A teaching demonstration content extraction method and device based on click interaction

Abstract

This application discloses a method and apparatus for extracting teaching demonstration content based on click-based interaction, belonging to the field of teaching demonstration technology. The method includes: decomposing the knowledge content of a demonstration document into hierarchically related node units based on a pre-constructed knowledge hierarchy tree; matching the click positions of click-based interactive operations during the presentation to obtain a candidate node set; filtering the candidate node set based on temporal validity to determine the effective target nodes; matching differentiated weight update rules based on the branch type of the target nodes and the operational characteristics of the click-based interactive operations, and updating the weights of each node level by level; and extracting the corresponding knowledge content based on the final updated weights of each node at the end of the presentation to generate teaching content. This application, through a node-level filtering mechanism and a node-level weight update mechanism, accurately quantifies the degree of emphasis on different knowledge points within the demonstration document, improving the accuracy of extracting key teaching content.

Description

Technical Field

[0001] This application relates to the field of instructional demonstration technology, and in particular to a method and apparatus for extracting instructional demonstration content based on click-interaction. Background Technology

[0002] In the field of intelligent teaching demonstration processing technology, in order to automatically extract the core knowledge content of slides during the teaching process, it is usually necessary to establish a corresponding association between the teacher's interactive operations such as clicking on slides, controlling animation timing, and page jumps, and the knowledge units in the courseware, so as to determine the importance of the corresponding knowledge content, and finally complete the automatic extraction and organization of key teaching content.

[0003] However, traditional technologies for extracting teaching demonstration content generally employ a page-level interaction statistics model. This model uses a single slide as the smallest statistical unit, summarizing the frequency and total duration of various interactions within a page to directly map the overall importance of the page's content. Lacking a matching mechanism with finer granularity than page-level statistics, traditional technologies cannot accurately map a teacher's individual interaction to specific knowledge points. This results in an inability to precisely quantify the importance of knowledge points within a page, leading to a significant discrepancy between the extracted teaching focus and the actual emphasis in the lesson. Consequently, this approach fails to meet the high-precision automatic extraction requirements of demonstration content in teaching scenarios. Summary of the Invention

[0004] The main purpose of this application is to propose a method and device for extracting teaching demonstration content based on click-interaction. Through a node-level filtering mechanism and a node-level weight update mechanism, the degree of emphasis of different knowledge points in the demonstration document is accurately quantified, thereby improving the accuracy of extracting key teaching content.

[0005] To achieve the above objectives, one aspect of this application proposes a method for extracting teaching demonstration content based on click-interaction, the method comprising: Obtain the knowledge hierarchy tree corresponding to the slide presentation document; wherein, the knowledge hierarchy tree has independent page content branches and page interaction sequence branches for each slide; In response to a user's click interaction on the slide presentation document during a teaching presentation, the system obtains the click location information and operation behavior characteristics corresponding to the click interaction, and obtains the current presentation sequence status of the current slide. Based on the click location information, a set of candidate nodes is obtained by matching from the page content branches and the page interaction time sequence branches; the candidate node set is then filtered for time sequence validity based on the current playback time sequence state to obtain several target nodes corresponding to the click interaction operation; Based on the operational behavior characteristics and the branch type to which each target node belongs, a node weight update rule corresponding to the click interaction operation is determined; based on the node weight update rule, the current weight value of each node in the knowledge hierarchy tree is updated level by level. After the slideshow presentation is completed, based on the updated final weight values ​​of each node, the corresponding knowledge content is extracted from the slideshow presentation and a structured teaching demonstration is generated.

[0006] Furthermore, in some embodiments, the process of constructing the knowledge hierarchy tree includes: Based on the content hierarchy attributes and first spatial coordinate attributes of each content element in the slide presentation document, multi-level content nodes are generated, and page content branches are constructed based on each content node. Based on the functional type attributes and second spatial coordinate attributes of each interactive control element in the slide presentation document, multiple interactive control nodes of different types are generated, and the page interaction sequence branch is constructed based on each interactive control node. Based on the original animation timing sequence number corresponding to each slide in the PowerPoint presentation document, a mapping relationship is constructed between the original animation timing sequence number, the display effective timing interval corresponding to the content node, and the interaction effective timing interval corresponding to the interactive control node; The knowledge hierarchy tree is constructed based on the content branches of each page, the interaction sequence branches of each page, and the mapping relationships of each page.

[0007] Furthermore, in some embodiments, the knowledge hierarchy tree further includes several heterogeneous association edges; the heterogeneous association edges are used to indicate non-hierarchical subordinate relationships between different nodes; The step of matching a set of candidate nodes from the page content branches and the page interaction sequence branches based on the click location information includes: Based on the matching relationship between the click location information and the first spatial coordinate attribute and the second spatial coordinate attribute, several initial candidate nodes are selected. For each initial candidate node, based on the heterogeneous association edges connected to the initial candidate node, the direct adjacent nodes that have the non-hierarchical subordinate association relationship with the initial candidate node are extracted as extended candidate nodes; The initial candidate nodes and the extended candidate nodes are deduplicated and merged to generate the candidate node set.

[0008] Furthermore, in some embodiments, the step of filtering out several initial candidate nodes based on the matching relationship between the click location information and the first spatial coordinate attribute and the second spatial coordinate attribute includes: Based on each of the first spatial coordinate attributes and each of the second spatial coordinate attributes, a first spatial bounding box corresponding to each of the content nodes and a second spatial bounding box corresponding to each of the interactive control nodes are constructed. The click location information is converted into a coordinate point to be matched corresponding to the coordinate system of the slide page, and the first spatial bounding box containing the coordinate point to be matched is marked as a candidate spatial bounding box; wherein, the candidate spatial bounding box corresponds to a minimum circumcircle; For each candidate bounding box, the hit proximity coefficient is calculated based on the Euclidean distance between the coordinate point to be matched and the center of the minimum circumcircle, and the radius of the minimum circumcircle. The candidate bounding boxes whose hit proximity coefficient is less than or equal to a preset proximity threshold are marked as target bounding boxes; the content nodes corresponding to the target bounding boxes are marked as initial candidate nodes. The interactive control node corresponding to the second spatial bounding box containing the coordinates to be matched is marked as the initial candidate node.

[0009] Furthermore, in some embodiments, the step of filtering the candidate node set based on the current playback timing state to obtain several target nodes corresponding to the click interaction operation includes: Extract the current animation sequence number corresponding to the current playback sequence state; For each content node to be verified in the candidate node set, the content node to be verified that contains the current animation sequence number in the display effective timing interval is marked as the target node; wherein, the display effective timing interval is used to indicate the range of consecutive animation sequence numbers in which the content node is in a visible display state; For each interactive control node to be verified in the candidate node set, the interactive control node to be verified that contains the current animation sequence number in the interaction effective timing interval is marked as the target node; wherein, the interaction effective timing interval is used to indicate the range of continuous animation sequence numbers in which the interactive control node is in a state that can respond to interactive operations.

[0010] Furthermore, in some embodiments, determining the node weight update rule corresponding to the click interaction operation based on the operation behavior characteristics and the branch type to which each target node belongs includes: For each target node, a basic weight value and weight update polarity are determined based on the branch type to which the target node belongs and the operation behavior characteristics. Based on the branch type to which the target node belongs, valid heterogeneous association edges are selected from each of the heterogeneous association edges; All nodes pointed to by the effective heterogeneous association edges are marked as effective nodes, and a weight transmission link is constructed based on the weight transmission rules corresponding to the effective heterogeneous association edges, each of the effective nodes, and the target node. When traversing each effective node in each weight transmission link, the effective transmission weight increment of the effective node is calculated based on the weight update polarity, the transmission attenuation coefficient in the weight transmission rule, and the transmission weight increment associated with the effective node in the previous level; wherein, during the first traversal, the calculation is performed using the base weight value of the target node as the transmission weight increment of the previous level. The basic weight values, weight update polarities, and effective weight increments in each weight transmission link are summarized to form the node weight update rule.

[0011] Furthermore, in some embodiments, the operation behavior characteristics include: operation type, click dwell time, and click frequency per unit time; the weight update polarity includes: positive polarity update for indicating an increase in weight value or negative polarity update for indicating a decrease in weight value; The step of determining the basic weight value and weight update polarity based on the branch type to which the target node belongs and the operation behavior characteristics includes: The node behavior intensity coefficient is calculated based on the click dwell time and the click frequency per unit time. When the target node is detected to belong to the page content branch, the node behavior intensity coefficient is multiplied by the preset node baseline weight value corresponding to the target node to obtain the base weight value; if the click frequency per unit time is greater than the preset click frequency threshold, the weight update polarity is determined to be the positive polarity update; if the click frequency per unit time is not greater than the preset click frequency threshold, the weight update polarity is determined to be the negative polarity update. When the target node is detected to belong to the page interaction timeline branch, the node behavior intensity coefficient is multiplied by the preset operation baseline weight value corresponding to the operation type to obtain the basic weight value; if the operation type is content playback, the weight update polarity is determined to be the positive polarity update; if the operation type is content skip, the weight update polarity is determined to be the negative polarity update.

[0012] Furthermore, in some embodiments, the heterogeneous association edge includes: a semantic association edge for characterizing the semantic similarity relationship between nodes and a temporal jump edge for indicating the jump direction relationship between nodes; The step of filtering valid heterogeneous association edges from the heterogeneous association edges according to the branch type to which the target node belongs includes: When the target node belongs to the page content branch, the semantic association edge that has a connection relationship with the target node is taken as the effective heterogeneous association edge; When the target node belongs to the page interaction timeline branch, the timeline jump edge that has a connection relationship with the target node is taken as the effective heterogeneous association edge.

