A method and system for adaptive construction of video shot data structure

Abstract

This invention discloses an adaptive construction method and system for video storyboard data structure, relating to the field of storyboard construction technology. It addresses the problems of scattered storage of related events and unreasonable hierarchical division of structures. The method constructs a video storyboard data structure by analyzing the narrative structure characteristics of the script to be processed, collecting character switching frequency and scene transition span, calculating narrative jump index to identify event nodes, statistically analyzing entity repetition frequency and action duration, generating event association strength and constructing an event association density map, extracting the path penetration depth of event links, generating structural hierarchy factors, stratifying event nodes to form storyboard structure information, statistically analyzing cross-layer reference frequency and previous frame dependency ratio, generating node popularity, and adaptively adjusting the storage structure of storyboard structure information accordingly. This improves the organization efficiency, retrieval efficiency, and structural consistency of video storyboard data.

Description

Technical Field

[0001] This invention relates to the field of storyboard construction technology, and more specifically, to an adaptive construction method and system for video storyboard data structure. Background Technology

[0002] With the continuous development of short video production, film and television content generation, and AI-based automated video production technology, the automatic conversion of scripts into video content has gradually become an important part of digital content production. In existing technologies, in order to support shot planning, scene organization, character scheduling, and subsequent video generation and processing, it is usually necessary to first convert the script content into a corresponding storyboard data structure, and then organize and manage the relationships between characters, scenes, actions, and events in the script through the storyboard data structure.

[0003] The existing technology has the following shortcomings: Currently, existing technologies typically use preset rules or fixed hierarchical structures to organize the script content into storyboards, lacking the ability to dynamically analyze the internal event relationships, the continuous participation of characters, and the evolutionary characteristics of the narrative chain. This results in the constructed storyboard data structure failing to accurately reflect the event organization relationships in complex narrative scenarios, and is prone to issues such as scattered storage of related events, unreasonable structural hierarchy, and decreased data access efficiency. Therefore, an adaptive construction method and system for video storyboard data structure is proposed.

[0004] The information disclosed in the background section is only intended to enhance the understanding of the background of this disclosure, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0005] To overcome the aforementioned deficiencies of the prior art, embodiments of the present invention provide an adaptive construction method and system for video storyboard data structure. This method and system address the problems mentioned in the background art by employing an event node recognition mechanism based on narrative transition features, an event association density construction mechanism based on entity reproduction and action continuity relationships, a storyboard structure layering mechanism based on path penetration depth, and an adaptive adjustment mechanism for storage structure based on node popularity.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an adaptive construction method for video storyboard data structure, comprising the following steps: Step S1: After receiving the script to be processed, perform text scanning on the script to be processed, collect the character switching frequency and scene transition span of the script to be processed, calculate the narrative leap index, and identify the event nodes of the script to be processed based on the narrative leap index. Step S2: Obtain the entity information of the event node and count the number of entity recurrences, detect the action duration of the event node, generate the association strength of the event node based on the number of entity recurrences, generate the event chain based on the association strength and construct the event association density map; Step S3: Extract the path penetration depth of the event link based on the event association density map, analyze the structural hierarchy factor based on the path penetration depth, use the structural hierarchy factor to index and classify the event nodes and output the storyboard structure information. Step S4: Calculate the cross-layer reference frequency and the previous frame dependency ratio of the storyboard structure information, generate node popularity by combining the cross-layer reference frequency and the previous frame dependency ratio, and determine whether to adjust the storage structure of the storyboard structure information based on the node popularity.

[0007] In a preferred embodiment, in step S1, after receiving the script to be processed, text scanning processing is performed on the script to be processed. Text scanning refers to scanning the script text sentence by sentence based on preset text units, and performing character identification and scene identification parsing on each text unit. Within each text unit, the dominant role identifier corresponding to each text unit is continuously compared. When the dominant role identifier of each text unit changes, it is determined as a role switching event. The number of role switching events is counted to obtain the number of role switching events. The role switching frequency is obtained by normalizing the number of role switching times with the number of sentences in the text unit.

[0008] In a preferred embodiment, in step S1, scene identifier parsing is performed on the text content in each text unit. When the scene semantic identifier of the text unit changes, a scene transition event is determined to have occurred. The number of scene transitions within the text unit is counted, and the scene transition span is calculated. The average scene transition span is obtained by summarizing and averaging all scene transition spans within the text unit, and a narrative leap index is constructed by combining the character switching frequency. When the narrative transition index is greater than or equal to the preset event recognition threshold, the text interval corresponding to the text unit is marked as a candidate event interval; When the narrative transition index is less than the preset event recognition threshold, it is determined that no significant narrative structure change occurs in the text interval corresponding to the text unit; Adjacent candidate event intervals are merged to form continuous event intervals as event nodes.

[0009] In a preferred embodiment, in step S2, entity information extraction processing is performed on each event node. Entity information refers to person entities, location entities, and key object entities extracted from the text unit set corresponding to the event node. The entity recurrence count is the cumulative number of times the entity identifier appears repeatedly in the event node; The action continuation length is the length of continuous text units covered by the same action predicate within the same event node; Calculate the reproducibility coefficient between two event nodes based on the number of entity reproducibility cycles; Construct an action consistency coefficient between two event nodes. The action consistency coefficient is a normalized similarity based on the difference in action duration. The association strength of event nodes is constructed using the product normalization method.

