Micro-short script intelligent optimization system based on semantic analysis

Abstract

This invention relates to the field of intelligent dialogue optimization technology. Specifically, it relates to a micro-drama dialogue intelligent optimization system based on semantic analysis. It includes a data acquisition and semantic matching module, a plot segment optimization and list update module, a dialogue reorganization and optimization module, and a full-drama dialogue coherence verification module. The data acquisition and semantic matching module collects dialogue data, scene data, and character data from the micro-drama, and performs plot semantic analysis based on the scene and character data. By setting up the data acquisition and semantic matching module, comprehensive collection of dialogue data, scene data, and character data from the micro-drama is achieved. Simultaneously, relying on the semantic matching algorithm, the overall plot semantics are accurately bound to each plot segment, giving each plot segment a clear and exclusive semantic anchor point. This effectively solves the problem of dialogue being disconnected from the plot and character design in existing technologies, ensuring that subsequent dialogue optimization always revolves around the core plot and character characteristics.

Description

Technical Field

[0001] This invention relates to the field of intelligent dialogue optimization technology, and more specifically, to a micro-drama dialogue intelligent optimization system based on semantic analysis. Background Technology

[0002] In the current era of rapid development in the micro-drama industry, dialogue, as the core carrier of content expression in micro-dramas, directly affects the presentation of the plot, the shaping of character settings, and the viewing experience of the audience. Therefore, related dialogue optimization technologies have emerged.

[0003] Existing dialogue optimization technologies allow creators to perform basic text corrections using tools, reducing the workload of manual line-by-line proofreading. However, these technologies cannot achieve a deep integration of dialogue with the semantics of the micro-drama's plot, character design, and scene setting. Optimized dialogue often only meets the requirement of textual fluency, but it is prone to problems such as disconnect between character design and dialogue and scene atmosphere, directly affecting the accuracy of character development and plot expression in micro-dramas. Secondly, existing technologies lack the ability to process plot segments in a refined manner. They cannot reasonably merge or split segments based on the semantic relevance of the plot, nor can they customize optimization strategies for the characteristics of different plot segments. As a result, the optimized dialogue is difficult to adapt to the fast-paced and high-conflict rhythm requirements of micro-dramas. To reduce this situation, a micro-drama dialogue intelligent optimization system based on semantic analysis is proposed. Summary of the Invention

[0004] The purpose of this invention is to provide a semantic analysis-based intelligent optimization system for dialogue in short dramas, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, a micro-drama dialogue intelligent optimization system based on semantic analysis is provided, including a data acquisition and semantic matching module, a plot segment optimization and list update module, a dialogue reorganization and optimization module, and a full-drama dialogue coherence verification module.

[0006] The data acquisition and semantic matching module is used to collect the dialogue data, scene data and character data of the micro-drama, perform plot semantic analysis based on the scene data and character data, divide the plot into segments and form a plot segment list based on the dialogue data, match the plot semantics with each plot segment, and determine the plot semantics corresponding to each plot segment.

[0007] The plot segment optimization and list update module is used to set the difference threshold corresponding to different scene types, calculate the difference rate of plot semantics between each plot segment, select the scene type of the micro-drama based on all plot semantics, merge the difference rate of adjacent plot segments with the difference threshold corresponding to the scene type of the micro-drama for judgment, and merge adjacent plot segments when the difference rate is less than the difference threshold and update the plot segment list.

[0008] The dialogue reorganization and optimization module is used to establish a reference dialogue database based on scene type, select plot segments in sequence, perform multi-dimensional correlation analysis on dialogue data with scene data, character data, and plot semantics, determine a baseline threshold based on the multi-dimensional correlation analysis results, filter dialogue data in the reference dialogue database based on the baseline threshold, reorganize the filtered dialogue data with the original dialogue data corresponding to the plot segment, obtain a reorganized dialogue data list, perform multi-dimensional correlation analysis on the reorganized dialogue data list simultaneously, select the optimal reorganized dialogue data from the analysis results to complete the dialogue optimization of the plot segment, and then complete the full optimization of the plot segment list in sequence.