[0013] Furthermore, in some embodiments, constructing a weight propagation link based on the weight propagation rules corresponding to the effective heterogeneous association edges, each of the effective nodes, and the target node includes: If the effective heterogeneous association edge is the semantic association edge, the first conduction attenuation coefficient and the weight conduction direction used to indicate the conduction along the semantic similarity association relationship between nodes are retrieved as the weight conduction rule; If the effective heterogeneous association edge is the timing jump edge, the second conduction attenuation coefficient and the weight conduction direction used to indicate conduction along the jump direction are retrieved as the weight conduction rule; wherein, the first conduction attenuation coefficient is less than the second conduction attenuation coefficient; Starting from the target node, connect each of the effective nodes sequentially along the weight propagation direction according to the associated hierarchy to generate a weight propagation link.

[0014] To achieve the above objectives, another aspect of this application proposes a teaching demonstration content extraction device based on click-interaction, the device comprising: The tree structure acquisition module is used to acquire the knowledge hierarchy tree corresponding to the slide presentation document; wherein, the knowledge hierarchy tree has independent page content branches and page interaction sequence branches for each slide; The click interaction response module is used to respond to the click interaction operation triggered by the user on the slide presentation document during the teaching presentation, obtain the click location information and operation behavior characteristics corresponding to the click interaction operation, and obtain the current presentation sequence status of the current slide. The node filtering module, based on the click location information, matches a set of candidate nodes from the page content branches and the page interaction time sequence branches; and filters the candidate node set for time sequence validity based on the current playback time sequence state to obtain several target nodes corresponding to the click interaction operation. The node weight update module is used to determine the node weight update rule corresponding to the click interaction operation based on the operation behavior characteristics and the branch type to which each target node belongs; and to update the current weight value of each node in the knowledge hierarchy tree level by level based on the node weight update rule. The teaching content extraction module is used to extract the corresponding knowledge content from the slide presentation document and generate structured teaching demonstration content based on the final weight values ​​of each node after the slide presentation document has finished playing.

[0015] The embodiments of this application include at least the following beneficial effects: This application provides a method and apparatus for extracting teaching demonstration content based on click-interaction. The scheme first obtains the knowledge hierarchy tree corresponding to the slide presentation document. Based on the page content branch and page interaction timing branch under each slide of the knowledge hierarchy tree, the knowledge content and interactive controls in the courseware can be decomposed into node units with hierarchical relationships. When responding to the click interaction operation triggered by the user during the presentation, the click position information, operation behavior characteristics, and the current presentation timing state of the current slide are collected simultaneously. First, a set of candidate nodes is obtained by matching from the two branches based on the click position information. Then, the timing validity filtering of the candidate nodes is performed in combination with the current presentation timing state to select several target nodes corresponding to the current click interaction. Subsequently, according to the operation behavior characteristics and the branch type to which each target node belongs, the corresponding node weight update rule is determined, and the current weight value of each node is updated level by level along the knowledge hierarchy tree. After detecting that the slide presentation document has finished playing, the corresponding knowledge content is extracted based on the final weight of each node after the update, and structured teaching demonstration content is generated. Compared to traditional technologies, this application, through the combined processing of location matching and temporal validity filtering, can refine the smallest granularity of interaction statistics from the entire slideshow to individual node units within the page. It effectively eliminates nodes unrelated to the clicked location and node units that are not currently active, preventing secondary knowledge points not emphasized in the actual teaching from being mistakenly included in the statistics. Furthermore, based on the branch types and operational behavior characteristics of the selected target nodes, this application derives differentiated weight update rules and completes the weight quantification of all levels of nodes through a hierarchical update process. This achieves precise differentiation of the actual emphasis on different knowledge points, solving the problem of large deviations in the quantification of knowledge point importance caused by the coarse-grained statistics of traditional technologies. This effectively improves the accuracy of extracting teaching demonstration content and can meet the needs of high-precision automatic extraction of key teaching content in teaching scenarios. Attached Figure Description

[0016] Figure 1 This is a flowchart illustrating a method for extracting teaching demonstration content based on click-interaction, as provided in an embodiment of this application. Figure 2 This is a schematic diagram of a dual-branch knowledge hierarchy tree structure provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the node weight propagation link structure provided in an embodiment of the present invention; Figure 4 A single-page slide diagram of a knowledge hierarchy tree for a corresponding teaching scenario provided in an embodiment of the present invention; Figure 5 This is a schematic diagram of the structure of a teaching demonstration content extraction device based on click-interaction provided in an embodiment of this application; Figure 6 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0018] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0019] As used in this application, the terms "several", "each", etc., "several" include one, two or more, "each" refers to each of the corresponding plurality, and "any" refers to any one of the plurality.

[0020] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0021] Before providing a detailed description of the embodiments of this application, some of the nouns and terms involved in the embodiments of this application will be explained first. The nouns and terms involved in the embodiments of this application are subject to the following interpretations.

[0022] (1) Knowledge hierarchy tree: This application forms a tree data structure after the structuring decomposition of the slide content. The single slide is the upper node, and it is split into independent page content branches and page interaction sequence branches to carry the hierarchical relationship, spatial attributes and temporal attributes of knowledge points and interactive controls.

[0023] (2) Heterogeneous association edge: Non-hierarchical subordinate association link connecting different nodes in the knowledge hierarchy tree, used to represent the semantic similarity between nodes and cross-level or cross-branch association relationships such as jump direction.

[0024] (3) Hit proximity coefficient: A coefficient used to quantify the degree of fit between the click position and the center of the content node. The value range is 0~1. The smaller the coefficient, the closer the click position is to the center of the node, and the higher the confidence of hitting the node.

[0025] Traditional teaching demonstration content extraction solutions typically employ a single-page granular statistical model, using the entire slide as the smallest statistical unit. Macro-level data such as page dwell time and page-turning frequency are used to directly map the overall importance of the content on that page. However, traditional solutions cannot analyze the spatial location of the teacher's clicks, thus failing to pinpoint specific knowledge elements. Lacking a finer-grained matching mechanism than page-level statistics, secondary knowledge points not emphasized by the teacher are easily mistakenly included in high-weighted statistics. The importance of different knowledge points cannot be accurately quantified, resulting in a significant discrepancy between the extracted teaching focus and the actual teaching emphasis. This insufficient content extraction accuracy fails to meet the demands of fine-grained and high-precision automatic extraction of demonstration content in smart teaching scenarios.

[0026] In view of this, this application provides a method and apparatus for extracting teaching demonstration content based on click-based interaction. This method constructs a knowledge hierarchy tree containing page content branches and page interaction timing branches, decomposing the knowledge content and interactive controls within the slides into fine-grained independent node units. After collecting click location information, operation behavior characteristics, and the current presentation timing state, candidate nodes are first screened through spatial matching, and then timing validity filtering is performed based on the presentation timing state to select target nodes in an active state. Subsequently, based on the branch type and operation behavior characteristics of the target nodes, differentiated node weight update rules are matched, and the weight values ​​of each node are updated level by level along the knowledge hierarchy tree. Finally, after the presentation ends, the core knowledge content is extracted based on the final weights of the updated nodes to generate structured teaching demonstration content. Compared to the traditional coarse-grained statistical model based on page granularity, this application refines the smallest granularity of interaction statistics from the entire slideshow to individual node units through the linkage of click location matching and time-series validity filtering. This effectively eliminates nodes unrelated to clicks and nodes that are not in an active state, preventing secondary knowledge points that are not actually emphasized from being mistakenly included in the statistical scope. At the same time, by combining branch type and operation behavior characteristics to match differentiated weight update rules, the weight quantification of all levels of nodes is completed through hierarchical transmission. This can accurately distinguish the actual degree of emphasis of different knowledge points, solving the problem of large deviations in the quantification of the importance of knowledge points caused by the coarse-grained statistics of traditional technologies. This effectively improves the accuracy of extracting teaching demonstration content and can meet the needs of high-precision automatic extraction of key teaching content in teaching scenarios.

[0027] Figure 1 This is an optional flowchart of a method for extracting teaching demonstration content based on click-interaction, provided in an embodiment of this application. Figure 1 The method may include, but is not limited to, steps S100 to S500: Step S100: Obtain the knowledge hierarchy tree corresponding to the slide presentation document; wherein, the knowledge hierarchy tree has independent page content branches and page interaction sequence branches for each slide. Schematic illustration: This application, by parsing the original file of a PowerPoint presentation document, can extract all content elements, interactive control elements, and animation timing data within each slide, and construct a corresponding knowledge hierarchy tree according to preset hierarchical rules. The knowledge hierarchy tree can separate the knowledge content and interactive controls within the same page into two independent branches, preserving both the hierarchical relationships of the content itself and independently maintaining the timing attributes of the interactive controls, thus achieving a structured decomposition of content and interaction.

[0028] Step S200: In response to the user's click interaction on the slide presentation document during the teaching presentation, obtain the click location information and operation behavior characteristics corresponding to the click interaction, and obtain the current presentation sequence state of the current slide. To illustrate, during a slideshow presentation, this application can monitor the teacher's mouse clicks or touch clicks in real time. When a click is detected, the application can simultaneously collect the screen coordinates corresponding to the click, the behavioral characteristics of the click, and the current animation playback status of the slideshow.

[0029] Step S300: Based on the click location information, a set of candidate nodes is obtained by matching from the page content branches and page interaction time sequence branches; the candidate node set is filtered for time sequence validity based on the current playback time sequence state to obtain several target nodes corresponding to the click interaction operation; Indicatively, this step is divided into two stages: initial screening and fine filtering. First, based on the spatial location of the click, all spatially matched nodes from the two branches are matched to form a candidate node set, completing coarse-grained positioning. Then, based on the current playback sequence state, nodes that are not currently in a display or interactive state are eliminated, and the truly valid target nodes corresponding to this click are selected to achieve fine-grained verification.