[0010] In a preferred embodiment, in step S2, event nodes are used as node units in the association density graph, and association strength is used as the basis for connecting nodes to generate event links. By performing a traversal analysis on all event nodes, an event link network is formed, consisting of the event node set and the event link set. The link connection status of each event node is statistically processed to calculate the event association density. The corresponding event association density is written into the event link network as a node attribute, and a topology mapping is performed according to the link connection relationship between event nodes to construct an event association density graph.

[0011] In a preferred embodiment, in step S3, each event node in the event association density graph is used as the starting node, and a continuous path search is performed along the connection direction in the event link network to obtain an effective through path. The number of event links contained in the effective through path is counted and recorded as the path length. The total number of valid connecting paths in the statistical event association density graph is calculated, and the path connection depth is calculated in combination with the path length. The event association density of all event nodes in the event association density graph is statistically analyzed to obtain the mean event association density. The structural hierarchy factor is constructed by combining the path penetration depth with the mean event association density. Using the integer part corresponding to the structural hierarchy factor as the single-layer coverage span, the hierarchy division interval is constructed in a continuous interval manner.

[0012] In a preferred embodiment, in step S3, the event nodes are subjected to index structure hierarchical processing according to the structural hierarchy factor, and the shortest link distance from the event node to the core event node of the event link network is calculated to form the event node hierarchical distance value. Define the event node with the highest event correlation density as the core event node; Event nodes are categorized into hierarchical levels based on their distance values, forming an event node index structure consisting of multiple index levels. The storyboard structure information is generated based on the event nodes contained in each index level, the connection relationship between levels, and the corresponding event link information. The storyboard structure information includes the event node identifier, the index level to which it belongs, and the event link pointing relationship.

[0013] In a preferred embodiment, in step S4, after the generation of the storyboard structure information is completed, the node access relationship in the storyboard structure information is statistically analyzed to obtain the cross-layer reference frequency and the previous frame dependency ratio of each storyboard node. A storyboard node refers to an event node that has completed the hierarchical arrangement of the index structure. Within the statistical period, all referencing behaviors occurring between storyboard nodes are cumulatively counted to obtain the total number of references; Select any storyboard node as the current statistical node. If the target storyboard node referenced by the storyboard node is at a different level than the current statistical node, a cross-level reference event is determined to have occurred. The cross-level reference frequency is obtained by dividing the number of cross-level reference events within the statistical period by the total number of references.

[0014] In a preferred embodiment, in step S4, when any selected storyboard node needs to call the data content of the preceding storyboard node during parsing, loading or rendering, it is determined that there is a previous frame dependency event, and the number of previous frame dependency events and the total number of times all storyboard nodes are accessed within the statistical period are counted. Divide the number of previous frame dependent events by the total number of times all split-view nodes are accessed within the statistical period to obtain the previous frame dependency percentage. Coupling normalized cross-layer reference frequency and normalized pre-frame dependency ratio generated node popularity: When the node's popularity is greater than or equal to a preset popularity threshold, the current storyboard node is determined to be a high-popularity node. For high-traffic nodes, the process of adjusting the storage structure of the storyboard information is triggered, and the high-traffic nodes are migrated from the original hierarchical storage area to the high-frequency access storage area. When the node's popularity is less than the preset popularity threshold, the current storyboard node is determined to be a normal node, and the current storage structure remains unchanged.

[0015] An adaptive video storyboard data structure construction system includes a script parsing module, a link construction module, a storyboard layering module, and a structure optimization module. The functions of each module are as follows: After receiving the script to be processed, the script parsing module performs a text scan on the script to be processed, collects the character switching frequency and scene transition span of the script to be processed, calculates the narrative leap index, identifies the event nodes of the script to be processed based on the narrative leap index, and passes the event nodes to the link construction module; The link construction module obtains the entity information of the event node and counts the number of entity recurrences, detects the duration of the event node's action, generates the association strength of the event node based on the number of entity recurrences, generates the event link based on the association strength and constructs the event association density map, and passes the event association density map into the scene layering module. The storyboard layering module extracts the path penetration depth of the event chain based on the event association density map, analyzes the structural hierarchy factor based on the path penetration depth, indexes the event nodes using the structural hierarchy factor, and outputs the storyboard structure information. The storyboard structure information is then passed to the structure optimization module. The structure optimization module counts the cross-layer reference frequency and the previous frame dependency ratio of the storyboard structure information, generates node popularity by combining the cross-layer reference frequency and the previous frame dependency ratio, and determines whether to adjust the storage structure of the storyboard structure information based on the node popularity.

[0016] The technical effects and advantages of this invention are as follows: This invention analyzes the narrative structure features of the script to be processed to construct a video storyboard data structure. Upon receiving the script, it collects character switching frequency and scene transition span, calculates the narrative leap index, and identifies event nodes based on the narrative leap index. Subsequently, it statistically analyzes the entity repetition frequency and action duration of each event node to generate event association strength, and constructs an event association density map based on this strength. Further, it extracts the path penetration depth of event links from the event association density map, generates structural hierarchy factors, and uses these factors to index and categorize event nodes, forming storyboard structure information. Finally, it statistically analyzes the cross-layer reference frequency and previous frame dependency ratio of the storyboard structure information to generate node popularity, and adaptively adjusts the storage structure of the storyboard structure information based on node popularity, thereby improving the organization efficiency, retrieval efficiency, and structural consistency of the video storyboard data. Attached Figure Description

[0017] Figure 1 This is a flowchart illustrating the implementation of an adaptive construction method for video scene data structure according to the present invention.