[0009] The full-drama dialogue coherence verification module is used to set corresponding relevant thresholds based on the difference rate of the plot semantics between each plot segment, perform correlation analysis on the recombined dialogue data of each plot segment after optimization, obtain the correlation of the recombined dialogue data between each plot segment, compare the correlation with the relevant threshold, and when the correlation is lower than the relevant threshold, reselect from the recombined dialogue data list until the correlation meets the relevant threshold.

[0010] As a further improvement to this technical solution, the data acquisition and semantic matching module connects to the micro-drama management terminal to collect micro-drama data that requires dialogue optimization. The micro-drama data includes dialogue data, scene data, and character data.

[0011] As a further improvement to this technical solution, the data acquisition and semantic matching module performs plot semantic analysis on the micro-drama based on scene data and character data to obtain the overall plot semantics of the micro-drama. At the same time, it divides the micro-drama into plot segments based on the dialogue data, divides the micro-drama into multiple plot segments, and summarizes the plot segments to obtain a plot segment list.

[0012] The overall plot semantics are segmented and matched with the corresponding plot fragments to obtain the plot semantics corresponding to each plot fragment.

[0013] As a further improvement to this technical solution, in the plot segment optimization and list update module, a difference threshold corresponding to different scene types is set, and then the semantic elements of the plot semantics corresponding to each plot segment are extracted. The semantic similarity algorithm is used to calculate the semantic element fit between any two plot segments, and then the plot semantic difference rate is obtained by reverse deduction through the fit rate, thereby obtaining the plot semantic difference rate between each plot segment.

[0014] Based on the plot semantics of each plot segment, scene type analysis is performed to obtain the proportion of different scene types corresponding to each plot semantics. Then, the proportion of different scene types corresponding to each plot semantics is summarized, and the scene type with the highest proportion is selected as the scene type corresponding to the micro-drama.

[0015] As a further improvement to this technical solution, according to the scene type corresponding to the micro-drama, a corresponding difference threshold is matched for the micro-drama, and then the matched difference threshold is compared with the difference rate between adjacent plot segments.

[0016] If the difference rate of plot semantics between adjacent plot segments is less than the difference threshold, the two plot segments are merged to update the plot segment list.

[0017] Conversely, if the difference rate of plot semantics between adjacent plot segments is greater than the difference threshold, the two plot segments will not be merged.

[0018] The merging of plot segments includes merging scene data, character data, dialogue data, and plot semantics.

[0019] As a further improvement to this technical solution, the dialogue reorganization and optimization module collects dialogue data of the same type of micro-drama that matches the scene type of the micro-drama, establishes a reference dialogue database based on the collected dialogue data, and uses the dialogue data in the reference dialogue database as reference dialogue data.

[0020] The dialogue data of the plot segments are correlated with scene data, character data, and plot semantics to obtain scene correlation, character correlation, and plot correlation. The three together constitute a multidimensional correlation, and the value of this multidimensional correlation is used as the benchmark threshold.

[0021] As a further improvement to this technical solution, based on the benchmark threshold, the reference dialogue data in the reference dialogue database is correlated with scene data, character data, and plot semantics to obtain the multidimensional correlation degree corresponding to each reference dialogue data, and then the multidimensional correlation degree corresponding to each reference dialogue data is compared with the benchmark threshold.

[0022] If the multidimensional correlation is greater than the baseline threshold, the reference dialogue data is retained.

[0023] Conversely, if the multidimensional correlation is less than the baseline threshold, the reference line data will not be retained and will be deleted from the reference line database.

[0024] The retained reference dialogue data is randomly semantically fused and recombined with the original dialogue data of the corresponding plot segment. The recombined dialogue data is summarized and a recombined dialogue data list is generated. At the same time, the multidimensional correlation degree of each recombined dialogue data in the recombined dialogue data list is calculated to obtain the multidimensional correlation degree corresponding to the recombined dialogue data. Then, the recombined dialogue data with the highest multidimensional correlation degree is selected to optimize the plot segment.

[0025] The optimization process prioritizes the first plot segment of a short drama as the first target for optimization. After optimizing this plot segment, subsequent plot segments are selected in the natural order of the plot segment list, and the optimization is repeated until all plot segments in the list are optimized.