[0030] Step S400: Based on the characteristics of the operation behavior and the branch type to which each target node belongs, determine the node weight update rule corresponding to the click interaction operation; based on the node weight update rule, update the current weight value of each node in the knowledge hierarchy tree level by level. Indicatively, for each target node selected, this application can match the corresponding node weight update rule by combining its branch type and the characteristics of the current click operation behavior; then, according to the update rule, the current weight value of the associated nodes is updated level by level along the hierarchical and association relationships of the knowledge hierarchy tree, thereby realizing the weight quantification of all nodes in a single click interaction.

[0031] Step S500: After the slide presentation is completed, based on the final weight values ​​of each node after the update, extract the corresponding knowledge content from the slide presentation and generate structured teaching demonstration content.

[0032] To illustrate, when the slideshow ends, this application can calculate the final weight values ​​of all nodes in the knowledge hierarchy tree after updates, sort them by weight to select core knowledge points, or extract knowledge points with weights greater than a preset threshold based on the final weight values ​​of all nodes after updates. Then, based on the hierarchical structure of the knowledge hierarchy tree, the content is integrated to generate structured teaching demonstration content, such as key teaching notes, lecture review materials, or knowledge point explanation documents, thus completing the automatic extraction from interactive behavior to teaching content. Specifically, when extracting knowledge content based on the final weight values ​​of each node after updates, extraction can be performed based on the final weight values ​​of the content nodes after updates.

[0033] Steps S100 to S500 as shown in this embodiment of the application realize fine-grained decomposition of slide content through a dual-branch knowledge hierarchy tree. In actual response to interactive operations, spatial matching and temporal filtering can be combined to complete the precise mapping of interaction and nodes. Then, the importance of knowledge points is quantified through differentiated weight rules. Compared with the traditional page granularity statistics scheme, this embodiment of the application can achieve precise extraction of teaching demonstration content at the node level, truly restore the knowledge focus in the teaching process, and greatly improve the accuracy of extracting teaching focus. It can adapt to the content review needs of various teaching scenarios such as smart classrooms, online teaching, and corporate training.

[0034] For step S100, in some embodiments, the process of constructing the knowledge hierarchy tree includes: Based on the content hierarchy attributes and first spatial coordinate attributes of each content element in the PowerPoint presentation document, multi-level content nodes are generated, and page content branches are constructed based on each content node. Based on the functional type attributes and second spatial coordinate attributes of each interactive control element in the PowerPoint presentation document, multiple interactive control nodes of different types are generated, and page interaction sequence branches are constructed based on each interactive control node. Based on the original animation timing sequence number of each slide in the PowerPoint presentation document, a mapping relationship is constructed between the original animation timing sequence number, the display effective timing interval corresponding to the content node, and the interaction effective timing interval corresponding to the interactive control node. A knowledge hierarchy tree is constructed based on the content branches of each page, the interaction sequence branches of each page, and the mapping relationships.

[0035] Specifically, in constructing the knowledge hierarchy tree, this application first processes the two types of elements within the slide separately. For example, for content elements such as titles, body text, and images, it extracts their title level and chapter-related content level attributes, as well as their coordinate range within the page and other first spatial coordinate attributes. Corresponding content nodes are then generated according to the hierarchy from page, chapter, knowledge point to detailed content, forming page content branches from top to bottom. For interactive control elements such as jump buttons and animation triggers, it extracts their jump and play function type attributes, as well as second spatial coordinate attributes, generating corresponding interactive control nodes and forming independent page interaction timing branches. Subsequently, it extracts the original animation timing sequence number of each slide, i.e., extracts the sequence numbers of multiple steps corresponding to the animation playback. For each content node, it matches a corresponding display activation timing interval, indicating from which animation step the node becomes visible and from which animation step it becomes hidden. Similarly, it matches a corresponding interaction activation timing interval for each interactive control node, indicating from which animation step the control becomes clickable. Finally, after establishing the mapping relationship between the animation timing sequence number and the two types of effective timing intervals, the two branches and the mapping relationship are integrated to form a complete knowledge hierarchy tree.

[0036] In some embodiments, please refer to Figure 2 , Figure 2 This is a schematic diagram of the overall structure of the knowledge hierarchy tree provided in this application embodiment. It can intuitively present the global architecture, node hierarchy relationship and heterogeneous association logic of the knowledge hierarchy tree corresponding to the slide presentation document. Among them, the root node of the slide presentation document is the top-level node of the entire presentation document. It is split into multiple parallel single-page slide nodes through hierarchical subordinate relationships. In this embodiment, the first page slide node and the second page slide node are used as examples for illustration. The remaining page nodes follow the same architecture rules.

[0037] Specifically, such as Figure 2 As shown, each slide node has two independent and non-overlapping branches: a page content branch and a page interaction sequence branch, which respectively carry the node-based data of knowledge content and interactive controls. Solid lines represent hierarchical relationships, dashed lines represent semantic associations, and solid lines with arrows represent temporal jumps, clearly distinguishing three different types of node association logic.

[0038] The page content branches adopt a multi-level tree structure, with first-level, second-level, and third-level content nodes arranged from top to bottom. Nodes are connected through hierarchical relationships, fully reproducing the chapter hierarchy and inclusion logic of the knowledge content within the slides. Furthermore, content nodes at the same level are connected by dashed semantic association edges, used to represent semantic similarity relationships between nodes that are not hierarchically subordinate. These semantic association edges can exist between content nodes on the same page or establish semantic associations between content nodes on different pages, thus enabling the association of the same topic knowledge points throughout the entire document.

[0039] Furthermore, the page interaction sequence branches also adopt a hierarchical arrangement, with first-level interactive control nodes and second-level interactive control nodes set from top to bottom. The nodes are connected by hierarchical subordinate relationships, corresponding to interactive control elements of different levels and functional types within the slide. Interactive control nodes are connected to the corresponding target content nodes by a sequence jump edge with a one-way arrow, used to represent the jump direction association of the interactive control; furthermore, the sequence jump edge can point to content nodes within the same page or to target content nodes on other pages, realizing the targeted mapping between interactive operations and knowledge content.

[0040] Therefore, the knowledge hierarchy tree architecture in this embodiment adopts a structural design of document root node, single-page slide node, dual-branch architecture node and multi-type heterogeneous association edge. It not only realizes the independent separation of each page's content and interaction, but also completes non-hierarchical associations across nodes and pages through heterogeneous association edges. This can provide a unified data structure foundation for subsequent spatial node matching, temporal validity filtering and weight propagation calculation.

[0041] This embodiment uses a two-branch hierarchical tree construction method to achieve a structured decomposition of slide content from the page level to the node level, classifying and managing knowledge content and interactive controls. This ensures the integrity of the content hierarchy while taking into account the timing characteristics of interactive controls. At the same time, it can automatically generate mappings based on the original animation timing and the effective intervals of different types of nodes, eliminating the need for manual annotation and enabling the rapid construction of hierarchical trees for slides of different formats.

[0042] For step S300, in some embodiments, based on the basic spatial matching logic, the embodiments of this application further add heterogeneous association edges to the knowledge hierarchy tree and further refine the matching process of the candidate node set, resulting in: The knowledge hierarchy tree also includes several heterogeneous association edges; heterogeneous association edges are used to indicate non-hierarchical subordinate relationships between different nodes; Based on click location information, a set of candidate nodes is obtained by matching from page content branches and page interaction sequence branches, including: Based on the matching relationship between the click location information and the first spatial coordinate attribute and the second spatial coordinate attribute, several initial candidate nodes are selected. For each initial candidate node, based on the heterogeneous association edges connected to the initial candidate node, the direct adjacent nodes that have non-hierarchical subordinate association relationships with the initial candidate node are extracted as extended candidate nodes. Each initial candidate node and each extended candidate node are deduplicated and merged to generate a candidate node set.

[0043] Understandably, in addition to the hierarchical parent-child node relationships, the knowledge hierarchy tree also adds heterogeneous association edges to carry non-hierarchical associations such as semantic similarity and jump directions between nodes. Specifically, in this embodiment, when performing spatial matching, the absolute click coordinates captured from the screen are first converted into relative coordinates within the slideshow page to obtain the coordinate points to be matched, thereby eliminating the influence of screen resolution and presentation window size; then, the coordinate points to be matched are compared one by one with the first spatial coordinates of all content nodes and the second spatial coordinates of all interactive control nodes. If the spatial range contains the click coordinates, it indicates a successful match, and the successfully matched node is taken as the initial candidate node, thus completing the initial screening of directly matched nodes.

[0044] After obtaining the initial candidate nodes, traverse all heterogeneous associated edges connected to each initial candidate node, extract the direct adjacent nodes that have non-hierarchical association with the initial candidate node, and use them as extended candidate nodes. This can expand the scope of the click's influence from the directly hit node to the associated node. Finally, deduplication is performed on the initial candidate nodes and extended candidate nodes, and they are merged to form the final candidate node set.

[0045] For example, when the instructor clicks on the text area of ​​a knowledge point in a chapter, the converted coordinate point to be matched falls within the spatial range of the secondary content node, and the node is marked as the initial candidate node; the node has a semantic association with another secondary content node in the same chapter through a semantic association edge, and when expanding based on the heterogeneous association edge, the adjacent node is extracted as the extended candidate node; finally, the initial candidate node and the extended candidate node are deduplicated and merged to generate the candidate node set corresponding to the click.

