CN-122020066-A - Statistical analysis method for using data of touch pen for learning behaviors of children

CN122020066ACN 122020066 ACN122020066 ACN 122020066ACN-122020066-A

Abstract

The invention relates to the technical field of intelligent analysis of child education data, in particular to a data statistical analysis method for a reading pen for learning behaviors of children. The quantized results are obtained through content field clustering, reading frequency pattern recognition and reading duration stability measurement, the three quantized results are input into a behavior feature fusion processor to generate a comprehensive learning behavior feature map, a learning state evaluation list with page codes and state quantized values is generated through a learning state evaluation network, and a learning resource planning deductor outputs a to-be-reviewed content queue and a personalized learning path of a next-stage reading content sequence. The method realizes accurate characterization of behavior data, multidimensional association fusion and page-level learning state quantification, and matches the learning behavior characteristics of individual children.

Inventors

ZHANG TENGFENG
XU BAIYONG

Assignees

深圳市小彼恩文教科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260409

Claims (10)

1. The statistical analysis method of the using data of the touch pen for learning behaviors of children is characterized by comprising the following steps: Acquiring an original use record containing page codes, reading times and time stamps from a child touch-and-talk pen to form an original use record set; Performing format standardization and content enhancement on the original usage record set to generate a structured click-to-read behavior record list, wherein the click-to-read behavior record list comprises learning content identifiers, multi-level reading frequency distribution and single-reading duration spans; Performing content domain clustering on the learning content identifiers to generate a content theme set, performing pattern recognition on the multi-level reading frequency distribution to obtain a reading frequency pattern, and performing stability measurement on the single reading duration span to obtain a reading duration stability index; The content theme set, the reading frequency mode and the reading time length stability index are input into a behavior feature fusion processor together to perform internal association analysis, and a comprehensive learning behavior feature diagram is generated; Based on the comprehensive learning behavior feature diagram, a learning state evaluation network is deployed to perform depth calculation, and a learning state evaluation list containing page codes and state quantization values is generated; And generating personalized learning path suggestions for individual learning of children through a learning resource planning deducer according to the learning state evaluation list, wherein the personalized learning path suggestions comprise a content queue to be reviewed and a next-stage reading content sequence.
2. The child learning behavior oriented stylus usage data statistical analysis method of claim 1, wherein content domain clustering is performed on the learning content identification, generating a content topic collection, comprising: Extracting key description phrases from the learning content identification; performing concept matching on the key description phrase and a preset child knowledge graph to obtain a basic concept label; carrying out semantic similarity calculation and domain attribution judgment on the basic concept labels, and aggregating to form a plurality of content theme clusters; And selecting basic concept labels with centrality from each content topic cluster as representatives, and summarizing to form the content topic set.
3. The method for statistical analysis of data used by a reading pen for learning behavior of children according to claim 2, wherein the step of performing pattern recognition on the multi-level reading frequency distribution to obtain a reading frequency pattern includes: in the multi-level reading frequency distribution, the reading times are reckoned according to different time scales to form a daily frequency sequence and a Zhou Du frequency sequence; Respectively carrying out periodic fluctuation detection and trend judgment on the daily frequency sequence and the Zhou Du frequency sequence; And integrating the periodic fluctuation detection result and the trend judgment result to generate the reading frequency mode describing the characteristics of the children's reading behaviors in the aspects of density, regularity and trend.
4. A method of statistical analysis of data used by a child learning behavior oriented stylus according to claim 3, wherein performing a stability metric on the single reading duration span results in a reading duration stability indicator, comprising: screening out a plurality of duration records aiming at the same learning content identification from the single reading duration span; calculating standard deviation or variation coefficients of a plurality of duration records aiming at the same learning content identification; calculating a stability score value by combining the average value of the duration records, wherein the stability score value is used for reflecting the fluctuation degree of the children on the reading duration of the same content; And summarizing the stability scoring values corresponding to all the learning content identifiers to form the reading duration stability index.
5. The method of statistical analysis of data used by a child-oriented learning behavior-based stylus according to claim 4, wherein calculating a stability score value in combination with the average of the plurality of duration records comprises: Taking the average value of the duration records as a reference of expected reading duration; Calculating the deviation degree of each duration record from the reference by taking the expected reading duration as the reference; And carrying out normalized weighted average on the deviation degree of all duration records to obtain a comprehensive score representing the overall fluctuation dispersion, namely the stability score value.