[0026] As a further improvement to this technical solution, in the full-drama dialogue coherence verification module, a relevant threshold is set in the plot segment list based on the difference rate of plot semantics between each plot segment;

[0027] Among them, the lower the difference rate, the lower the correlation threshold;

[0028] The higher the difference rate, the higher the correlation threshold.

[0029] As a further improvement to this technical solution, in the full-drama dialogue coherence verification module, the optimized recombined dialogue data of each plot segment is subjected to correlation analysis, the semantic, expression style and sentence structure features of the optimized recombined dialogue data of each plot segment are extracted, the degree of fit of the recombined dialogue data of adjacent plot segments in semantic, expression style and sentence structure features is calculated, and this is used as the correlation between the recombined dialogue data of each plot segment. Then the correlation is compared with the relevant threshold.

[0030] If the relevance is lower than the relevance threshold, the recombined dialogue data list will be re-selected until the relevance meets the relevance threshold.

[0031] Conversely, if the correlation is higher than the correlation threshold, monitoring continues.

[0032] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0033] 1. This intelligent optimization system for micro-drama dialogue based on semantic analysis achieves comprehensive collection of dialogue, scene, and character data by setting up data acquisition and semantic matching modules. At the same time, it relies on semantic matching algorithms to accurately bind the overall plot semantics with each plot segment, giving each plot segment a clear and exclusive semantic anchor point. This effectively solves the problem of dialogue being disconnected from the plot and character settings in existing technologies, ensuring that subsequent dialogue optimization always revolves around the core of the plot and the characteristics of the characters, and fundamentally avoiding situations where character settings are fragmented and dialogue is incongruous with the scene.

[0034] 2. This intelligent dialogue optimization system for micro-dramas based on semantic analysis intelligently organizes the initial list of plot segments by calculating the semantic difference rate of the plot and combining it with the difference threshold corresponding to the scene type. This module can automatically merge semantically similar segments, eliminate redundant and fragmented content, and form processing units that better meet the needs of dialogue reorganization. Simultaneously, during the merging process, it integrates four types of data: scene, character, dialogue, and semantics. This not only reduces the workload of subsequent dialogue reorganization and optimization, improving the overall optimization efficiency of the system, but also further strengthens the semantic connections between plot segments, laying the foundation for the coherence of the entire drama's dialogue.

[0035] 3. This intelligent dialogue optimization system for micro-dramas based on semantic analysis constructs a vertical database of reference dialogues based on the core scene types of micro-dramas, ensuring a high degree of compatibility between the reference corpus and the target micro-drama's category and scene. Simultaneously, this module innovatively introduces a multi-dimensional correlation calculation model, incorporating scene correlation, character correlation, and plot semantic correlation into a quantitative analysis system. Reference dialogue data is then filtered using these as benchmark thresholds, and optimized dialogues are generated through random semantic fusion and recombination. This overcomes the limitations of existing technologies that rely on single-dimensional optimization, allowing the recombined dialogues to retain the core information of the original plot while also fitting the scene atmosphere and character design in terms of expression. This significantly improves the quality and expressiveness of the micro-drama dialogues, while the sequential iterative optimization logic adapts to the fast-paced production rhythm of micro-dramas. Attached Figure Description

[0036] Figure 1 This is a flowchart illustrating the intelligent optimization system for micro-drama dialogue based on semantic analysis of the present invention.

[0037] Figure 2 This is a flowchart illustrating the data acquisition and semantic matching module of the present invention;

[0038] Figure 3 This is a flowchart illustrating the plot segment optimization and list update module of the present invention;

[0039] Figure 4 This is a flowchart illustrating the dialogue reorganization and optimization module of the present invention;

[0040] Figure 5 This is a flowchart illustrating the process of the dialogue coherence verification module for the entire drama in this invention. Detailed Implementation

[0041] 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.

[0042] Please see Figures 1-5 As shown, the purpose of this embodiment is to provide a micro-drama dialogue intelligent optimization system based on semantic analysis, including a data acquisition and semantic matching module, a plot segment optimization and list update module, a dialogue reorganization and optimization module, and a full drama dialogue coherence verification module.