[0018] Figure 2 This is a module framework diagram of an adaptive video scene data structure construction system according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] This invention analyzes the narrative structure features in the script to be processed to construct a video storyboard data structure. Upon receiving the script, it collects character switching frequency and scene transition span, calculates the narrative leap index, and identifies event nodes based on the narrative leap index. Subsequently, it counts the number of entity repetitions and action duration of each event node to generate event association strength, and constructs an event association density map based on this strength. Further, it extracts the path penetration depth of event links from the event association density map, generates structural hierarchy factors, and uses these factors to index and categorize event nodes, forming storyboard structure information. Finally, it counts the cross-layer reference frequency and previous frame dependency ratio of the storyboard structure information to generate node popularity, and adaptively adjusts the storage structure of the storyboard structure information based on the node popularity.

[0021] Example 1, as Figure 1 As shown, an adaptive construction method for video storyboard data structure includes the following steps: Step S1: After receiving the script to be processed, perform text scanning on the script to be processed, collect the character switching frequency and scene transition span of the script to be processed, calculate the narrative leap index, and identify the event nodes of the script to be processed based on the narrative leap index. Step S2: Obtain the entity information of the event node and count the number of entity recurrences, detect the action duration of the event node, generate the association strength of the event node based on the number of entity recurrences, generate the event chain based on the association strength and construct the event association density map; Step S3: Extract the path penetration depth of the event link based on the event association density map, analyze the structural hierarchy factor based on the path penetration depth, use the structural hierarchy factor to index and classify the event nodes and output the storyboard structure information. Step S4: Calculate the cross-layer reference frequency and the previous frame dependency ratio of the storyboard structure information, generate node popularity by combining the cross-layer reference frequency and the previous frame dependency ratio, and determine whether to adjust the storage structure of the storyboard structure information based on the node popularity.

[0022] The specific implementation is as follows: In step S1, after receiving the script to be processed, text scanning processing is performed on the script to be processed.

[0023] Text scanning refers to scanning the script text sentence by sentence based on preset text units, and performing character identification and scene identification analysis on each text unit. Character identification analysis is used to extract the dominant character identification of the text unit, and scene identification analysis is used to extract the scene semantic identification in the text unit.

[0024] It should be noted that the preset text unit refers to the basic analysis unit formed after the script text is segmented according to a pre-set division rule before the script is scanned. This unit is used to carry out subsequent processing such as character identification, scene identification, and event node extraction. Character identification resolution refers to the semantic character attribution identification processing of each text unit in the script to extract the dominant character identifier corresponding to that text unit in the narrative structure. Scene identification resolution refers to the scene semantic classification processing of each text unit in the script to extract the scene semantic identifier in that text unit.

[0025] After scanning the script to be processed, a sequence of text units is obtained.

[0026] The collection of role switching frequency is based on the role identifier parsing results. Within each text unit, the dominant role identifier corresponding to each text unit is continuously compared. When the dominant role identifier of each text unit changes, it is determined as a role switching event.

[0027] Count the number of character switching events to obtain the character switching count.

[0028] The role-switching frequency is obtained by normalizing the number of role switching times with the number of sentences in the text unit: ; in, Indicates the number of sentences in a text unit. Indicates the cumulative number of handovers. Indicates the frequency of character switching.

[0029] Scene identifiers are parsed for the text content in each text unit. When the scene semantic identifier of a text unit changes, a scene transition event is determined. The number of scene transitions within a text unit is counted, and its scene transition span is calculated. The calculation method is as follows: ; in, Indicates the span of scene transition. This indicates the number of scene transition events that occur within a text unit. This indicates the total number of sentences in a text unit.

[0030] It should be noted that the scene transition span reflects the intensity of abrupt changes in scene switching during the continuous narrative process of the script. The larger the value, the higher the degree of semantic break between scenes and the more significant the jump in narrative space.

[0031] The average scene transition span is obtained by summing and averaging all scene transition spans within a text unit. The calculation method is as follows: ; in, This represents the total number of times the contextual semantic identifier of a text unit changes. Indicates the average scene transition span. Indicates the span of scene transition.

[0032] It should be noted that when When this time, it indicates that no scene transition has occurred within the text unit, and the average scene transition span is [value missing]. .

[0033] After obtaining the character switching frequency and average scene transition span, use The normalization method processes the character switching frequency and maps its value range to... The normalized role switching frequency was obtained.

[0034] A narrative transition index is constructed based on the normalized character switching frequency and average scene transition span to characterize the overall intensity of change in the narrative structure within the current text unit. Its calculation method is as follows: in, Indicates the narrative leap index. Indicates the normalized role switching frequency Indicates the average scene transition span. and The preset weighting coefficient is used to balance the contribution ratio of the character dimension and the scene dimension to the narrative leap index, thereby realizing the quantitative representation of the degree of leap in the script's narrative structure.

