtain specific behavioral events. In other words: based on classroom analysis data, after-class analysis data, and non-teaching scenario analysis data to form a text library, perform text correction on the text library to obtain a corrected text library; identify behavioral events and subject names in the corrected text library through text matching and name entity recognition to obtain behavioral events and name entities; extract behavioral events from the initial behavioral event library and identify them through the name entities of the behavioral events, obtain matching behavioral events, and associate them with the name entities to obtain behavioral events corresponding to the name entities.

[0107] In addition, as the amount of data increases, the initial behavior event library will also be updated. The specific update process is as follows: continuously collect user data and establish a basic event library, which is the initial behavior event library; when a new event is added, semantic matching is performed to determine whether it is in the original initial behavior event. If it does not exist, a new initial behavior event is added. If it exists, the definition boundary of the original initial behavior event is expanded to achieve the improvement and expansion of the initial behavior event library.

[0108] In step S400, image preprocessing is performed on the attendance list image data to obtain subject information analysis data, and the post-class analysis data and subject information analysis data are combined to obtain stage feedback data, which includes the following steps:

[0109] S410: In this embodiment, image tilt correction and identification code detection are first performed on the roll call image data. This is achieved by using a table corner point detection algorithm to obtain corner point position information. Based on the relative positions of the corner point position information, identification code area position information is obtained. Image tilt correction is then performed based on the corner point position information. In this embodiment, identification codes include but are not limited to QR codes and barcodes.

[0110] S420: Using the corner point position information and the prior information of the rows and columns in the roll call table, the roll call table is evenly split into cells, and then the subject information position data is obtained by expanding the cells and using a detection and positioning model;

[0111] S430: Using the character recognition model and combining the subject information location data, the characters are accurately located to obtain a character recognition result;

[0112] S440: Based on the actual scenario, the character recognition results are corrected using the model through the column formulation mode. The correction logic corrects and proofreads the character recognition results based on the first k results of the character recognition results according to the character column mode to obtain subject information analysis data;

[0113] S450: Summarize and analyze the subject information analysis data and post-class analysis data to generate evaluation data for the period, and then adjust and optimize the evaluation data for the period. In this embodiment, based on the subject information analysis data, classroom performance data, homework status data, and in-class exercise data are obtained. A preset subject comprehension threshold is set, and the subject information analysis data is divided into comprehension levels based on the preset subject comprehension threshold to obtain comprehension data.

[0114] S460: Summarize and analyze the post-class analysis data, and adjust and optimize the evaluation data to obtain stage feedback data. This stage feedback data has been adjusted and optimized to accurately represent the specific situation at a certain stage. Here, subject comprehension refers to the student's mastery of each subject, which is commonly referred to as knowledge absorption. This comprehension data can provide feedback on the student's learning status. This comprehension data can quickly and in real time understand the student's learning situation, allowing teachers to customize training plans and related tutoring plans for each student.

[0115] In this embodiment, the character recognition model is obtained by:

[0116] Acquire an original data set and preprocess the original data set to obtain a training data set, wherein the training data set is a grayscale image, and the preprocessing includes at least data enhancement and data synthesis;

[0117] Build a character recognition pre-training model and train it using the training dataset to obtain a character recognition model;

[0118] Among them, the character recognition pre-training model is implemented through the Beam Search algorithm. During decoding, the top K characters with the highest probability of the current decoded character are screened and saved. Specifically, the first list is constructed to store the largest K results obtained each time, and the decoding length S is traversed; the first list is traversed to decode a candidate set for each of the largest K results so far; each decoded K result is stored in the second list; the top K are selected as the first list according to the confidence level; and the traversal is continued until the decoding length S is obtained.

[0119] To avoid the influence of other colors, such as background color and font color, in this implementation, the image must first be grayscaled. The dataset is then expanded, such as through data augmentation or data synthesis. To enhance the robustness of the model, additional data augmentation methods are added, as follows: Font expansion: OpenCV is used to randomly perform dilation operations on the image at different levels, primarily to simulate different font thicknesses; Scaling: Pillow and OpenCV are used to scale the image to varying degrees, as scaling may affect the key features and boundaries of the image; Small-angle rotation: Randomly generate small rotation angles and then rotate the target image, primarily to enhance the model's ability to recognize different angles.