[0043] The data acquisition and semantic matching module is used to collect dialogue data, scene data and character data of micro-dramas. Based on the scene data and character data, the module performs plot semantic analysis. At the same time, it combines the dialogue data to divide the plot into segments and form a list of plot segments. The module matches the plot semantics with each plot segment to determine the plot semantics corresponding to each plot segment.

[0044] It undertakes the foundational work for the entire process of raw data acquisition, semantic analysis, fragment segmentation, and semantic matching, providing standardized and semantic data support for all subsequent optimization modules, which is a prerequisite for the operation of the entire system;

[0045] In the data acquisition and semantic matching module, the micro-drama management terminal is connected to collect micro-drama data that needs dialogue optimization. The micro-drama data includes dialogue data, scene data, and character data.

[0046] In the data acquisition and semantic matching module, the short drama is analyzed for plot semantics based on scene data and character data to obtain the overall plot semantics. Simultaneously, the short drama is divided into plot segments based on dialogue data, resulting in a list of plot segments. The steps are as follows:

[0047] Based on the collected scene data and character data, the overall plot semantics of the micro-drama are mined using natural language semantic analysis methods to obtain the overall plot semantics of the micro-drama. Then, based on the collected dialogue data, the plot segments of the micro-drama are divided according to the semantic logic of the dialogue and the dialogue intervals to obtain several independent plot segments. All plot segments are summarized and organized to generate an initial plot segment list.

[0048] The overall plot semantics are segmented and matched with the corresponding plot fragments to obtain the plot semantics corresponding to each plot fragment.

[0049] Based on the semantic matching algorithm, the overall plot semantics of the micro-drama are segmented and matched with each plot segment in the initial plot segment list. The matching degree between each plot segment and the overall plot semantics is calculated, and the exclusive plot semantics corresponding to each plot segment are determined based on the matching degree results.

[0050] The plot segment optimization and list update module is used to set the difference threshold corresponding to different scene types, calculate the difference rate of plot semantics between each plot segment, select the scene type of the short drama based on all plot semantics, and merge the difference rate of adjacent plot segments with the difference threshold corresponding to the scene type of the short drama for judgment. When the difference rate is less than the difference threshold, the adjacent plot segments are merged and the plot segment list is updated.

[0051] Based on the semantic fragment data output by the data acquisition module, the initial list of plot fragments is optimized and reorganized through the difference rate calculation, scene type determination and fragment merging logic to generate processing units that better meet the needs of dialogue reorganization, while realizing the integration and unification of fragment data and semantics.

[0052] In the plot segment optimization and list update module, a difference threshold is set for different scene types. Then, the semantic elements corresponding to the plot semantics of each plot segment are extracted. The semantic similarity algorithm is used to calculate the semantic element fit between any two plot segments. Then, the difference rate of plot semantics is derived by reverse derivation from the fit rate, thereby obtaining the difference rate of plot semantics between each plot segment. The steps are as follows:

[0053] Preset the semantic difference thresholds for different scene types in the micro-drama, establish a correspondence table between scene types and difference thresholds, complete the basic threshold configuration, extract the semantic elements of the exclusive plot semantics for each plot segment, vectorize the semantic elements, and form a standardized semantic feature vector.

[0054] The semantic similarity optimization algorithm is used to calculate the semantic element fit between any two plot segments. The semantic difference rate between the two segments is then derived by reverse reasoning from the fit rate. Finally, the semantic difference rate between all plot segments in the entire series is calculated. The formula is as follows:

[0055] ;

[0056] Where D(A,B) is the semantic difference rate between any two plot segments A and B, with a value range of [0,1]. The closer the value is to 1, the greater the semantic difference between the two. Sim(A,B) is the semantic element fit degree between plot segments A and B.

[0057] Based on the plot semantics of each plot segment, scene type analysis is performed to obtain the proportion of different scene types corresponding to each plot semantics. Then, the proportion of different scene types corresponding to each plot semantics is summarized, and the scene type with the highest proportion is selected as the scene type corresponding to the micro-drama.