[0035] It should be noted that the preset weighting coefficients and The statistical learning approach based on historical script sample sets is determined by selecting a standard script sample set of the same type as the script to be processed. For each sample, the normalized character switching frequency and average scene transition span are calculated. The intensity of real narrative transitions defined by rules is used as the supervision target, and an error function is constructed while satisfying constraints. and Under the premise of minimizing the error function, the optimal weight combination that minimizes the deviation between the narrative transition index and the reference transition strength is obtained by optimizing the parameters.

[0036] After calculating the narrative leap index, event nodes in the script to be processed are identified based on the narrative leap index. Specifically, the narrative leap index of each text unit is compared with a preset event identification threshold. When the narrative leap index is greater than or equal to the preset event identification threshold, it is determined that a significant narrative structural change has occurred in the text interval corresponding to that text unit, and the text interval corresponding to that text unit is marked as a candidate event interval. When the narrative leap index is less than the preset event identification threshold, it is determined that no significant narrative structural change has occurred in the text interval corresponding to that text unit.

[0037] It should be noted that the preset event recognition threshold is used as a boundary standard to characterize the intensity of narrative transitions. Its value is obtained based on the statistical distribution of narrative transition index corresponding to the narrative structure mutation points in historical script samples. It is determined by doubling the standard deviation of the mean, so that event node recognition is triggered only when the intensity of local narrative changes is significantly higher than the level of normal narrative fluctuations, thereby reducing the probability of misjudgment caused by minor local character switching or scene fine-tuning.

[0038] To avoid overlap and redundancy between text units, adjacent candidate event intervals are merged. When two adjacent text units simultaneously satisfy the narrative transition index being greater than or equal to the preset event recognition threshold, they are merged into a continuous event interval, and this continuous interval is used as the event node.

[0039] In step S2, after the event node identification is completed, entity information extraction processing is performed on each event node to obtain a set of structured semantic entities corresponding to the event node.

[0040] Entity information refers to the person entity, location entity, and key object entity extracted from the set of text units corresponding to the event node. The text inside the event node is parsed by the named entity recognition model, and the parsing results are standardized into a set of entity identifiers and used as entity information.

[0041] It should be noted that the named entity recognition model refers to an information recognition model used to automatically identify and extract objects with clear semantic references from script text. The objects it identifies include character entities, location entities, organization entities, prop entities, and other key narrative objects.

[0042] The number of entity recurrences is obtained by performing recurrence statistics on each entity identifier in the entity information. The number of entity recurrences is the cumulative number of times the entity identifier appears repeatedly in the event node, reflecting the continuity of the entity's participation in the narrative structure of the entire script. The larger the value, the higher the degree of penetration of the entity in the narrative chain and the stronger the structural correlation. Recurrence refers to the repeated appearance of the entity identifier in different text units in the form of explicit name, referential relationship or semantic synonym mapping.

[0043] Action semantic parsing is performed on the text units within each event node to detect the action continuation length. The action continuation length is the coverage length of consecutive text units formed around the same action predicate within the same event node. Specifically, the number of consecutive text units with the same action predicate in the event node is identified to obtain the action continuation length, which reflects the continuity of the behavioral narrative within the event node. The larger the value, the more complete the action chain within the event node and the more stable the narrative progression.

[0044] The formula for calculating the duration of an action is: ; in, This indicates the text unit position where the action predicate first appears. The text unit position that indicates the end of a series of action predicates. Indicates the duration of an action.

[0045] The recurrence consistency coefficient between two event nodes is calculated based on the number of entity recurrences. This coefficient represents the intersection strength of the two event nodes at the entity participation level. The larger the value, the higher the overlap between the two event nodes in terms of narrative participation objects. The calculation formula is as follows: ; in, Represents the reproducibility coefficient. and The set of entity identifiers representing an event node. and This represents the index value of the event node. This represents the shared set of entities between two event nodes. This represents the union of entities of two event nodes. Indicates the number of times an entity is reproduced.

[0046] A motion consistency coefficient is constructed between two event nodes. This coefficient is a normalized similarity based on the difference in action duration, reflecting the closeness of the two event nodes in terms of narrative rhythm continuity. The larger the value, the more consistent the action progression structure of the two nodes. The calculation formula is as follows: ; in, Indicates the consistency coefficient of actions. and Indicates the duration of the action of the event node. and This represents the index value of the event node, and .

[0047] The association strength of event nodes is constructed using the product normalization method: ; in, Indicates the strength of association between event nodes. Indicates the consistency coefficient of actions. This represents the reproducibility coefficient.

[0048] The correlation strength of event nodes reflects the overall coupling degree between event nodes in the entity co-occurrence structure and action continuity structure. The larger the value, the stronger the consistency of the two event nodes in the narrative structure and the tighter the connection relationship of the event link.

[0049] It should be noted that when and All When the condition is met, it indicates that the two event nodes have no correlation in terms of entity co-occurrence and action continuity; in this case, the correlation strength is 0. .

[0050] After obtaining the association strength between each event node, the event node is used as the node unit in the association density graph, and the association strength is used as the basis for connecting the nodes to generate event links.

[0051] Specifically, the association strength between any two event nodes is analyzed. When the association strength is greater than or equal to the preset link establishment threshold, it is determined that there is a stable narrative connection between the two event nodes, and an event link is established between the two event nodes. When the association strength is less than the preset link establishment threshold, it is considered that there is no effective narrative connection between the two, and no link connection is established.

[0052] By performing the above traversal analysis on all event nodes, an event link network is formed, consisting of the event node set and the event link set.