[0120] Since the real data collected manually is limited, data synthesis processing is also adopted. The required characters are randomly selected, and the original images are transformed at different levels, such as size, color, etc., and then posted at random positions in the cells to achieve the purpose of expanding the data set.

[0121] In addition, during inference, the image is expanded to different degrees based on the detected text location, such as 0, 10%, and 15%. After obtaining images of different scales, inference is performed separately, resulting in three inference results of different sizes, providing more options.

[0122] Perform pattern judgment and matching for each column of characters in the table, that is, formulate a pattern for the column based on the occurrence of each character in the column. For example, for column ABCD, combine the multiple candidates obtained in steps 1 and 2 and select the one with the highest confidence as the final output result of the cell.

[0123] In step S500, the third-party operation behavior data is analyzed and processed to obtain operation analysis data, and the behavior analysis data is then corrected and optimized to obtain optimized analysis data, including the following steps:

[0124] S510: Based on the student's operation behavior records on the terminal device, such as a tablet device or other hardware device, obtain third-party operation behavior data, and obtain operation analysis data through data analysis and data processing;

[0125] S520: Combined with the operation analysis data, the behavior analysis data obtained by comprehensive analysis in other scenarios is subjected to secondary correction and optimization to obtain optimized analysis data.

[0126] In step S600, multivariate analysis is performed on the optimized analysis data to obtain student multivariate evaluation data, including the following steps:

[0127] S610: Slice the behavior analysis data in the optimization analysis data to obtain short word data, remove short words and matching hot words from the short word data, and generate a student hot word profile;

[0128] S620: Summarize the behavior analysis data in the optimized analysis data, analyze the behavior characteristics and student types, and obtain behavior characteristic data and student type data;

[0129] S630: Analyze the behavior analysis data in the optimization analysis data and combine it with expert advice to obtain targeted guidance suggestions;

[0130] S640: Summarize the stage feedback data in the optimization analysis data, analyze the learning behavior characteristics, and obtain learning behavior data.

[0131] Finally, the student hot word portraits, behavioral characteristics data, student type data, targeted guidance suggestions and learning behavior data are displayed through the interface display module, such as Figure 3 In other words, the entire process can be understood as follows: after obtaining classroom analysis data, the student performance of each class is summarized and fed back to the teacher in real time; after obtaining post-class evaluation data, the teacher's input voice and corresponding label information are displayed based on the student dimension and event dimension. During the display process, the teacher can perform secondary editing and corrections on the obtained information.

[0132] After obtaining the subject information analysis data, teachers can view the data through the corresponding account through the electronic data. The resulting spreadsheet will automatically calculate routine data statistics, including score calculations and quantity statistics. For example, it can calculate the class average score, or count all students who have received five-pointed stars or honors.

[0133] By obtaining non-classroom analysis data, psychological and behavioral data are displayed on mobile phones and web pages based on student dimensions and event dimensions, and a comprehensive analysis of student status changes is conducted. At the same time, during the data display process, based on the need to protect student data and student privacy, permission control is performed in the data display part. Students can choose to block some data for teachers and parents. At the same time, parents can upload data and view the results of analysis and processing of the relevant data uploaded by themselves. Parent permissions are set so that parents only have viewing permissions for the open part of the data uploaded by some teachers and students.

[0134] After analyzing the third-party operation data to obtain the operation analysis data, the behavior analysis data and stage feedback data will be corrected and optimized to obtain the optimization analysis data. Finally, the optimized analysis data will be used to obtain the student multi-evaluation data. The student multi-evaluation data can be understood as including student hot word portraits, behavior characteristics data, student type data, targeted guidance suggestions and learning behavior data, and of course other related data. These data will eventually be displayed on the interface, which can be referred to as Figure 3 As shown, it is possible to intuitively know the diversified evaluation and targeted guidance of each student, providing a guidance basis for the next stage.