[0058] Based on the unique plot semantics of each plot segment, a scene type attribution analysis is conducted to determine the scene type attribution corresponding to a single plot semantic. The number of each scene type corresponding to all plot semantics in the entire series is counted, and the proportion of each scene type in the overall plot semantics is calculated. Then, the proportion results of all scene types are summarized, and the scene type with the highest proportion value is selected as the scene type of the micro-drama. The formula is as follows:

[0059] ;

[0060] Where P(T) is the proportion of a certain scene type T in the semantics of the whole drama, N(T) is the number of plot segments belonging to scene type T in the whole drama, Ti is the i-th scene type in the whole drama, and n is the total number of scene types involved in the whole drama.

[0061] Based on the scene type corresponding to the micro-drama, a corresponding difference threshold is matched for the micro-drama, and then the matched difference threshold is compared with the difference rate between adjacent plot segments.

[0062] If the difference rate of plot semantics between adjacent plot segments is less than the difference threshold, the two plot segments are merged to update the plot segment list.

[0063] Conversely, if the difference rate of plot semantics between adjacent plot segments is greater than the difference threshold, the two plot segments will not be merged.

[0064] The merging of plot segments includes merging scene data, character data, dialogue data, and plot semantics.

[0065] The dialogue reorganization and optimization module is used to establish a reference dialogue database based on scene type, select plot fragments in sequence, perform multi-dimensional correlation analysis on dialogue data with scene data, character data, and plot semantics, determine a baseline threshold based on the multi-dimensional correlation analysis results, filter dialogue data in the reference dialogue database based on the baseline threshold, reorganize the filtered dialogue data with the original dialogue data corresponding to the plot fragments, obtain a reorganized dialogue data list, perform multi-dimensional correlation analysis on the reorganized dialogue data list simultaneously, select the optimal reorganized dialogue data from the analysis results to complete the dialogue optimization of the plot fragments, and then complete the full optimization of the plot fragment list in sequence.

[0066] As the system's optimization execution module, it takes the optimized plot segments as the processing object and achieves precise optimization of the dialogue of a single segment through reference library construction, benchmark threshold setting, dialogue screening and reorganization, and optimal selection logic. Then, it completes the optimization of the entire list of segments through sequential iteration, directly solving the dialogue quality problem.

[0067] In the dialogue reorganization and optimization module, dialogue data from similar micro-dramas matching the scene type are collected. A reference dialogue database is established based on the collected dialogue data, and the dialogue data in the reference dialogue database is used as reference dialogue data. The steps are as follows:

[0068] Based on the identified short drama scene types, we collect dialogue data from short dramas of the same scene type and category to ensure that the data highly matches the scene characteristics and category style of the target short drama. At the same time, we standardize and clean the collected dialogue data to remove invalid, redundant, and semantically incomplete data. Based on the cleaned and effective data, we construct a reference dialogue database and define all effective dialogue data in the database as reference dialogue data.

[0069] The dialogue data of the plot segments are correlated with scene data, character data, and plot semantics to obtain scene correlation, character correlation, and plot correlation. The three together constitute a multidimensional correlation, and the value of this multidimensional correlation is used as the benchmark threshold.

[0070] Semantic features of the dialogue data of the plot segment to be optimized are extracted. At the same time, the features of the scene data, character data and plot semantics bound to the plot segment are extracted. All features are vectorized to generate standardized feature vectors. Then, the semantic similarity optimization algorithm is used to calculate the matching degree of the feature vectors of the plot segment dialogue data with the scene data, character data and plot semantics respectively, and obtain three single-dimensional correlation indicators of scene correlation, character correlation and plot correlation.

[0071] Configure customizable weight coefficients for the three single-dimensional correlations. The weight coefficients can be adjusted according to the creation needs and category characteristics of the micro-drama to adapt to the priority requirements of different optimization scenarios.

[0072] Based on the configured weight coefficients, the three single-dimensional correlations are weighted and summed to obtain the multi-dimensional correlation. The specific value of the calculated multi-dimensional correlation is set as the benchmark threshold for this reference dialogue data screening and dialogue reorganization, serving as the quantitative basis for subsequent reference dialogue data screening.