[0046] The matching process in this embodiment adopts a matching logic of direct hit and related expansion. First, the direct target node clicked by the instructor is located by matching spatial coordinates to ensure the accuracy of the matching. Then, the matching is expanded to related nodes by heterogeneous related edges. This takes into account the actual teaching logic of clicking a single knowledge point while associating similar knowledge points. It avoids the omission of related knowledge point weights caused by only matching directly hit nodes, so that the subsequent weight statistics are more in line with the knowledge emphasis logic of actual teaching.

[0047] This embodiment adopts a matching mode of initial candidate nodes plus extended candidate nodes, which not only retains the accurate positioning of direct clicks, but also expands the scope of association through heterogeneous association edges. This can fit the actual scenario of emphasizing the linkage of knowledge points during the teaching process and avoid the loss of weight of related knowledge points. At the same time, the deduplication and merging avoids duplicate node statistics and ensures the accuracy of subsequent weight calculation.

[0048] In some embodiments, to further improve the matching accuracy of initial candidate nodes and filter out invalid interference nodes caused by accidental clicks on the edge, this application performs a refined hierarchical determination of the initial candidate node determination process, resulting in: Based on the matching relationship between the click location information and the first spatial coordinate attribute and the second spatial coordinate attribute, several initial candidate nodes are selected, including: Based on each first spatial coordinate attribute and each second spatial coordinate attribute, construct the first spatial bounding box corresponding to each content node and the second spatial bounding box corresponding to each interactive control node; The click location information is converted into a coordinate point to be matched corresponding to the coordinate system of the slide page, and the first spatial bounding box containing the coordinate point to be matched is marked as a candidate spatial bounding box; wherein, the candidate spatial bounding box corresponds to a minimum circumcircle; For each candidate bounding box, the hit proximity coefficient is calculated based on the Euclidean distance between the coordinate point to be matched and the center of the minimum circumcircle, and the radius of the minimum circumcircle. Candidate bounding boxes with a proximity coefficient less than or equal to a preset proximity threshold are marked as target bounding boxes; content nodes corresponding to target bounding boxes are marked as initial candidate nodes. Mark the interactive control node corresponding to the second spatial bounding box containing the coordinates to be matched as the initial candidate node.

[0049] Understandably, due to the dense arrangement of content elements on the page, accidental touches are prone to occur in the edge areas. This embodiment quantifies the click intent through a proximity coefficient, which can effectively filter out accidental touches at the edges and improve matching accuracy. Interactive controls are usually sparsely arranged and the click target is clear, so a direct judgment rule is adopted to balance accuracy and interactive experience.

[0050] Specifically, in this embodiment, when determining initial candidate nodes, content nodes and interactive control nodes employ differentiated judgment rules. For content nodes, a first bounding box is first generated for each node, and candidate bounding boxes containing the coordinates to be matched are selected. Then, a corresponding minimum circumcircle is drawn for each candidate bounding box. The final hit proximity coefficient is obtained by dividing the Euclidean distance from the click position to the center of the circle by the circle radius. The smaller the hit proximity coefficient, the closer the click position is to the node center, and the clearer the instructor's intention to hit the node. Only when the proximity coefficient does not exceed a preset threshold is the node determined as an initial candidate node, thereby excluding cases where the click falls on the node edge or is a misclick.

[0051] For interactive control nodes, since the clickable range of the control itself is clear and the probability of accidental touch is low, this embodiment directly determines the node by using the second spatial bounding box. As long as the clicked position falls within the second spatial bounding box, it is marked as an initial candidate node, ensuring the responsiveness of interactive operations.

[0052] For example, the first rectangular bounding box of a content node has a width of 200px and a height of 80px, with a minimum outer circle radius of approximately 108px and the center of the circle located at the geometric center of the rectangle. The Euclidean distance between the instructor's click position and the center of the circle is approximately 32px, resulting in a hit proximity coefficient of approximately 0.3. Since the preset proximity threshold is 0.6, and 0.3 is less than the threshold of 0.6, this node is determined to be an initial candidate node. Furthermore, if the click position falls on the edge of the node and the proximity coefficient is 0.7, exceeding the threshold of 0.6, it is determined to be a false click and is not included in the initial candidate node list.

[0053] This embodiment employs differentiated initial candidate determination rules for content nodes and interactive control nodes. Content nodes are determined through a proximity coefficient to achieve refined judgment, effectively filtering out interference from accidental clicks at the edge and improving the accuracy of node matching. Interactive control nodes are determined using direct determination rules to ensure the response efficiency of interactive operations.

[0054] In some embodiments, due to the sequential nature of slideshow animation, content nodes and interactive controls are not always active, and undisplayed content and inactive controls cannot be the actual targets for the instructor's clicks. This application, through temporal interval coverage verification, can eliminate candidate nodes that are not active at the current moment, filtering out invalid matches caused by spatial overlap and predicted clicks. Therefore, based on the initial screening of candidate nodes, to further eliminate invalid interactions corresponding to inactive nodes and ensure the consistency between node matching results and the actual lecture playback progress, this application can further specify the temporal validity filtering process for candidate nodes, namely: Based on the current playback sequence status, the candidate node set is filtered for temporal validity to obtain several target nodes corresponding to the click interaction, including: Extract the current animation sequence number corresponding to the current playback sequence state; For each content node to be verified in the candidate node set, the content node to be verified whose effective time interval includes the current animation time sequence number will be marked as the target node; where the effective time interval is used to indicate the range of consecutive animation time sequence numbers in which the content node is in a visible display state. For each interactive control node to be verified in the candidate node set, the interactive control node whose interaction effective timing interval contains the current animation timing number is marked as the target node; where the interaction effective timing interval is used to indicate the range of continuous animation timing numbers in which the interactive control node is in a state that can respond to interactive operations.

[0055] To illustrate, the timing validity filtering is based on the matching relationship between the node's effective timing interval and the current animation timing. First, the current animation timing number is extracted from the current playback timing state, indicating which animation step is currently being played. Then, each node in the candidate node set is verified: for content nodes, it is checked whether their effective timing interval includes the current timing number; if so, the node is currently visible, the click is valid, and it is included in the target node list. For interactive control nodes, it is checked whether their effective interaction timing interval includes the current timing number; if so, the control is currently responsive and is included in the target node list. Nodes that do not meet the conditions are filtered out, thus excluding invalid clicks corresponding to ineffective nodes.

[0056] For example, the current animation sequence number is step 2, and the candidate node set contains three secondary content nodes; the display effective range of the first secondary content node is [2,4], which includes the current sequence number 2, so it is determined to be valid and marked as the target node; the display ranges of the other two secondary content nodes are [3,4] and [4,4], respectively, neither of which includes the current sequence number 2, so they are determined to be invalid and are directly filtered.

[0057] This embodiment filters candidate nodes by using the playback sequence status to eliminate nodes that are not currently in a display or interactive active state. This effectively excludes invalid interactions corresponding to inactive content, avoids including knowledge points that have not yet appeared in the weight statistics, and makes the node matching results more in line with the actual teaching playback progress, thus greatly improving the accuracy of click interaction and knowledge point mapping.

[0058] For step S400, in some embodiments, based on the emphasis logic of actual teaching, it is known that direct clicks have the highest emphasis, while the emphasis of related nodes decreases layer by layer. In order to make the weight quantification results of each node more closely match the actual knowledge emphasis, this application can also realize refined weight quantification of all related nodes for a single click, then: Based on the characteristics of the operation behavior and the branch type to which each target node belongs, the node weight update rules corresponding to the click interaction operation are determined, including: For each target node, the basic weight value and weight update polarity are determined based on the branch type to which the target node belongs and the characteristics of its operation behavior. Based on the branch type to which the target node belongs, select the valid heterogeneous association edges from each heterogeneous association edge; All nodes pointed to by valid heterogeneous association edges are marked as valid nodes, and a weight transmission link is constructed based on the weight transmission rules corresponding to the valid heterogeneous association edges, each valid node, and the target node. When traversing each effective node in each weight propagation link, the effective propagation weight increment of the effective node is calculated based on the weight update polarity, the propagation attenuation coefficient in the weight propagation rule, and the propagation weight increment associated with the effective node in the previous level. During the first traversal, the calculation is performed using the base weight value of the target node as the propagation weight increment of the previous level. The basic weight values, weight update polarities, and effectively transmitted weight increments in each weight transmission link are summarized to form the node weight update rules.

[0059] When selecting valid heterogeneous edges from various heterogeneous edges, the type of the valid heterogeneous edge can be determined according to the branch type to which the target node belongs. At the same time, heterogeneous edges of the corresponding type are selected layer by layer and all are considered as valid heterogeneous edges.

[0060] Optionally, in the actual selection of multiple valid nodes in this embodiment, the target node is used as the starting center point. First, all heterogeneous association edges in the knowledge hierarchy tree that have direct or indirect connection with the target node are traversed. Then, the effective heterogeneous association edges are filtered in batches according to the branch attributes of the target node. Subsequently, all nodes pointed to by all effective heterogeneous association edges are uniformly collected into a set of candidate valid nodes. The same node that is repeatedly pointed to by multiple association edges is deduplicated to ensure that the same node is marked only once, avoiding the error of repeated superposition during subsequent weight transmission.

[0061] Specifically, when the target node belongs to a page content branch, the valid nodes retained after filtering include both first-level adjacent nodes directly connected to the target node through semantic association edges and second-level adjacent nodes connected to the first-level adjacent nodes through semantic association edges. Therefore, the first-level and second-level adjacent nodes pointed to by the semantic association edges will be included in the set of valid nodes. When a target node belongs to a page interaction timeline branch, if the target node corresponds to multiple jump links and points to different target content nodes, then all content nodes pointed to by the jump directions will be marked as valid nodes. Simultaneously, based on the unidirectional pointing attribute of the timeline jump edges, valid nodes only include content nodes on the jump endpoint side and will not extend backward along the associated edges, thus ensuring the directional transmission logic of knowledge content through interactive operations. After completing the selection and layering of all valid nodes, the system then combines the weight transmission rules of the corresponding type of associated edges to build a layer-by-layer weight transmission link according to the association level from near to far, providing the link topology for subsequent layer-by-layer incremental calculations.