[0135] Example 2:

[0136] A multi-evaluation system based on multi-scenario educational information analysis, such as Figure 2 As shown, it includes a data acquisition module 100, a speech pre-processing module 200, an event library establishment module 300, an image pre-processing module 400, a correction and optimization module 500 and a multivariate analysis module 600;

[0137] The data acquisition module 100 acquires classroom interaction audio data, post-class voice comment data, roll call image data, non-teaching scene voice data, and third-party operation behavior data;

[0138] The speech preprocessing module 200 performs speech preprocessing on the classroom interactive audio data, the post-class speech commentary data, and the non-teaching scene speech data to obtain classroom analysis data, post-class analysis data, and non-teaching scene analysis data;

[0139] The event library establishment module 300 establishes an initial behavior event library, combines classroom analysis data, after-class analysis data, and non-teaching scenario analysis data, matches the initial behavior event library through sentence matching, obtains specific behavior events, and performs evaluation and analysis based on the specific behavior events in different scenarios to obtain behavior analysis data;

[0140] The image preprocessing module 400 performs image preprocessing on the roll call image data to obtain subject information analysis data, and combines the after-class analysis data and the subject information analysis data to obtain stage feedback data;

[0141] The correction and optimization module 500 analyzes and processes the third-party operation behavior data to obtain operation analysis data, and corrects and optimizes the behavior analysis data and stage feedback data based on the operation analysis data to obtain optimized analysis data;

[0142] The multivariate analysis module 600 performs multivariate analysis on the optimized analysis data to obtain student multivariate evaluation data.

[0143] Various changes and modifications can be made without departing from the spirit and scope of the present invention, and all equivalent technical solutions also fall within the scope of the present invention.

[0144] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments, and the same or similar parts between the various embodiments can be referenced to each other.

[0145] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, apparatus, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0146] The present invention is described with reference to the flowcharts and / or block diagrams of the method, terminal device (system), and computer program product according to the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing terminal device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the process in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0147] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0148] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device so that a series of operating steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable terminal device to implement the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0149] It should be noted that:

[0150] References in this specification to "one embodiment" or "an embodiment" mean that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the present invention. Therefore, appearances of the phrases "one embodiment" or "an embodiment" in various places throughout this specification do not necessarily refer to the same embodiment.

[0151] Furthermore, it should be noted that the specific embodiments described in this specification may vary in the shapes and names of their components. Any equivalent or simple variations based on the structure, features, and principles described in the patented concept of this invention are included within the scope of protection of this patent. Persons skilled in the art may make various modifications, additions, or substitutions to the described specific embodiments, and these modifications, as long as they do not deviate from the structure of the invention or exceed the scope defined by the claims, shall fall within the scope of protection of this invention.