[0073] Based on the benchmark threshold, the reference lines data in the reference lines database are correlated with scene data, character data, and plot semantics to obtain the multidimensional correlation degree corresponding to each reference lines data. Then, the multidimensional correlation degree corresponding to each reference lines data is compared with the benchmark threshold.

[0074] If the multidimensional correlation is greater than the baseline threshold, the reference dialogue data is retained.

[0075] Conversely, if the multidimensional correlation is less than the baseline threshold, the reference line data will not be retained and will be deleted from the reference line database.

[0076] The retained reference dialogue data is randomly semantically fused and recombined with the original dialogue data of the corresponding plot segment. The recombined dialogue data is summarized and a recombined dialogue data list is generated. At the same time, the multidimensional correlation degree of each recombined dialogue data in the recombined dialogue data list is calculated to obtain the multidimensional correlation degree corresponding to the recombined dialogue data. Then, the recombined dialogue data with the highest multidimensional correlation degree is selected to optimize the plot segment.

[0077] Extract the semantics of the reference lines data that have been filtered and retained, and perform random semantic fusion and recombination with the semantics of the original lines data of the corresponding plot segments. During the recombination process, the original plot information of the lines is retained, and only the expression is optimized. All recombination results are summarized to generate a list of recombined lines data.

[0078] The optimization process prioritizes the first plot segment of a short drama as the first target for optimization. After optimizing this plot segment, subsequent plot segments are selected in the natural order of the plot segment list, and the optimization is repeated until all plot segments in the list are optimized.

[0079] The dialogue coherence verification module is used to set corresponding relevant thresholds based on the difference rate of the plot semantics between each plot segment. It performs correlation analysis on the recombined dialogue data of each plot segment after optimization, obtains the correlation between the recombined dialogue data of each plot segment, and compares the correlation with the relevant threshold. When the correlation is lower than the relevant threshold, it reselects from the recombined dialogue data list until the correlation meets the relevant threshold.

[0080] Focusing on the semantic coherence between segments, the logic of setting relevant thresholds, analyzing relevance, comparing thresholds, and making secondary adjustments is used to avoid the problem of semantic fragmentation in the whole series after optimizing a single segment, ensuring the integrity of the dialogue in the whole series after optimization, and forming a closed loop of optimization-verification-adjustment.

[0081] In the whole drama dialogue coherence verification module, relevant thresholds are set in the plot segment list based on the difference rate of plot semantics between each plot segment;

[0082] Among them, the lower the difference rate, the lower the correlation threshold;

[0083] The higher the difference rate, the higher the relevant threshold. The steps are as follows:

[0084] The semantic difference rate between adjacent plot segments, calculated during the plot segment optimization process, is retrieved. Based on the positive correlation between the difference rate and the relevant threshold, a specific relevant threshold is set for each pair of adjacent segments. The lower the difference rate, the lower the corresponding relevant threshold; the higher the difference rate, the higher the corresponding relevant threshold. This forms a mapping table between adjacent segment pairs and specific relevant thresholds, as shown in the following formula:

[0085] ;

[0086] Where Thr is the specific relevance threshold corresponding to a pair of adjacent segments, k is the proportional coefficient, D is the semantic difference rate of the plot of the pair of adjacent segments, and b is the correction coefficient. Both can be calibrated according to the micro-drama industry data to ensure that the threshold setting is reasonable.

[0087] In the whole drama dialogue coherence verification module, the optimized and recombined dialogue data of each plot segment are analyzed, and the semantic, expression style and sentence structure features of the optimized and recombined dialogue data of each plot segment are extracted. The degree of fit of the recombined dialogue data of adjacent plot segments in semantic, expression style and sentence structure features is calculated, which is used as the correlation between the recombined dialogue data of each plot segment. Then the correlation is compared with the relevant threshold.

[0088] Features of all plot segments and optimized recombined dialogue data are extracted, specifically covering three dimensions: semantic features, expression style features, and sentence structure features. The three types of features are vectorized to generate standardized feature vectors.