6. The method for statistical analysis of data used by a touch and talk pen for learning behavior of children according to claim 5, wherein the step of inputting the content theme set, the reading frequency pattern and the reading time duration stability index together into a behavior feature fusion processor to perform internal association analysis, and the step of generating a comprehensive learning behavior feature map comprises the following steps: The behavior feature fusion processor receives the content theme set, the reading frequency mode and the reading time length stability index as initial feature nodes; in the behavior feature fusion processor, a dynamic association channel is established between the initial feature nodes according to a preset feature interaction rule; through the dynamic association channel, performing multiple-round state iteration and information complementation among the content theme set, the reading frequency mode and the reading time length stability index; after reaching the preset convergence condition, the behavior feature fusion processor outputs the comprehensive learning behavior feature graph formed by the final feature nodes and the weighted connection relation among the nodes.
7. The method for analyzing data statistics of a touch-and-talk pen for learning behavior of children according to claim 6, wherein based on the comprehensive learning behavior feature map, a learning state evaluation network is deployed to perform depth calculation, and a learning state evaluation list including page codes and state quantization values is generated, including: taking the final feature nodes in the comprehensive learning behavior feature graph as input features of the learning state evaluation network; The learning state evaluation network comprises a plurality of cascaded characteristic transformation layers, and the input characteristics are sequentially abstract and fused layer by layer through the plurality of cascaded characteristic transformation layers; after being processed by the plurality of cascaded characteristic transformation layers, outputting a state quantization value corresponding to each original page code; And grouping and sorting all page codes and state quantized values corresponding to the page codes according to the numerical range of the state quantized values to form the learning state evaluation list.
8. The method of statistical analysis of usage data of a stylus for learning behavior for children according to claim 7, wherein generating personalized learning path suggestions for individual learning of children through a learning resource planning deriver according to the learning state evaluation list comprises: Acquiring a preset child learning outline and a knowledge point association structure; Loading the learning state evaluation list and the learning outline and knowledge point association structure of the child into the learning resource planning deducer; In the learning resource planning deducer, a decision model of knowledge mastering level and learning sequence is constructed, wherein the decision model takes a state quantization value in the learning state evaluation list as the input of the knowledge mastering level, and takes the learning outline and the knowledge point association structure of the child as the basis of sequence constraint; And solving the decision model, outputting a learning content arrangement sequence optimized for the current knowledge mastering level on the premise of following the sequence constraint, and forming the content queue to be reviewed and the next stage reading content sequence.
9. The child learning behavior oriented stylus use data statistical analysis method according to claim 8, wherein solving the decision model, outputting a learning content arrangement order optimized for a current knowledge mastering level while following the order constraint, comprises: Defining a pre-knowledge condition which can be accessed for each knowledge point in the child learning outline and knowledge point association structure in the decision model; comparing the pre-knowledge condition of each knowledge point with the state quantization value of the relevant page in the learning state evaluation list, and judging whether the current knowledge mastering level meets the pre-knowledge condition or not; For knowledge points which do not meet the pre-knowledge condition, preferentially adding the related pages or the pre-knowledge pages thereof into the to-be-reviewed content queue; And for the knowledge points meeting the pre-knowledge condition, calculating the suitability score of the learning according to the knowledge difficulty and the state quantization value of the related page in the learning state evaluation list, and determining the position of the knowledge points in the reading content sequence of the next stage according to the suitability score.
10. The method for statistical analysis of usage data of a stylus for learning behavior for children according to claim 9, further comprising, after the step of generating personalized learning path suggestions for individual learning of children through a learning resource planning deriver according to the learning state evaluation list: Extracting core planning information from the personalized learning path suggestion, wherein the core planning information comprises an identification of a content queue to be reviewed and a starting point of a next-stage reading content sequence; Converting the core planning information into a reading task planning file which can be identified and executed by a child touch-and-talk pen; And pushing the reading task plan file to a specific point-reading pen device associated with the target child.