[0053] It should be noted that the preset link establishment threshold represents the minimum association strength required for event nodes to form a stable association relationship. Specifically, it is set by statistically analyzing the association strength distribution of all event node pairs in the historical script sample set. First, the association strength of each pair of event nodes is calculated to form an association strength sample set. Then, the threshold is determined based on the statistical characteristics of this sample set, such as using the mean plus standard deviation or a specified quantile as the preset link establishment threshold. Traversal analysis refers to the process of checking and calculating all event node pairs in the event node set one by one in a predetermined order. The event link network refers to the graph structure network composed of the event node set and the event link set.

[0054] After forming the event link network, the link connection status of each event node is statistically processed to calculate the event association density: ; in, This indicates the number of event nodes that have event link connections with the current event node. This indicates the maximum number of nodes in the event link network that can establish connections, excluding the current event node. This represents the total number of event nodes in the event link network. Indicates the event correlation density. This represents the index value of the event node.

[0055] Event association density characterizes the connectivity activity of the current event node in the overall event link network. The larger the value, the more narrative connections the event node has with other event nodes, and the higher its connectivity centrality in the overall narrative structure. The smaller the value, the more the event node mainly participates in the local narrative structure, and the more limited its connections with other event nodes.

[0056] After calculating the event association density for all event nodes, each event node is treated as a node object in the association density graph. The corresponding event association density is written into the event link network as a node attribute. Then, topology mapping is performed according to the link connection relationship between event nodes to construct the event association density graph.

[0057] The event association density map represents the overall association distribution between event nodes in the script to be processed, as well as the degree of connection concentration of event nodes in the narrative structure, providing basic data support for subsequent path connection depth extraction and structural hierarchy analysis.

[0058] It should be noted that topology mapping refers to the process of spatially organizing and structurally representing the link connections between event nodes in a graph structure.

[0059] In step S3, the event nodes and event links in the event association density graph are used as the analysis objects, and the path penetration depth of the event links is extracted.

[0060] Specifically, each event node in the event association density graph is taken as the starting node, and a continuous path search is performed along the connection direction in the event link network. When it is possible to reach other event nodes continuously from the starting event node through multiple intermediate event nodes, the continuous connection sequence is defined as a valid through path.

[0061] The number of event links contained in an effective connecting path is recorded as the path length. The path length represents the extent to which the continuous connection between event nodes extends. The larger the value, the more event nodes the effective connecting path covers and the more complete the narrative connection chain.

[0062] After completing the search of all valid connecting paths, count the total number of valid connecting paths in the event association density graph and calculate the path connection depth: ; in, This represents the total number of valid connecting paths in the event association density graph; Indicates the first The path length of a valid through path; Indicates the depth of the path.

[0063] The depth of path connectivity reflects the vertical extension capability of the event link network in the overall narrative structure. The larger the value, the longer the continuous connection between event nodes in the script can be formed, and the more the narrative structure tends to be organized in a layered and progressive manner. The smaller the value, the more the event link is concentrated in local connection relationships, and the more dispersed the overall narrative structure is.

[0064] Based on the path connectivity depth analysis of the structural hierarchy factor, the event association density of all event nodes in the event association density graph is statistically analyzed to obtain the mean event association density. The mean event association density represents the average connection level of the overall event link network. The larger the value, the denser the association between event nodes.

[0065] Calculate the mean of event association density: ; in, Indicates the total number of event nodes; Indicates the event correlation density. This represents the mean of the event correlation density.

[0066] Construct a structural hierarchy factor by combining the path penetration depth and the mean of event association density: ; in, Indicates the structural hierarchy factor; This represents the mean of the event correlation density; Indicates the depth of the path.

[0067] The structural hierarchy factor comprehensively characterizes the vertical connectivity and horizontal connection capabilities of the event link network. The larger the value, the longer the narrative connection path between event nodes and the higher the connection density, with the overall narrative structure exhibiting obvious hierarchical organizational characteristics. The smaller the value, the more limited the connection range and connection depth between event nodes, with the narrative structure tending towards a planar distribution.

[0068] After obtaining the structural hierarchy factor, it is used as the granularity parameter for hierarchy division. The integer part corresponding to the structural hierarchy factor is used as the single-layer coverage span, denoted as: ; in, This represents the function that takes the maximum value. Indicates the span of hierarchical division; This indicates that the structural hierarchy factor is rounded down.

[0069] Hierarchical interval division is constructed using the continuous interval method. Based on the hierarchical factor, the event nodes are indexed and hierarchically processed. The shortest link distance from the event node to the core event node in the event link network is calculated to form the event node hierarchical distance value. Specifically, the event node with the highest event association density is defined as the core event node. Taking the core event node as the hierarchical starting point, the hierarchy is expanded outward layer by layer along the connection direction in the event link network. The minimum number of event links required to get from the core event node to the target event node is defined as the event node hierarchical distance value.

[0070] It should be noted that the event node hierarchy distance value reflects the hierarchy depth of the current event node in the overall narrative structure. The smaller the value, the closer the event node is to the main narrative area and the more central it is in the event link network. The larger the value, the farther the event node is from the core narrative link and the more likely it is to serve as a branch expansion, plot supplement, or local transition.

[0071] Event nodes are categorized into hierarchical levels based on their distance values. When the distance values ​​of event nodes fall within the same interval, the corresponding event nodes are grouped into the same index level, thus forming an event node index structure consisting of multiple index levels.