Claims

1. A multi-evaluation method based on multi-scenario educational information analysis, characterized in that: The following steps are involved: Obtain classroom interaction audio data, post-class voice comment data, roll call image data, non-teaching scene voice data, and third-party operation behavior data; Perform voice preprocessing on classroom interaction audio data, post-class voice comment data, and non-teaching scene voice data to obtain classroom analysis data, post-class analysis data, and non-teaching scene analysis data; Establish an initial behavioral event library, combine classroom analysis data, after-class analysis data, and non-teaching scenario analysis data, match the initial behavioral event library through sentence matching, obtain specific behavioral events, and conduct evaluation and analysis based on specific behavioral events in different scenarios to obtain behavioral analysis data; Perform image preprocessing on the roll call image data to obtain subject information analysis data, and combine the post-class analysis data and subject information analysis data to obtain stage feedback data; Performing corner point detection on the point list image data to obtain corner point position information, correcting the point list image data using the corner point position information to obtain corrected image data, and obtaining identification code area data using the relative position of the corner point position information; Based on the roll call image data, we obtain row and column information data, and combine the corner point position information and the detection and positioning model to obtain the subject information position data; Combining the character recognition model and the subject information location data, the subject information recognition data is obtained, and then character correction and character proofreading are performed on the subject information recognition data, and combined with the identification code area data, the subject information analysis data is obtained; The character recognition model is obtained in the following way: Acquire an original data set and preprocess the original data set to obtain a training data set, wherein the training data set is a grayscale image, and the preprocessing includes at least data enhancement and data synthesis; Build a character recognition pre-training model and train it using the training dataset to obtain a character recognition model; The character recognition pre-training model is implemented using the Beam Search algorithm. During decoding, the top K characters with the highest probability of the currently decoded character are filtered and saved. Specifically, the first list is constructed to store the largest K results obtained each time, and the decoding length S is traversed; the first list is traversed to decode a candidate set for each of the largest K results so far; each decoded K result is stored in the second list; the top K are selected as the first list according to the confidence level; and the traversal is continued until the decoding length S is reached. Analyze and process third-party operation behavior data to obtain operation analysis data, and use the operation analysis data to correct and optimize the behavior analysis data and stage feedback data to obtain optimized analysis data; Multivariate analysis is performed on the optimized analysis data to obtain students' multivariate evaluation data.

2. The multivariate evaluation method based on multi-scenario educational information analysis according to claim 1 is characterized in that: The method of performing voice preprocessing on the classroom interactive audio data to obtain classroom analysis data includes the following steps: Split and value analyze the classroom interaction audio data to obtain the classroom interaction audio data after value analysis; Using speech recognition technology, the classroom interaction audio data after value analysis is converted to obtain converted classroom interaction audio data; Conduct preliminary analysis and text extraction on the converted classroom interactive audio data to obtain key information data; Conduct deep semantic analysis on key information data to obtain semantic positive and negative expression data, behavior description data and corresponding relationship data, and then obtain classroom analysis data through behavior classification label matching.

3. The multivariate evaluation method based on multi-scenario educational information analysis according to claim 1 is characterized in that: The method of performing voice preprocessing on the post-class voice comment data to obtain post-class analysis data includes the following steps: By using speech recognition technology, the post-class speech comment data is converted into text to obtain post-class comment text data; The post-class review text data is subjected to in-depth semantic analysis to obtain positive and negative review data, review behavior data, and review correspondence data, and then the post-class analysis data is obtained through behavioral classification label matching.

4. The multivariate evaluation method based on multi-scenario educational information analysis according to claim 1 is characterized in that: The method of performing speech preprocessing on the non-teaching scene speech data to obtain non-teaching scene parsed data includes the following steps: Through data analysis and data matching, the non-teaching scene voice data is extracted to obtain non-teaching voice segment data; Deep semantic analysis is performed on non-teaching voice segment data to obtain non-teaching positive and negative data, non-teaching behavior data, and non-teaching correspondence data, and then non-teaching scene analysis data is obtained through behavior classification label matching.

5. The multivariate evaluation method based on multi-scenario educational information analysis according to claim 1 is characterized in that: The step of obtaining a specific behavior event comprises the following steps: Based on students' daily behavior events, daily behavior data is obtained, and a basic event database, namely an initial behavior event database, is established through the daily behavior data; Through semantic understanding, classroom analysis data, after-class analysis data, and non-teaching scenario analysis data are matched with daily behavior data to obtain specific behavioral events, specifically: Based on the classroom analysis data, the after-class analysis data, and the non-teaching scenario analysis data, a text library is formed, and text error correction is performed on the text library to obtain a corrected text library; Identify the behavioral events and subject names in the corrected text library through text matching and name entity recognition to obtain behavioral events and name entities; Through the name entity of the behavior event, the behavior events in the initial behavior event library are extracted and identified, the matching behavior events are obtained and associated with the name entity to obtain the behavior event corresponding to the name entity.