Description

Statistical analysis method for using data of touch pen for learning behaviors of children Technical Field The invention relates to the technical field of intelligent analysis of child education data, in particular to a statistical analysis method of data used by a touch-and-talk pen for learning behaviors of children. Background The existing data processing mode for the child touch and talk pen only collects three types of original use records including page coding, reading times and time stamps, only carries out simple frequency accumulation and total duration accounting on the original data, does not carry out format standardization and content enhancement processing on the original records, and does not construct a structured touch and talk behavior record list comprising learning content identification, multi-level reading frequency distribution and single reading duration span. The existing processing mode does not carry out content field clustering operation aiming at the learning content identification, does not carry out pattern recognition on multi-level reading frequency distribution, does not carry out stability measurement on single reading duration span, and only stays at the shallow data statistics level. The existing data processing scheme of the touch and talk pen cannot form independent content theme sets, reading frequency modes and reading duration stability indexes, does not need a behavior feature fusion processing link for carrying out internal correlation analysis on three types of data, and cannot generate a comprehensive learning behavior feature map. The existing scheme does not deploy a learning state evaluation network to carry out depth calculation, cannot generate a learning state evaluation list carrying page codes and state quantization values, does not form a learning path plan for a child individual through a learning resource planning deductor, and cannot output a content queue to be reviewed and a next-stage reading content sequence. The scheme needs to complete the structural conversion and multidimensional special quantization processing of the original record, needs to realize the association fusion of multi-class behavior data and page-level learning state quantization, and generates a customized learning content queue and sequence planning based on a quantization result. Disclosure of Invention The invention aims to solve the defects in the prior art, and provides a data statistical analysis method for the use of a touch-and-talk pen facing the learning behavior of children. In order to achieve the purpose, the invention adopts the following technical scheme that the method for statistically analyzing the data of the use of the touch-and-talk pen facing the learning behavior of children comprises the following steps: Acquiring an original use record containing page codes, reading times and time stamps from a child touch-and-talk pen to form an original use record set; Performing format standardization and content enhancement on the original usage record set to generate a structured click-to-read behavior record list, wherein the click-to-read behavior record list comprises learning content identifiers, multi-level reading frequency distribution and single-reading duration spans; Performing content domain clustering on the learning content identifiers to generate a content theme set, performing pattern recognition on the multi-level reading frequency distribution to obtain a reading frequency pattern, and performing stability measurement on the single reading duration span to obtain a reading duration stability index; The content theme set, the reading frequency mode and the reading time length stability index are input into a behavior feature fusion processor together to perform internal association analysis, and a comprehensive learning behavior feature diagram is generated; Based on the comprehensive learning behavior feature diagram, a learning state evaluation network is deployed to perform depth calculation, and a learning state evaluation list containing page codes and state quantization values is generated; And generating personalized learning path suggestions for individual learning of children through a learning resource planning deducer according to the learning state evaluation list, wherein the personalized learning path suggestions comprise a content queue to be reviewed and a next-stage reading content sequence. As a further aspect of the present invention, performing content domain clustering on the learning content identifier, to generate a content topic set, including: Extracting key description phrases from the learning content identification; performing concept matching on the key description phrase and a preset child knowledge graph to obtain a basic concept label; carrying out semantic similarity calculation and domain attribution judgment on the basic concept labels, and aggregating to form a plurality of content theme clusters; And selecting b