[0062] Schematic, this embodiment constructs a node weight update rule in four stages: First, for each target node, the basic weight value corresponding to this interaction is calculated based on its branch type and click behavior characteristics, while simultaneously determining the polarity of the weight update; Second, based on the branch type of the target node, corresponding effective heterogeneous association edges are selected to determine the associated effective nodes. A complete weight transmission link is built, starting with the target node and using the effective nodes as transmission nodes; Third, each transmission link is traversed layer by layer in a specified direction. The transmission weight of the previous level is used as the base, multiplied by the transmission attenuation coefficient to obtain the effective transmission weight increment of the current level node. The first transmission uses the basic weight value as the base; Fourth, the basic weights, update polarities, and transmission weight increments of each node in all transmission links are summarized to form a complete set of node weight update rules, used to guide the weight update of all levels of nodes during a single click interaction by the teacher.

[0063] Specifically, in the calculation stage of layer-by-layer traversal, this embodiment follows the transmission logic of chain decay, that is, the influence of weight is transmitted layer by layer along the transmission link. The weight increment of each effective node at each level is determined only by the transmission increment of the related node at the previous level and the corresponding transmission decay coefficient. The farther away from the starting target node, the smaller the weight increment. This is consistent with the inherent logic that the further away from the knowledge point in actual teaching, the less emphasis it receives. Among them, the calculation of the effective nodes in the first level is the first traversal. The basic weight value of the target node is directly used as the transmission weight increment of the previous level for calculation. Each subsequent level uses the effective transmission weight increment calculated by the previous level as the input base for the current level calculation. The increment of the current level is solved by combining the weight update polarity and the corresponding transmission decay coefficient.

[0064] Specifically, when the target node is a knowledge point node under a content branch, with a base weight value of 1.5 and determined to be a positive polarity update, the selected semantic association edges can form a two-level transmission link. The first level consists of first-level adjacent content nodes directly connected to the target node, and the second level consists of second-level adjacent content nodes connected to the first-level adjacent content nodes. Both belong to valid nodes pointed to by valid heterogeneous association edges. When traversing the first level of valid nodes for the first time, the initial transmission weight value of 1.5 is used as the base, multiplied by the first transmission attenuation coefficient of 0.6 corresponding to the semantic association edge, resulting in an effective transmission weight increment of 0.9 for the first-level adjacent content nodes. When traversing to the second level of valid nodes, the transmission increment of 0.9 obtained from the previous level is used as the base, multiplied by the first transmission attenuation coefficient of 0.6, resulting in an effective transmission weight increment of 0.54 for the second-level adjacent content nodes. The weight decreases step by step as the association level increases.

[0065] In the context of interactive temporal branch propagation, if the target node is a control node corresponding to the content playback class, with a base weight of 1.2 and determined to be a positive polarity update, the selected temporal jump edge forms a single-level propagation link. The target content node pointed to by the jump is the first-level effective node. When calculating, the initial propagation weight value of 1.2 is used as the base, multiplied by the second propagation attenuation coefficient of 0.8 corresponding to the temporal jump edge, resulting in a propagation weight increment of 0.96 for the effective node. Comparing the calculation results of the two types of related edges, it can be seen that under similar initial propagation weights, the semantic related edge has a larger weight attenuation and higher propagation loss, corresponding to its weak association attribute characteristics; the temporal jump edge has a higher weight retention ratio and lower propagation loss, corresponding to its strong association pointing attribute.

[0066] Understandable, Figure 3 This diagram illustrates the node weight propagation link structure corresponding to the aforementioned weight propagation logic, fully presenting the hierarchical propagation process and numerical calculation relationships under both types of branches. Figure 3 In the semantic association propagation link, the target node of the content node is taken as the starting node, and its basic weight value of 1.5 is the initial propagation value. It propagates to the right layer by layer along the semantic association edge. For each level node, the weight increment is multiplied by the first propagation attenuation coefficient of 0.6. The propagation increment of the first-level adjacent node in the first layer is 1.5×0.6=0.9. The second-level adjacent node in the second layer continues to attenuate based on the propagation increment of the previous layer, resulting in a propagation increment of 0.9×0.6=0.54.

[0067] for Figure 3 In the time-series jump propagation link, the target node of the control node is used as the starting node, and its basic weight value of 1.2 is used as the starting propagation value. It propagates unidirectionally to the target content node along the time-series jump edge, and the propagation increment is 1.2×0.8=0.96. The overall attenuation is smaller and the weight retention ratio is higher, which corresponds to the strong correlation attribute of the jump direction.

[0068] This embodiment uses a combination of basic weights and transmission weights to quantify the emphasis of directly clicked nodes and reasonably transmit the weight influence to related nodes, avoiding the one-sidedness of weights caused by updating only a single node. At the same time, each transmission link in this embodiment is calculated independently, which can accurately reflect the weight influence of a single click on the entire knowledge point link, making the weight quantification more in line with the knowledge linkage logic of actual teaching.

[0069] Furthermore, the click behavior of nodes in different branches corresponds to completely different semantics. For example, the click behavior of content nodes directly reflects the instructor's attention to the knowledge points, with duration and frequency being the criteria for judgment; the click behavior of interactive control nodes reflects the instructor's teaching rhythm intention, with operation type being the core criterion for judgment. Therefore, based on the overall framework of the weight update rules, to further improve the accuracy of basic weight calculation and accurately identify the teaching semantics corresponding to the click behavior of different branch nodes, this application can further refine and limit the determination process of basic weight values ​​and weight update polarity, resulting in: Operational behavior characteristics include: operation type, click dwell time, and click frequency per unit time; weight update polarity includes: positive polarity update indicating an increase in weight value or negative polarity update indicating a decrease in weight value; Based on the branch type and operational behavior characteristics of the target node, determine the basic weight value and weight update polarity, including: The node behavior intensity coefficient is calculated based on the click dwell time and the click frequency per unit time. When a target node is detected to belong to a branch of page content, the node behavior intensity coefficient is multiplied by the preset node baseline weight value corresponding to the target node to obtain the basic weight value. If the click frequency per unit time is greater than the preset click frequency threshold, the weight update polarity is determined to be positive; if the click frequency per unit time is not greater than the preset click frequency threshold, the weight update polarity is determined to be negative. When the target node is detected to belong to the page interaction sequence branch, the node behavior intensity coefficient is multiplied by the preset operation baseline weight value corresponding to the operation type to obtain the basic weight value; if the operation type is content playback, the weight update polarity is determined to be positive; if the operation type is content skip, the weight update polarity is determined to be negative.

[0070] Understandably, the base weight value is obtained by multiplying the node behavior intensity coefficient by the corresponding baseline weight value. The node behavior intensity coefficient can be calculated using the formula "click dwell time / baseline dwell time × click frequency per unit time / baseline click frequency," and is used to quantify the emphasis of a single interaction. Dwell time reflects the depth of attention for a single click, while click frequency per unit time reflects the density of the interaction. By comparing these two values ​​with the corresponding baseline value, the node behavior intensity coefficient can be determined. A higher node behavior intensity coefficient indicates a higher degree of emphasis for the node during the interaction.

[0071] The polarity of weight updates is determined differently based on branch type. For target nodes in content branches, high-frequency clicks indicate that the instructor is emphasizing the knowledge point, so the target node is judged as a positive polarity update, and the weight increases; while a single light touch indicates that the instructor is not paying attention, so it is judged as a negative polarity update, and the weight decreases. For nodes in interactive sequence branches, content playback operations indicate that the instructor has emphasized the content, so they correspond to a positive polarity update; content skip operations indicate that the instructor has skipped the current content, so they correspond to a negative polarity update.

[0072] For example, let the preset click frequency threshold be 1 click / second. When the target node belongs to a branch of page content, its click frequency is 2 clicks / second. Since this is greater than 1 click / second, the weight update polarity of the target node is determined to be a positive polarity update.

[0073] This embodiment quantifies the emphasis of clicks through a behavior intensity coefficient, enabling refined calculation of basic weights. Differentiated polarity judgment logic is adopted for content branches and interaction sequence branches, which can accurately identify the teaching intentions of different types of clicks, distinguish the different effects of emphasis-type operations and skip-type operations on weights, avoid misjudgment of weight direction caused by a single rule, and further improve the accuracy of quantifying the importance of knowledge points.

[0074] Furthermore, the business meanings and application scenarios of heterogeneous association edges are completely different. Semantic association serves the knowledge linkage between content nodes, while temporal jump serves the linkage from interactive controls to content. To avoid the confusion of weight logic caused by cross-category transmission, this application can classify heterogeneous association edges and clearly define the selection rules for valid heterogeneous association edges, thus: Heterogeneous association edges include: semantic association edges used to characterize the semantic similarity relationship between nodes and temporal jump edges used to indicate the jump direction relationship between nodes; Based on the branch type to which the target node belongs, valid heterogeneous association edges are selected from all heterogeneous association edges, including: When the target node belongs to a page content branch, the semantically related edges that are connected to the target node will be considered as valid heterogeneous related edges. When the target node belongs to the page interaction timeline branch, the timeline jump edge that has a connection relationship with the target node will be regarded as a valid heterogeneous association edge.