[0089] A semantic similarity algorithm is used to calculate the feature vector fit of each pair of adjacent plot segment reconstructed dialogue data in three dimensions: semantics, expression style, and sentence structure. This yields three single-dimensional fit indices. Customizable weighting coefficients are assigned to these three dimensions (adjustable according to the genre of the short drama; for example, suspense dramas have increased weighting for the semantic dimension, while romantic dramas have increased weighting for the expression style dimension). The comprehensive relevance of each pair of adjacent segment reconstructed dialogue is calculated through weighted summation, serving as a quantitative indicator of coherence. The formula is as follows:

[0090] ;

[0091] Among them, C Y For single-dimensional fit, it can represent semantic fit, stylistic fit, and sentence structure fit, respectively. V iY Let V(i+1) be the feature vector corresponding to the dimension of the reconstructed dialogue for the i-th plot segment. Y The feature vector corresponding to the dimension of the recombined dialogue for the (i+1)th plot segment.

[0092] Retrieve the specific relevance thresholds for the corresponding adjacent segment pairs from the mapping table, and compare the calculated overall relevance with the specific relevance thresholds one by one;

[0093] If the relevance is lower than the relevance threshold, the recombined dialogue data list will be re-selected until the relevance meets the relevance threshold.

[0094] Conversely, if the correlation is higher than the correlation threshold, monitoring continues.

[0095] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely preferred examples and are not intended to limit the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A micro-series script intelligent optimization system based on semantic analysis, characterized in that: This includes modules for data collection and semantic matching, plot segment optimization and list updating, dialogue reorganization and optimization, and overall dialogue coherence verification. The data acquisition and semantic matching module is used to collect the dialogue data, scene data and character data of the micro-drama, perform plot semantic analysis based on the scene data and character data, divide the plot into segments and form a plot segment list based on the dialogue data, match the plot semantics with each plot segment, and determine the plot semantics corresponding to each plot segment. The plot segment optimization and list update module is used to set the difference threshold corresponding to different scene types, calculate the difference rate of plot semantics between each plot segment, select the scene type of the micro-drama based on all plot semantics, merge the difference rate of adjacent plot segments with the difference threshold corresponding to the scene type of the micro-drama for judgment, and merge adjacent plot segments when the difference rate is less than the difference threshold and update the plot segment list. The dialogue reorganization and optimization module is used to establish a reference dialogue database based on scene type, select plot segments in sequence, perform multi-dimensional correlation analysis on dialogue data with scene data, character data, and plot semantics, determine a baseline threshold based on the multi-dimensional correlation analysis results, filter dialogue data in the reference dialogue database based on the baseline threshold, reorganize the filtered dialogue data with the original dialogue data corresponding to the plot segment, obtain a reorganized dialogue data list, perform multi-dimensional correlation analysis on the reorganized dialogue data list simultaneously, select the optimal reorganized dialogue data from the analysis results to complete the dialogue optimization of the plot segment, and then complete the full optimization of the plot segment list in sequence. The full-drama dialogue coherence verification module is used to set corresponding relevant thresholds based on the difference rate of the plot semantics between each plot segment, perform correlation analysis on the recombined dialogue data of each plot segment after optimization, obtain the correlation of the recombined dialogue data between each plot segment, compare the correlation with the relevant threshold, and when the correlation is lower than the relevant threshold, reselect from the recombined dialogue data list until the correlation meets the relevant threshold.

2. The system for semantic analysis based micro-series script intelligent optimization according to claim 1, characterized in that: In the data acquisition and semantic matching module, the micro-drama management terminal is connected to collect micro-drama data that needs dialogue optimization. The micro-drama data includes dialogue data, scene data, and character data. 3.The micro-series script intelligent optimization system based on semantic analysis of claim 1, wherein: In the data acquisition and semantic matching module, the short drama is analyzed for plot semantics based on scene data and character data to obtain the overall plot semantics of the short drama. At the same time, the short drama is divided into plot segments based on the dialogue data, and the short drama is divided into multiple plot segments. The plot segments are then summarized to obtain a plot segment list. The overall plot semantics are segmented and matched with the corresponding plot fragments to obtain the plot semantics corresponding to each plot fragment.