[0072] After the index structure is hierarchically organized, storyboard structure information is generated based on the event nodes contained in each index level, the connection relationship between levels, and the corresponding event link information. The storyboard structure information includes event node identifiers, the index level to which they belong, and the event link pointing relationship, which represents the hierarchical organization structure and narrative development path of event nodes in the script to be processed, thus providing a structured foundation for subsequent storyboard data structure storage and optimization.

[0073] In step S4, after the storyboard structure information is generated, the node access relationship in the storyboard structure information is statistically analyzed to obtain the cross-layer reference frequency and previous frame dependency ratio of each storyboard node, and the node popularity is generated based on the above statistical results to determine whether the storage structure of the storyboard structure information needs to be adjusted.

[0074] A storyboard node refers to an event node after the index structure has been hierarchically organized. As the smallest organizational unit in the storyboard structure information, it is used to carry event node identifiers, hierarchical identifiers, and event link association information. A storyboard structure information is formed by multiple storyboard nodes organized according to hierarchical and link relationships.

[0075] In the process of analyzing the storyboard structure information, the statistical period refers to the complete data access interval that starts with the loading, parsing or reconstructing of one storyboard structure information and ends with the end of that process. Within the statistical period, all reference behaviors that occur between storyboard nodes are cumulatively counted to obtain the total number of references.

[0076] Select any storyboard node as the current statistical node. If the target storyboard node referenced by the current statistical node belongs to a different level, a cross-level reference event is determined to have occurred. The number of cross-level reference events within the statistical period is accumulated and counted. The cross-level reference frequency is obtained by dividing the number of cross-level reference events within the statistical period by the total number of references. The calculation formula is as follows: ; in, This indicates the total number of cross-level reference events within the statistical period; Indicates the total number of citations; Indicates the frequency of cross-level references.

[0077] It should be noted that in the process of analyzing the storyboard structure information, the referencing behavior refers to the operation event in which one storyboard node calls, points to, or obtains data from another storyboard node due to narrative connection, content dependence, or access needs; the target storyboard node refers to the storyboard node that is pointed to or accessed by the currently analyzed storyboard node in the process of statistical analysis of the reference relationship of the storyboard structure information, that is, the storyboard node that exists as the referenced object in a referencing behavior.

[0078] It represents the degree of dependence of storyboard nodes on cross-level data. The larger the value, the more frequent the inter-level access is in the current storyboard structure, and the weaker the level independence is.

[0079] Simultaneously, the access process of each storyboard node is analyzed within the statistical period. When any selected storyboard node needs to call the data content of the preceding storyboard node during parsing, loading, or rendering, it is determined that there is a previous frame dependency event. The number of previous frame dependency events and the total number of times all storyboard nodes are accessed within the statistical period are counted. The previous frame dependency ratio is obtained by dividing the number of previous frame dependency events by the total number of times all storyboard nodes are accessed within the statistical period. The calculation formula is as follows: ; in, Indicates the number of events dependent on the previous frame; This represents the total number of times all mirror nodes were accessed within the statistical period. This indicates the percentage of dependencies from the previous frame.

[0080] The percentage of previous frame dependency represents the degree to which the current storyboard node depends on the preceding narrative content. The larger the value, the more the current storyboard node depends on the continuous narrative structure in terms of content organization.

[0081] It should be noted that a preceding storyboard node refers to a storyboard node that precedes the current storyboard node and has a temporal or narrative sequence relationship with the current storyboard node in the sequential organization of storyboard structure information.

[0082] After obtaining the cross-layer reference frequency and the proportion of previous frame dependencies, the two metrics were analyzed separately. Normalization is performed to obtain normalized cross-layer reference frequency and normalized previous frame dependency ratio. Then, the normalized cross-layer reference frequency and normalized previous frame dependency ratio are coupled to generate node popularity. ; in, Indicates node popularity; Indicates the normalized cross-layer reference frequency; This indicates the percentage of dependencies in the previous frame after normalization.

[0083] Node popularity represents the activity level and access concentration of a storyboard node in the overall storyboard structure. The higher the value, the higher the cross-layer access demand and the stronger the dependency relationship with the preceding content of the storyboard node, and the higher the call frequency and structural influence in the storyboard structure.

[0084] After obtaining the node popularity, the node popularity is compared with a preset popularity threshold. When the node popularity is greater than or equal to the preset popularity threshold, the current storyboard node is determined to be a high-popularity node; when the node popularity is less than the preset popularity threshold, the current storyboard node is determined to be a normal node.

[0085] For high-popularity nodes, the process of adjusting the storage structure of the storyboard information is triggered, and the high-popularity nodes are migrated from the original hierarchical storage area to the high-frequency access storage area, and a direct index mapping relationship is established to reduce the number of cross-layer retrievals and the number of previous frame related queries. At the same time, the storyboard nodes that have a direct relationship with the high-popularity node are processed for adjacent storage to shorten the access path length. For ordinary nodes, the current storage structure remains unchanged.