[0075] Schematic, semantic association edges represent the semantic thematic associations between knowledge points, while temporal jump edges represent the jump direction relationship after a control is clicked. In this embodiment, when filtering valid heterogeneous association edges, edges of the corresponding category can be matched according to the branch type of the target node. If the target node is a content node, only semantic association edges are filtered as valid edges, and the weight is only transmitted between content nodes through semantic relationships. If the target node is an interactive control node, only temporal jump edges are filtered as valid edges, and the weight is only transmitted along the jump direction to the target content node. The two types of edges do not intersect, avoiding logical confusion in weight transmission.

[0076] For example, when the target node is a second-level content node and belongs to a page content branch, only semantically related edges connected to it are selected as valid heterogeneous related edges; temporal jump edges are not included in the selection. If the target node is a jump control node and belongs to a page interaction temporal branch, only temporal jump edges connected to it are selected as valid heterogeneous related edges.

[0077] This embodiment achieves differentiated control over the weight transmission path of different branch nodes by classifying and filtering effective heterogeneous association edges. The weight of content nodes is transmitted only along semantic associations, and the weight of interactive controls is transmitted only along jump directions. This ensures the reasonable diffusion of association weights and avoids weight distortion caused by invalid transmission, thereby improving the accuracy and interpretability of weight transmission.

[0078] Based on the above classification and filtering rules, differentiated filtering processes can be performed according to the actual number and type of heterogeneous related edges corresponding to the target node, in order to adapt to slide page structures of different complexities, as follows: In some embodiments, when the target node belongs to the page content branch and there is only one semantic association edge, or belongs to the page interaction time sequence branch and there is only one time sequence jump edge, the heterogeneous association edge of the corresponding type is directly determined as a valid heterogeneous association edge, and a single-level weight transmission link can be constructed based on the single valid node pointed to by the edge.

[0079] In some embodiments, when there are multiple heterogeneous association edges of the type corresponding to the target node, all heterogeneous association edges of the corresponding type are determined as valid heterogeneous association edges, and all valid nodes pointed to by each edge are simultaneously included in the propagation range. Multiple independent weight propagation links are constructed with the target node as the starting point. If multiple links point to the same valid node, the propagation increment of that node is deduplicated and merged to ensure that the same node is only superimposed with the same level of propagation weight once, avoiding weight quantization distortion caused by repeated superposition.

[0080] In some embodiments, when the target node does not have a corresponding type of heterogeneous associated edge, valid heterogeneous associated edges are not filtered, nor are valid associated nodes expanded. The weight propagation link of a single node is formed solely by the target node itself, and the weight update only applies to the target node itself and does not propagate outward. In this scenario, the basic weight value is directly used as the unique weight increment of the node, ensuring clear boundaries for weight quantification and avoiding unfounded weight diffusion.

[0081] In some embodiments, when the target node has only heterogeneous associated edges of a different type, such as nodes in the page content branch only being connected to time-series jump edges, or nodes in the page interaction time-series branch only being connected to semantic associated edges, these non-corresponding associated edges are not included in the selection range of valid heterogeneous associated edges. They are treated as if there are no heterogeneous associated edges of a different type, that is, only the weight of the target node itself is updated, and cross-type weight transmission is not triggered to avoid confusion in the transmission logic.

[0082] In some embodiments, based on the effective edge classification and screening, in order to further match the strength differences of different associations and make the weight transmission results more consistent with the knowledge association logic of actual teaching, this application sets differentiated weight transmission rules for different types of heterogeneous association edges, so that different weight transmission rules correspond to different weight transmission links. Regarding the construction process of the weight transmission links, the following applies: Based on the weight propagation rules corresponding to effective heterogeneous association edges, each effective node, and the target node, a weight propagation link is constructed, including: If the effective heterogeneous association edge is a semantic association edge, the first transmission attenuation coefficient and the weight transmission direction used to indicate the transmission along the semantic similarity association relationship between nodes are retrieved as the weight transmission rule. If the effective heterogeneous association edge is a time-series jump edge, the second conduction attenuation coefficient and the weight conduction direction used to indicate the conduction along the jump direction are retrieved as the weight conduction rule; wherein, the first conduction attenuation coefficient is less than the second conduction attenuation coefficient; Starting from the target node, connect each valid node sequentially along the weight propagation direction according to the associated hierarchy to generate a weight propagation link.

[0083] It is understandable that different types of effective heterogeneous association edges correspond to different weight transmission rules, with the main differences being in the transmission attenuation coefficient and transmission direction. Semantic association edges correspond to the first transmission attenuation coefficient and bidirectional transmission direction, where weights can be transmitted along the association edge to nodes in both directions. Its attenuation coefficient is smaller than that of temporal jump edges, meaning the weight retention ratio is lower. Temporal jump edges correspond to the second transmission attenuation coefficient and unidirectional transmission direction, where weights can only be transmitted from the control node to the target content node along the jump direction. Its attenuation coefficient is larger than that of semantic association edges, meaning the weight retention ratio is higher.

[0084] Specifically, since semantic similarity is a bidirectional equivalence, it propagates bidirectionally; while jump direction is a unidirectional direction, it propagates unidirectionally. Furthermore, since semantic association is a weak association, the weight propagation loss is greater, resulting in a lower weight retention rate; because jump direction is a clear strong association, the weight propagation loss is smaller, resulting in a higher weight retention rate.

[0085] This embodiment sets differentiated transmission attenuation coefficients and transmission directions for two types of heterogeneous related edges, matching different association strengths and attributes of semantic association and jump direction. Strong associations have high retention while weak associations have low retention. The direction is consistent with the association logic, making the quantification results of weight transmission more in line with the actual impact of different association relationships, and further improving the rationality and accuracy of weight quantification.

[0086] To more clearly illustrate the overall execution process of this application, the following example uses a single-slide teaching presentation scenario from a university microeconomics course. Specifically, this slide presents the teaching content on "Price Elasticity of Demand." The page's content branches include a first-level content node, "Overview of Price Elasticity of Demand," and three second-level content nodes: "Definition and Connotation of Elasticity," "Classification Criteria of Elasticity," and "Practical Applications of Elasticity." These three nodes are sequentially linked by semantic association edges; that is, "Definition and Connotation of Elasticity" is connected to "Classification Criteria of Elasticity," and "Classification Criteria of Elasticity" is connected to "Practical Applications of Elasticity." The page's interactive timing branch includes a "Quickly Skip Practical Applications of Elasticity" control, which points to the second-level content node "Practical Applications of Elasticity" via a timing jump edge. The preset click frequency threshold is 1 click / second. The preset node baseline weight value corresponding to the second-level content node "elastic definition connotation" is 1.0. The preset operation baseline weight value corresponding to the operation type is 1.0 when it is content playback and 1.2 when it is content skip. The first transmission attenuation coefficient corresponding to the semantic association edge is 0.5, and the second transmission attenuation coefficient corresponding to the temporal jump edge is 0.75. The initial weight of all nodes is 1.0, and the maximum transmission level is set to 2 levels.

[0087] Please see Figure 4 , Figure 4 This is a single-page slide diagram of the knowledge hierarchy tree corresponding to the teaching scenario in this embodiment, used to present the dual-branch node topology and heterogeneous relationships of this slide. Specifically, as shown... Figure 4As shown, the single-page slide node serves as the top-level node of the page, branching downwards into independent page content branches and page interaction sequence branches. The page content branches employ a multi-level tree structure. The top level is the first-level content node "Overview of Demand Price Elasticity," and the lower level connects three second-level content nodes through hierarchical relationships: "Definition of Elasticity," "Classification Standards of Elasticity," and "Practical Application of Elasticity," from left to right. These three second-level content nodes are sequentially linked by dashed semantic association edges, forming a hierarchical semantic association link. The page interaction sequence branch includes a "Quickly Skip to Practical Application of Elasticity" interactive control node, which points to the second-level content node "Practical Application of Elasticity" via a time-jump edge with a one-way arrow.

[0088] When the instructor clicks on the secondary content node "Flexible Definition of Connotation" for detailed explanation, the click duration is 1.2 seconds and the click frequency is 1.8 times / second. After being determined as a valid target node through time-series validity verification, the node behavior strength coefficient of 1.2 is first calculated, and then multiplied by the preset node baseline weight value of 1.0 corresponding to the secondary content node "Flexible Definition of Connotation" to obtain the basic weight value of 1.2. At the same time, since 1.8 times / second is greater than 1 time / second, and the target node belongs to the page content branch, the weight update polarity is determined to be positive.

[0089] Subsequently, based on the content branch attributes, all semantically related edges are filtered as effective heterogeneous related edges, resulting in two effective nodes: the first-level adjacent node "Elastic Classification Standard" and the second-level adjacent node "Elastic Practical Application," constructing a two-level transmission link. During layer-by-layer calculation, since "Elastic Definition Connotation" is the target node, its corresponding base weight value is 1.2. Therefore, the first-level adjacent node "Elastic Classification Standard" is multiplied by the first transmission attenuation coefficient 0.5, resulting in an effective transmission weight increment of 0.6. Then, the second-level adjacent node "Elastic Practical Application" receives the transmission increment from the previous level, which is based on the effective transmission weight increment of the first-level adjacent node "Elastic Classification Standard" (0.6) and multiplied by 0.5, resulting in an effective transmission weight increment of 0.3. It is easy to see that the weight of each node gradually decreases with increasing transmission level, aligning with the logic of university classroom teaching where greater distances result in lower emphasis.