6. The multivariate evaluation method based on multi-scenario educational information analysis according to claim 1 is characterized in that: The step of obtaining the stage feedback data comprises the following steps: Based on the preset subject understanding threshold, the subject information analysis data is analyzed and processed to obtain understanding data; By analyzing and summarizing the post-class analysis data and subject information analysis data, we can obtain evaluation data; The evaluation data is adjusted and optimized to obtain stage feedback data.

7. The multivariate evaluation method based on multi-scenario educational information analysis according to claim 1 is characterized in that: The multivariate analysis of the optimized analysis data comprises the following steps: Slicing the behavior analysis data in the optimization analysis data to obtain short word data, removing short words and matching hot words from the short word data to generate a student hot word portrait; Summarize the behavior analysis data in the optimization analysis data, analyze the behavior characteristics and student types, and obtain behavior characteristic data and student type data; Analyze the behavioral analysis data in the optimization analysis data and combine it with expert advice to obtain targeted guidance and suggestions; The stage feedback data in the optimization analysis data is summarized, and the learning behavior characteristics are analyzed to obtain the learning behavior data.

8. The multivariate evaluation method based on multi-scenario educational information analysis according to claim 1 is characterized in that: The following steps are also included: The student's multi-dimensional evaluation data is displayed through the interface display module, wherein the displayed content at least includes student hot word portraits, behavioral characteristic data, student type data, targeted guidance suggestions and learning behavior data.

9. A multi-evaluation system based on multi-scenario educational information analysis, characterized by: It includes data acquisition module, speech preprocessing module, event library establishment module, image preprocessing module, correction and optimization module and multivariate analysis module; The data acquisition module acquires classroom interaction audio data, post-class voice comment data, roll call image data, non-teaching scene voice data, and third-party operation behavior data; The speech preprocessing module performs speech preprocessing on the classroom interactive audio data, the post-class speech commentary data, and the non-teaching scene speech data to obtain classroom analysis data, post-class analysis data, and non-teaching scene analysis data; The event library establishment module establishes an initial behavior event library, combines classroom analysis data, after-class analysis data, and non-teaching scenario analysis data, matches the initial behavior event library through sentence matching, obtains specific behavior events, and performs evaluation and analysis based on the specific behavior events in different scenarios to obtain behavior analysis data; The image preprocessing module performs image preprocessing on the roll call image data to obtain subject information analysis data, and combines the after-class analysis data and the subject information analysis data to obtain stage feedback data; Performing corner point detection on the point list image data to obtain corner point position information, correcting the point list image data using the corner point position information to obtain corrected image data, and obtaining identification code area data using the relative position of the corner point position information; Based on the roll call image data, we obtain row and column information data, and combine the corner point position information and the detection and positioning model to obtain the subject information position data; Combining the character recognition model and the subject information location data, the subject information recognition data is obtained, and then character correction and character proofreading are performed on the subject information recognition data, and combined with the identification code area data, the subject information analysis data is obtained; The character recognition model is obtained in the following way: Acquire an original data set and preprocess the original data set to obtain a training data set, wherein the training data set is a grayscale image, and the preprocessing includes at least data enhancement and data synthesis; Build a character recognition pre-training model and train it using the training dataset to obtain a character recognition model; The character recognition pre-training model is implemented using the Beam Search algorithm. During decoding, the top K characters with the highest probability of the currently decoded character are filtered and saved. Specifically, the first list is constructed to store the largest K results obtained each time, and the decoding length S is traversed; the first list is traversed to decode a candidate set for each of the largest K results so far; each decoded K result is stored in the second list; the top K are selected as the first list according to the confidence level; and the traversal is continued until the decoding length S is reached. The correction and optimization module analyzes and processes the third-party operation behavior data to obtain operation analysis data, and corrects and optimizes the behavior analysis data and stage feedback data based on the operation analysis data to obtain optimized analysis data; The multivariate analysis module performs multivariate analysis on the optimized analysis data to obtain student multivariate evaluation data.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

11. A multi-evaluation device based on multi-scenario educational information analysis, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 8 is implemented.

Images