[0086] It should be noted that the preset popularity threshold is used to distinguish between high-popularity and normal-popularity states of storyboard nodes. It is determined by statistical analysis of the node popularity distribution in historical storyboard structure information. First, the node popularity values ​​of all storyboard nodes in the historical script sample are collected to construct a popularity distribution sequence, and its mean and standard deviation or quantile distribution characteristics are calculated. Based on this, the dividing point that can distinguish between high-frequency access nodes and normal access nodes is selected as the preset popularity threshold. The original hierarchical storage area refers to the basic storage space organized according to the index structure hierarchical results in the storyboard structure information, which is used to store storyboard node data at different levels. The high-frequency access storage area refers to the storyboard nodes with high access frequency or strong dependencies that are migrated or mapped from the original hierarchical storage area to the optimized storage space during the operation of the storyboard structure, based on the node popularity judgment results.

[0087] By using a dynamic adjustment mechanism based on node popularity, the storage structure of storyboard information is adaptively optimized, enabling storyboard nodes with high frequency of access and high dependencies to obtain a better storage organization, thereby improving the retrieval efficiency and access response efficiency of subsequent storyboard data.

[0088] Example 2, as Figure 2 As shown, an adaptive video storyboard data structure construction system includes a script parsing module, a link construction module, a storyboard layering module, and a structure optimization module. The functions of each module are as follows: After receiving the script to be processed, the script parsing module performs a text scan on the script to be processed, collects the character switching frequency and scene transition span of the script to be processed, calculates the narrative leap index, identifies the event nodes of the script to be processed based on the narrative leap index, and passes the event nodes to the link construction module; The link construction module obtains the entity information of the event node and counts the number of entity recurrences, detects the duration of the event node's action, generates the association strength of the event node based on the number of entity recurrences, generates the event link based on the association strength and constructs the event association density map, and passes the event association density map into the scene layering module. The storyboard layering module extracts the path penetration depth of the event chain based on the event association density map, analyzes the structural hierarchy factor based on the path penetration depth, indexes the event nodes using the structural hierarchy factor, and outputs the storyboard structure information. The storyboard structure information is then passed to the structure optimization module. The structure optimization module counts the cross-layer reference frequency and the previous frame dependency ratio of the storyboard structure information, generates node popularity by combining the cross-layer reference frequency and the previous frame dependency ratio, and determines whether to adjust the storage structure of the storyboard structure information based on the node popularity.

[0089] Finally, it should be noted that in this paper, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0090] Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0091] In this document, the singular forms “a,” “an,” and “the” may also include the plural forms unless the context clearly indicates otherwise. It should also be understood that terms such as “comprising / including” or “having” specify the presence of the stated features, integrals, steps, operations, components, parts, or combinations thereof, but do not preclude the possibility of the presence or addition of one or more other features, integrals, steps, operations, components, parts, or combinations thereof. Meanwhile, the term “and / or” as used in this specification includes any and all combinations of the associated listed items.

[0092] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0093] The above description of the disclosed embodiments will enable those skilled in the art to make or use various modifications to these embodiments. It will be readily apparent to those skilled in the art that the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An adaptive construction method for video storyboard data structure, characterized in that: Includes the following steps: Step S1: After receiving the script to be processed, perform text scanning on the script to be processed, collect the character switching frequency and scene transition span of the script to be processed, calculate the narrative leap index, and identify the event nodes of the script to be processed based on the narrative leap index. Step S2: Obtain the entity information of the event node and count the number of entity recurrences, detect the action duration of the event node, generate the association strength of the event node based on the number of entity recurrences, generate the event chain based on the association strength and construct the event association density map; Step S3: Extract the path penetration depth of the event link based on the event association density map, analyze the structural hierarchy factor based on the path penetration depth, use the structural hierarchy factor to index and classify the event nodes and output the storyboard structure information. Step S4: Calculate the cross-layer reference frequency and the previous frame dependency ratio of the storyboard structure information, generate node popularity by combining the cross-layer reference frequency and the previous frame dependency ratio, and determine whether to adjust the storage structure of the storyboard structure information based on the node popularity.

2. The adaptive construction method for video storyboard data structure according to claim 1, characterized in that: In step S1, after receiving the script to be processed, text scanning processing is performed on the script to be processed. Text scanning refers to scanning the script text sentence by sentence based on preset text units, and parsing the character identifier and scene identifier for each text unit. Within each text unit, the dominant role identifier corresponding to each text unit is continuously compared. When the dominant role identifier of each text unit changes, it is determined as a role switching event. The number of role switching events is counted to obtain the number of role switching events. The role switching frequency is obtained by normalizing the number of role switching times with the number of sentences in the text unit.

3. The adaptive construction method for video scene data structure according to claim 2, characterized in that: In step S1, scene identifier parsing is performed on the text content in each text unit. When the scene semantic identifier of the text unit changes, a scene transition event is determined to have occurred. The number of scene transitions within the text unit is counted, and the scene transition span is calculated. The average scene transition span is obtained by summarizing and averaging all scene transition spans within the text unit, and a narrative leap index is constructed by combining the character switching frequency. When the narrative transition index is greater than or equal to the preset event recognition threshold, the text interval corresponding to the text unit is marked as a candidate event interval; When the narrative transition index is less than the preset event recognition threshold, it is determined that no significant narrative structure change occurs in the text interval corresponding to the text unit; Adjacent candidate event intervals are merged to form continuous event intervals as event nodes.