[0090] In the context of skipping operations within the page interaction sequence branches, when the instructor clicks the "Quickly Skip Flexible Practical Applications" control on the page, this control belongs to the page interaction sequence branch, the operation type is content skip, the click dwell time is 0.6s, and the click frequency is 1 time / second. After timing validity verification, it is determined to be a valid target node. First, combining the preset baseline dwell time of 0.9s and the baseline click frequency of 2 times / second, the node behavior strength coefficient is calculated to be approximately 0.333. Multiplying this by the preset baseline weight value of 1.2 for the content skip category yields a basic weight value of approximately 0.4. Since the target node belongs to the page interaction sequence branch and the operation type is content skip, it indicates that the instructor quickly skipped the corresponding content without emphasizing it. Therefore, the weight update polarity is determined to be negative.

[0091] Subsequently, based on the interaction timing branch attributes, the timing jump edge is selected as a valid heterogeneous association edge. The skip control is directed to the secondary content node "Elastic Practical Application" through the timing jump edge, which is the skipped content node corresponding to this operation. Therefore, the target content node is marked as a valid node, and a single-level unidirectional weight transmission link is constructed. The transmission direction is consistent with the jump direction, and the second transmission attenuation coefficient of 0.75 is matched with the timing jump edge. During layer-by-layer calculation, "Quick Skip Elastic Application" is used as a skip control, and its own node weight increment is -0.4. The effective content node "Elastic Application" pointed to by the time-series jump edge uses the absolute value of the base weight value of 0.4 as the starting transmission base, multiplied by the second transmission attenuation coefficient of 0.75 to obtain an effective transmission weight increment of 0.3. Furthermore, combined with the negative polarity update rule, the actual effective weight increment of the "Elastic Application" node is -0.3. Since the initial weight of all nodes is 1.0, and the effective transmission weight increment based on the previous click operation is 0.3, the final weight update of the "Elastic Application" node after the update is: 1.0 (initial weight) + 0.3 (previous effective weight increment) + (-0.3) (current effective weight increment) = 1.0.

[0092] Comparing the propagation results of the two types of related edges reveals that, at the same propagation level, semantically related edges, using a smaller first propagation attenuation coefficient of 0.5, exhibit greater weight decay with each level and higher propagation loss, corresponding to weakly related attributes with similar semantics. Conversely, temporal jump edges, using a larger second propagation attenuation coefficient of 0.75, retain a higher weight ratio and exhibit lower propagation loss, corresponding to strongly related attributes to which the jump points. Furthermore, active clicks on content branches correspond to a positive weight increase, while skipping interactive branches corresponds to a negative weight decrease. This two-way interaction more accurately quantifies the actual importance of knowledge points, ensuring a high degree of match between the final weight results and the knowledge emphasis in actual teaching.

[0093] Therefore, as can be seen from the above embodiments, the differentiated weight transmission mechanism of this application not only ensures the dominant weight of directly clicked nodes, but also achieves refined gradual attenuation based on the strength of the association relationship. The final node weight can truly restore the knowledge emphasis in the teaching process, providing reliable data support for the accurate extraction of subsequent teaching content.

[0094] For step S500, in some embodiments, to ensure that the final extracted teaching demonstration content accurately matches the knowledge focus of the actual lecture and has a clear hierarchical logic and a directly reusable structured form, this embodiment can, based on the cumulatively updated node final weights and combined with the original hierarchical structure of the knowledge hierarchy tree, complete the selection and structured generation of core knowledge content, thus: After the slideshow presentation is completed, the final weight values ​​of all content nodes in the knowledge hierarchy tree are first collected. Knowledge point nodes that meet the weight criteria are then selected according to the preset weight screening threshold. Subsequently, based on the original hierarchical relationship within the knowledge hierarchy tree, the selected knowledge point nodes are categorized and arranged in the order of first-level chapters and second-level knowledge points. At the same time, the corresponding text, chart annotations, and other knowledge content in the original slides for each node are extracted, and auxiliary information such as the weight ratio and interaction trigger frequency of the corresponding nodes are added. Finally, the content is integrated to generate a structured teaching demonstration that is clear in hierarchy and highlights the key points.

[0095] To illustrate, when filtering content, the weighted filtering threshold can be flexibly adjusted according to different application scenarios. For example, if a concise review summary is needed, the threshold can be increased to retain only the core knowledge points with the highest weight; if a complete lecture handout is needed, the threshold can be lowered to cover more secondary points. During structured generation, the original logical framework of the knowledge hierarchy tree is strictly reused. For example, parent nodes serve as chapter titles, and child nodes serve as sub-knowledge points, ensuring that the knowledge logic of the generated content is completely consistent with the original courseware and avoiding logical inconsistencies.

[0096] This embodiment uses the final weight of nodes accumulated throughout the entire process as the basis for determining the importance of content, ensuring that the extraction results are entirely determined by the actual teaching interaction behavior, and can objectively restore the instructor's knowledge emphasis rhythm. At the same time, it directly reuses the hierarchical relationship of the knowledge hierarchy tree to organize the output structure, without the need to construct additional knowledge logic, which not only ensures the logical coherence of the generated content, but also improves the execution efficiency of content generation. Therefore, this embodiment achieves accurate extraction of core content through weight threshold filtering, and can automatically filter out secondary knowledge points that are not emphasized, solving the problems of excessive redundant content and lack of emphasis in traditional page-level extraction schemes. Moreover, the structured explanation content generated based on the original hierarchical tree structure can be directly used as classroom handouts, review materials, or training debriefing materials without the need for manual secondary organization, effectively reducing the manual cost of teaching debriefing.

[0097] As can be seen from the above embodiments, this application optimizes the traditional coarse-grained statistical model for pages by addressing the issues of insufficient extraction accuracy and large deviations in the measurement of the importance of knowledge points. It achieves node-level decomposition of slideshow content through a dual-branch knowledge hierarchy tree and implements categorized management of content and interactions. This application also employs a two-step node matching mechanism of spatial initial screening and temporal fine filtering. First, it directly matches and associates nodes through coordinate matching, and then removes ineffective nodes through temporal intervals, effectively filtering out accidental touches and invalid interactions, significantly improving the accuracy of node matching. Simultaneously, it uses differentiated judgment rules for content nodes and interactive control nodes, balancing matching accuracy and interaction response efficiency.

[0098] At the weight quantification level, this application determines the basic weights and update polarities based on branch type and operational behavior characteristics. Then, it constructs a transmission link through heterogeneous association edges selected by classification. Differential attenuation coefficients and transmission directions are set for different association relationships, achieving refined weight quantification from the target node to all nodes in the link, accurately reflecting the degree of emphasis on knowledge points during the teaching process. Furthermore, this application automatically parses slide files to complete the hierarchical tree construction, eliminating the need for manual node and time sequence annotation. It is highly adaptable and can be widely applied to various teaching scenarios such as smart classrooms, online education, and corporate training, effectively reducing the manual cost of teaching review and improving the efficiency and quality of digital teaching.

[0099] Please see Figure 5 This application also provides a device for extracting teaching demonstration content based on click-interaction, which can implement the above-mentioned method for extracting teaching demonstration content based on click-interaction. The device includes: The tree structure acquisition module 601 is used to acquire the knowledge hierarchy tree corresponding to the slide presentation document; wherein, the knowledge hierarchy tree has independent page content branches and page interaction sequence branches for each slide; The click interaction response module 602 is used to respond to the click interaction operation triggered by the user on the slide presentation document during the teaching presentation, obtain the click location information and operation behavior characteristics corresponding to the click interaction operation, and obtain the current presentation sequence state of the current slide. The node filtering module 603, based on the click location information, matches a set of candidate nodes from the page content branches and the page interaction sequence branches; and filters the candidate node set for time sequence validity based on the current playback sequence state to obtain several target nodes corresponding to the click interaction operation. The node weight update module 604 is used to determine the node weight update rule corresponding to the click interaction operation based on the operation behavior characteristics and the branch type to which each target node belongs; and to update the current weight value of each node in the knowledge hierarchy tree level by level based on the node weight update rule. The teaching content extraction module 605 is used to extract the corresponding knowledge content from the PowerPoint presentation document and generate structured teaching demonstration content based on the final weight values ​​updated at each node after the presentation is completed.

[0100] It is understood that the content of the above method embodiments is applicable to the present device embodiments. The specific functions implemented by the present device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0101] It should be noted that the device embodiments described above are merely illustrative. The modules described as separate components may or may not be physically separate, and the components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided in this application, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0102] Those skilled in the art will understand that, for convenience and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0103] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned method for extracting teaching demonstration content based on click-based interaction. This electronic device can include any smart terminal such as a tablet computer or an in-vehicle computer.

[0104] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0105] Please see Figure 6 , Figure 6 This illustrates the hardware structure of an electronic device according to another embodiment, the electronic device comprising: The processor 701 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 702 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 702 can store the operating system and other application programs. When the technical solutions provided in the embodiments of this application are implemented through software or firmware, the relevant program code is stored in the memory 702 and is called and executed by the processor 701. The input / output interface 703 is used to implement information input and output; The communication interface 704 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, Wi-Fi, Bluetooth, etc.). Bus 705 transmits information between various components of the device (e.g., processor 701, memory 702, input / output interface 703, and communication interface 704); The processor 701, memory 702, input / output interface 703 and communication interface 704 are connected to each other within the device via bus 705.

[0106] The processor 701 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting various parts of the terminal device through various interfaces and lines.

[0107] The memory 702 can be used to store computer programs. The processor implements various functions of the terminal device by running or executing the computer programs stored in the memory and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, applications required for at least one function, etc.; the data storage area may store data created based on the use of the mobile phone, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory or plug-in hard disk, smart media card (SMC), secure digital card (SD), flash card, at least one disk storage device, flash memory device or other volatile solid-state storage device.

[0108] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for extracting teaching demonstration content based on click-interaction.