4. The system, as recited in claim 1, wherein the system is characterized by: In the plot segment optimization and list update module, a difference threshold corresponding to different scene types is set, then the semantic elements of the plot semantics corresponding to each plot segment are extracted, the semantic similarity algorithm is used to calculate the semantic element fit between any two plot segments, and then the plot semantic difference rate is obtained by reverse deduction through the fit rate, thereby obtaining the plot semantic difference rate between each plot segment. Based on the plot semantics of each plot segment, scene type analysis is performed to obtain the proportion of different scene types corresponding to each plot semantics. Then, the proportion of different scene types corresponding to each plot semantics is summarized, and the scene type with the highest proportion is selected as the scene type corresponding to the micro-drama.

5. The system, as claimed in claim 4, wherein the system is based on semantic analysis of micro-series dialogues. Based on the scene type corresponding to the micro-drama, a corresponding difference threshold is matched for the micro-drama, and then the matched difference threshold is compared with the difference rate between adjacent plot segments. If the difference rate of plot semantics between adjacent plot segments is less than the difference threshold, the two plot segments are merged to update the plot segment list. Conversely, if the difference rate of plot semantics between adjacent plot segments is greater than the difference threshold, the two plot segments will not be merged. The merging of plot segments includes merging scene data, character data, dialogue data, and plot semantics.

6. The system, as claimed in claim 1, wherein the semantic analysis based micro-episode dialogue intelligent optimization system is characterized by: In the dialogue reorganization and optimization module, dialogue data of the same type of micro-drama that matches the scene type of the micro-drama are collected, a reference dialogue database is established based on the collected dialogue data, and the dialogue data in the reference dialogue database is used as reference dialogue data. The dialogue data of the plot segments are correlated with scene data, character data, and plot semantics to obtain scene correlation, character correlation, and plot correlation. The three together constitute a multidimensional correlation, and the value of this multidimensional correlation is used as the benchmark threshold.

7. The semantic analysis based micro-series script intelligent optimization system according to claim 6, characterized in that: Based on the benchmark threshold, the reference lines data in the reference lines database are correlated with scene data, character data, and plot semantics to obtain the multidimensional correlation degree corresponding to each reference lines data. Then, the multidimensional correlation degree corresponding to each reference lines data is compared with the benchmark threshold. If the multidimensional correlation is greater than the baseline threshold, the reference dialogue data is retained. Conversely, if the multidimensional correlation is less than the baseline threshold, the reference line data will not be retained and will be deleted from the reference line database. The retained reference dialogue data is randomly semantically fused and recombined with the original dialogue data of the corresponding plot segment. The recombined dialogue data is summarized and a recombined dialogue data list is generated. At the same time, the multidimensional correlation degree of each recombined dialogue data in the recombined dialogue data list is calculated to obtain the multidimensional correlation degree corresponding to the recombined dialogue data. Then, the recombined dialogue data with the highest multidimensional correlation degree is selected to optimize the plot segment. The optimization process prioritizes the first plot segment of a short drama as the first target for optimization. After optimizing this plot segment, subsequent plot segments are selected in the natural order of the plot segment list, and the optimization is repeated until all plot segments in the list are optimized.

8. The system, as claimed in claim 1, wherein the semantic analysis based micro-episode dialogue intelligent optimization system is characterized by: In the full-drama dialogue coherence verification module, relevant thresholds are set in the plot segment list based on the difference rate of plot semantics between each plot segment. Among them, the lower the difference rate, the lower the correlation threshold; The higher the difference rate, the higher the correlation threshold.

9. The semantic analysis based micro-series script intelligent optimization system according to claim 8, characterized in that: In the full-drama dialogue coherence verification module, the optimized recombined dialogue data of each plot segment is analyzed for correlation. The semantic, expressive style and sentence structure features of the optimized recombined dialogue data of each plot segment are extracted. The degree of fit of the recombined dialogue data of adjacent plot segments in terms of semantic, expressive style and sentence structure features is calculated. This is used as the correlation between the recombined dialogue data of each plot segment. Then, the correlation is compared with the relevant threshold. If the relevance is lower than the relevance threshold, the recombined dialogue data list will be re-selected until the relevance meets the relevance threshold. Conversely, if the correlation is higher than the correlation threshold, monitoring continues.

Images