4. The adaptive construction method for video storyboard data structure according to claim 1, characterized in that: In step S2, entity information extraction processing is performed on each event node. Entity information refers to the person entity, location entity, and key object entity extracted from the text unit set corresponding to the event node. The entity recurrence count is the cumulative number of times the entity identifier appears repeatedly in the event node; The action continuation length is the length of continuous text units covered by the same action predicate within the same event node; Calculate the reproducibility coefficient between two event nodes based on the number of entity reproducibility cycles; Construct an action consistency coefficient between two event nodes. The action consistency coefficient is a normalized similarity based on the difference in action duration. The association strength of event nodes is constructed using the product normalization method.

5. The adaptive construction method for video storyboard data structure according to claim 4, characterized in that: In step S2, event nodes are used as node units in the association density graph, and association strength is used as the basis for connecting nodes to generate event links. By performing a traversal analysis on all event nodes, an event link network is formed, consisting of the event node set and the event link set. The link connection status of each event node is statistically processed to calculate the event association density. The corresponding event association density is written into the event link network as a node attribute, and a topology mapping is performed according to the link connection relationship between event nodes to construct an event association density graph.

6. The adaptive construction method for video scene data structure according to claim 1, characterized in that: In step S3, each event node in the event association density graph is used as the starting node, and a continuous path search is performed along the connection direction in the event link network to obtain an effective through path. The number of event links contained in the effective through path is counted and recorded as the path length. The total number of valid connecting paths in the statistical event association density graph is calculated, and the path connection depth is calculated in combination with the path length. The event association density of all event nodes in the event association density graph is statistically analyzed to obtain the mean event association density. The structural hierarchy factor is constructed by combining the path penetration depth with the mean event association density. Using the integer part corresponding to the structural hierarchy factor as the single-layer coverage span, the hierarchy division interval is constructed in a continuous interval manner.

7. The adaptive construction method for video scene data structure according to claim 6, characterized in that: In step S3, the event nodes are indexed and hierarchically processed according to the hierarchical factor, and the shortest link distance from the event node to the core event node of the event link network is calculated to form the event node hierarchical distance value. Define the event node with the highest event correlation density as the core event node; Event nodes are categorized into hierarchical levels based on their distance values, forming an event node index structure consisting of multiple index levels. The storyboard structure information is generated based on the event nodes contained in each index level, the connection relationship between levels, and the corresponding event link information. The storyboard structure information includes the event node identifier, the index level to which it belongs, and the event link pointing relationship.

8. The adaptive construction method for video scene data structure according to claim 7, characterized in that: In step S4, after the storyboard structure information is generated, the node access relationship in the storyboard structure information is statistically analyzed to obtain the cross-layer reference frequency and the previous frame dependency ratio of each storyboard node. A storyboard node refers to an event node that has completed the hierarchical arrangement of the index structure. Within the statistical period, all referencing behaviors occurring between storyboard nodes are cumulatively counted to obtain the total number of references; Select any storyboard node as the current statistical node. If the target storyboard node referenced by the storyboard node is at a different level than the current statistical node, a cross-level reference event is determined to have occurred. The cross-level reference frequency is obtained by dividing the number of cross-level reference events within the statistical period by the total number of references.

9. The adaptive construction method for video scene data structure according to claim 8, characterized in that: In step S4, when any selected storyboard node needs to call the data content of the preceding storyboard node during parsing, loading or rendering, it is determined that there is a previous frame dependency event, and the number of previous frame dependency events and the total number of times all storyboard nodes are accessed within the statistical period are counted. Divide the number of previous frame dependent events by the total number of times all split-view nodes are accessed within the statistical period to obtain the previous frame dependency percentage. Coupling normalized cross-layer reference frequency and normalized pre-frame dependency ratio generated node popularity: When the node's popularity is greater than or equal to a preset popularity threshold, the current storyboard node is determined to be a high-popularity node. For high-traffic nodes, the process of adjusting the storage structure of the storyboard information is triggered, and the high-traffic nodes are migrated from the original hierarchical storage area to the high-frequency access storage area. When the node's popularity is less than the preset popularity threshold, the current storyboard node is determined to be a normal node, and the current storage structure remains unchanged.

10. A video storyboard data structure adaptive construction system, used to implement the video storyboard data structure adaptive construction method according to any one of claims 1-9, characterized in that: It includes a script analysis module, a link construction module, a storyboard layering module, and a structure optimization module. The functions of each module are as follows: After receiving the script to be processed, the script parsing module performs a text scan on the script to be processed, collects the character switching frequency and scene transition span of the script to be processed, calculates the narrative leap index, identifies the event nodes of the script to be processed based on the narrative leap index, and passes the event nodes to the link construction module; The link construction module obtains the entity information of the event node and counts the number of entity recurrences, detects the duration of the event node's action, generates the association strength of the event node based on the number of entity recurrences, generates the event link based on the association strength and constructs the event association density map, and passes the event association density map into the scene layering module. The storyboard layering module extracts the path penetration depth of the event link based on the event association density map, analyzes the structural hierarchy factor based on the path penetration depth, indexes the event nodes using the structural hierarchy factor, and outputs the storyboard structure information. The storyboard structure information is then passed to the structure optimization module. The structure optimization module counts the cross-layer reference frequency and the previous frame dependency ratio of the storyboard structure information, generates node popularity by combining the cross-layer reference frequency and the previous frame dependency ratio, and determines whether to adjust the storage structure of the storyboard structure information based on the node popularity.

Images