[0109] It is understood that the content of the above method embodiments is applicable to the present computer-readable storage medium embodiments. The specific functions implemented by the present computer-readable storage medium embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0110] Those skilled in the art will understand that all or some of the steps, apparatuses, or functional modules / units in the methods disclosed above can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0111] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for extracting teaching demonstration content based on click-based interaction, characterized in that, The method includes: Obtain the knowledge hierarchy tree corresponding to the slide presentation document; wherein, the knowledge hierarchy tree has independent page content branches and page interaction sequence branches for each slide; In response to a user's click interaction on the slide presentation document during a teaching presentation, the system obtains the click location information and operation behavior characteristics corresponding to the click interaction, and obtains the current presentation sequence status of the current slide. Based on the click location information, a set of candidate nodes is obtained by matching from the page content branches and the page interaction time sequence branches; the candidate node set is then filtered for time sequence validity based on the current playback time sequence state to obtain several target nodes corresponding to the click interaction operation; Based on the operational behavior characteristics and the branch type to which each target node belongs, a node weight update rule corresponding to the click interaction operation is determined; based on the node weight update rule, the current weight value of each node in the knowledge hierarchy tree is updated level by level. After the slideshow presentation is completed, based on the updated final weight values ​​of each node, the corresponding knowledge content is extracted from the slideshow presentation and a structured teaching demonstration is generated.

2. The method for extracting teaching demonstration content based on click-interaction according to claim 1, characterized in that, The process of constructing the knowledge hierarchy tree includes: Based on the content hierarchy attributes and first spatial coordinate attributes of each content element in the slide presentation document, multi-level content nodes are generated, and page content branches are constructed based on each content node. Based on the functional type attributes and second spatial coordinate attributes of each interactive control element in the slide presentation document, multiple interactive control nodes of different types are generated, and the page interaction sequence branch is constructed based on each interactive control node. Based on the original animation timing sequence number corresponding to each slide in the PowerPoint presentation document, a mapping relationship is constructed between the original animation timing sequence number, the display effective timing interval corresponding to the content node, and the interaction effective timing interval corresponding to the interactive control node; The knowledge hierarchy tree is constructed based on the content branches of each page, the interaction sequence branches of each page, and the mapping relationships of each page.

3. The method for extracting teaching demonstration content based on click-interaction according to claim 2, characterized in that, The knowledge hierarchy tree also includes several heterogeneous association edges; these heterogeneous association edges are used to indicate non-hierarchical subordinate relationships between different nodes. The step of matching a set of candidate nodes from the page content branches and the page interaction sequence branches based on the click location information includes: Based on the matching relationship between the click location information and the first spatial coordinate attribute and the second spatial coordinate attribute, several initial candidate nodes are selected. For each initial candidate node, based on the heterogeneous association edges connected to the initial candidate node, the direct adjacent nodes that have the non-hierarchical subordinate association relationship with the initial candidate node are extracted as extended candidate nodes; The initial candidate nodes and the extended candidate nodes are deduplicated and merged to generate the candidate node set.

4. The method for extracting teaching demonstration content based on click-interaction according to claim 3, characterized in that, The step of filtering out several initial candidate nodes based on the matching relationship between the click location information and the first spatial coordinate attribute and the second spatial coordinate attribute, respectively, includes: Based on each of the first spatial coordinate attributes and each of the second spatial coordinate attributes, a first spatial bounding box corresponding to each of the content nodes and a second spatial bounding box corresponding to each of the interactive control nodes are constructed. The click location information is converted into a coordinate point to be matched corresponding to the coordinate system of the slide page, and the first spatial bounding box containing the coordinate point to be matched is marked as a candidate spatial bounding box; wherein, the candidate spatial bounding box corresponds to a minimum circumcircle; For each candidate bounding box, the hit proximity coefficient is calculated based on the Euclidean distance between the coordinate point to be matched and the center of the minimum circumcircle, and the radius of the minimum circumcircle. The candidate bounding boxes whose hit proximity coefficient is less than or equal to a preset proximity threshold are marked as target bounding boxes; the content nodes corresponding to the target bounding boxes are marked as initial candidate nodes. The interactive control node corresponding to the second spatial bounding box containing the coordinates to be matched is marked as the initial candidate node.

5. The method for extracting teaching demonstration content based on click-interaction according to claim 2, characterized in that, The step of filtering the candidate node set based on the current playback time sequence state to obtain several target nodes corresponding to the click interaction operation includes: Extract the current animation sequence number corresponding to the current playback sequence state; For each content node to be verified in the candidate node set, the content node to be verified that contains the current animation sequence number in the display effective timing interval is marked as the target node; wherein, the display effective timing interval is used to indicate the range of consecutive animation sequence numbers in which the content node is in a visible display state; For each interactive control node to be verified in the candidate node set, the interactive control node to be verified that contains the current animation sequence number in the interaction effective timing interval is marked as the target node; wherein, the interaction effective timing interval is used to indicate the range of continuous animation sequence numbers in which the interactive control node is in a state that can respond to interactive operations.

6. The method for extracting teaching demonstration content based on click-interaction according to claim 3, characterized in that, The step of determining the node weight update rule corresponding to the click interaction operation based on the operation behavior characteristics and the branch type to which each target node belongs includes: For each target node, a basic weight value and weight update polarity are determined based on the branch type to which the target node belongs and the operation behavior characteristics. Based on the branch type to which the target node belongs, valid heterogeneous association edges are selected from each of the heterogeneous association edges; All nodes pointed to by the effective heterogeneous association edges are marked as effective nodes, and a weight transmission link is constructed based on the weight transmission rules corresponding to the effective heterogeneous association edges, each of the effective nodes, and the target node. When traversing each effective node in each weight transmission link, the effective transmission weight increment of the effective node is calculated based on the weight update polarity, the transmission attenuation coefficient in the weight transmission rule, and the transmission weight increment associated with the effective node in the previous level; wherein, during the first traversal, the calculation is performed using the base weight value of the target node as the transmission weight increment of the previous level. The basic weight values, weight update polarities, and effective weight increments in each weight transmission link are summarized to form the node weight update rule.

7. The method for extracting teaching demonstration content based on click-interaction according to claim 6, characterized in that, The operational behavior characteristics include: operation type, click dwell time, and click frequency per unit time; the weight update polarity includes: positive polarity update for indicating an increase in weight value or negative polarity update for indicating a decrease in weight value; The step of determining the basic weight value and weight update polarity based on the branch type to which the target node belongs and the operation behavior characteristics includes: The node behavior intensity coefficient is calculated based on the click dwell time and the click frequency per unit time. When the target node is detected to belong to the page content branch, the node behavior intensity coefficient is multiplied by the preset node baseline weight value corresponding to the target node to obtain the base weight value; if the click frequency per unit time is greater than the preset click frequency threshold, the weight update polarity is determined to be the positive polarity update; if the click frequency per unit time is not greater than the preset click frequency threshold, the weight update polarity is determined to be the negative polarity update. When the target node is detected to belong to the page interaction timeline branch, the node behavior intensity coefficient is multiplied by the preset operation baseline weight value corresponding to the operation type to obtain the basic weight value; if the operation type is content playback, the weight update polarity is determined to be the positive polarity update; if the operation type is content skip, the weight update polarity is determined to be the negative polarity update.

8. The method for extracting teaching demonstration content based on click-interaction according to claim 6, characterized in that, The heterogeneous association edges include: semantic association edges used to characterize the semantic similarity relationship between nodes and temporal jump edges used to indicate the jump direction relationship between nodes; The step of filtering valid heterogeneous association edges from the heterogeneous association edges according to the branch type to which the target node belongs includes: When the target node belongs to the page content branch, the semantic association edge that has a connection relationship with the target node is taken as the effective heterogeneous association edge; When the target node belongs to the page interaction timeline branch, the timeline jump edge that has a connection relationship with the target node is taken as the effective heterogeneous association edge.

9. The method for extracting teaching demonstration content based on click-interaction according to claim 8, characterized in that, The construction of a weight transmission link based on the weight transmission rules corresponding to the effective heterogeneous association edges, each effective node, and the target node includes: If the effective heterogeneous association edge is the semantic association edge, the first conduction attenuation coefficient and the weight conduction direction used to indicate the conduction along the semantic similarity association relationship between nodes are retrieved as the weight conduction rule; If the effective heterogeneous association edge is the timing jump edge, the second conduction attenuation coefficient and the weight conduction direction used to indicate conduction along the jump direction are retrieved as the weight conduction rule; wherein, the first conduction attenuation coefficient is less than the second conduction attenuation coefficient; Starting from the target node, connect each of the effective nodes sequentially along the weight propagation direction according to the associated hierarchy to generate a weight propagation link.

10. A device for extracting teaching demonstration content based on click-interaction, characterized in that, The device includes: The tree structure acquisition module is used to acquire the knowledge hierarchy tree corresponding to the slide presentation document; wherein, the knowledge hierarchy tree has independent page content branches and page interaction sequence branches for each slide; The click interaction response module is used to respond to the click interaction operation triggered by the user on the slide presentation document during the teaching presentation, obtain the click location information and operation behavior characteristics corresponding to the click interaction operation, and obtain the current presentation sequence status of the current slide. The node filtering module, based on the click location information, matches a set of candidate nodes from the page content branches and the page interaction time sequence branches; and filters the candidate node set for time sequence validity based on the current playback time sequence state to obtain several target nodes corresponding to the click interaction operation. The node weight update module is used to determine the node weight update rule corresponding to the click interaction operation based on the operation behavior characteristics and the branch type to which each target node belongs; and to update the current weight value of each node in the knowledge hierarchy tree level by level based on the node weight update rule. The teaching content extraction module is used to extract the corresponding knowledge content from the slide presentation document and generate structured teaching demonstration content based on the final weight values ​​of each node after the slide presentation document has finished playing.

Images