Stroma detection method and device for script water injection, electronic equipment and storage medium

Abstract

The invention provides a script water injection plot detection method and device, electronic equipment and a storage medium. The method comprises the following steps: dividing a to-be-detected script into a plurality of plot units, and generating a plot abstract for each plot unit; for each plot unit, determining a first contribution degree of the plot unit to a core plot element according to the plot abstract of the plot unit and the full-plot plot abstract; determining the plot units of which the corresponding first contribution degrees are lower than a first preset threshold value as potential water injection plots; according to the script original text of the potential water injection plot and the full-drama plot abstract, determining a second contribution degree of the potential water injection plot to a core plot element; and determining the potential water injection plot of which the corresponding second contribution degree is lower than a second preset threshold value as a target water injection plot. Therefore, the accurate quantification of the actual narrative utility of each plot unit is realized, the reliability of water injection detection is remarkably improved, and a scientific decision basis is provided for script quality control.

Description

Technical Field

[0001] The present application relates to the technical field of script analysis, and in particular to a method, device, electronic device and storage medium for detecting water-injected script plots. Background Art

[0002] With the rapid development of film, television, and online dramas, script quality assessment has become a crucial component of the production process. Script bloating (i.e., redundant plots unrelated to the main plot) directly impacts the narrative rhythm and audience experience. Existing script bloating detection techniques primarily rely on manual annotation or simple rule matching. For example, the likelihood of bloating can be inferred by counting the proportion of supporting characters or the length of side plots.

[0003] However, traditional detection methods are unable to accurately measure the actual narrative effectiveness of each plot unit, resulting in insufficient reliability of water injection detection results and making it difficult to provide a scientific basis for script quality control. Summary of the Invention

[0004] The purpose of the embodiments of this application is to provide a method, device, electronic device, and storage medium for detecting water-filled plots in scripts, so as to address the problem that traditional detection methods cannot accurately measure the actual narrative effectiveness of each plot unit, resulting in insufficient reliability of water-filled detection results and difficulty in providing a scientific basis for script quality control. The specific technical solution is as follows:

[0005] In a first aspect, the present application provides a method for detecting water-injected plots in scripts, comprising:

[0006] Divide the script to be tested into several plot units and generate a plot summary for each plot unit;

[0007] For each plot unit, determining the first contribution of the plot unit to the core plot elements based on the plot summary of the plot unit and the plot summary of the entire play;

[0008] Determining a plot unit corresponding to a first contribution degree lower than a first preset threshold as a potential water injection plot;

[0009] Determining a second contribution of the potential bloated plot to the core plot elements based on the original script of the potential bloated plot and the plot summary of the entire play;

[0010] The potential water injection scenario corresponding to the second contribution degree lower than the second preset threshold is determined as the target water injection scenario.

[0011] In one possible implementation, determining the first contribution of the plot unit to the core plot elements based on the plot summary of the plot unit and the plot summary of the entire play includes:

[0012] inputting the plot summary and the full-episode plot summary into a pre-trained large language model to perform multi-dimensional analysis by the large language model, including: analyzing a driving effect of the plot summary on the main plot development; analyzing a characterization effect of the plot summary on the main character's personality characteristics; and analyzing an interpretation effect of the plot summary on the expression of the work theme;

[0013] performing weighted summation operation on each dimension analysis result output by the large language model to obtain the first contribution degree.

[0014] In one possible implementation, the method further includes:

[0015] inputting the plot summary and the full-episode plot summary into a pre-trained large language model to perform multi-dimensional analysis by the large language model, including: analyzing a driving effect of the plot summary on the main plot development; analyzing a characterization effect of the plot summary on the main character's personality characteristics; and analyzing an interpretation effect of the plot summary on the expression of the work theme;

[0016] performing weighted summation operation on each dimension analysis result output by the large language model to obtain the first contribution degree.

[0017] In one possible implementation, the method further includes:

[0018] analyzing the plot content of each target water injection plot to identify target characters or target branch plots involved in each target water injection plot;

[0019] clustering and integrating multiple target water injection plots involving the same target character or the same target branch plot to form a corresponding water injection branch.

[0020] In one possible implementation, the method further includes:

[0021] For each water injection branch, calculating a ratio of the total number of words of the water injection branch to the total number of words of the to-be-detected script to obtain a branch full-episode water injection proportion;

[0022] calculating a ratio of the total number of words of the water injection branch in each episode to the total number of words of each episode to obtain a branch single-episode water injection proportion;

[0023] generating a water injection branch distribution heat map according to the branch full-episode water injection proportion and the branch single-episode water injection proportion.

[0024] In one possible implementation, the method further includes:

[0025] dividing the to-be-detected script by episodes to obtain episode texts.

[0026] For each episode text, a ratio of a total number of target waterlogging plots in the episode text to a total number of words in the episode text is calculated to obtain a plot single-episode waterlogging proportion.

[0027] In one possible implementation, the method further includes:

[0028] Counting a total number of plot waterlogging words of all target waterlogging plots in the script to be detected;

[0029] A ratio of the total number of plot waterlogging words to a total number of words in the script to be detected is calculated to obtain a plot waterlogging proportion.

[0030] In a second aspect, the present application provides a script waterlogging plot detection device, comprising:

[0031] A division module is configured to divide a script to be detected into a plurality of plot units, and generate a plot abstract for each plot unit;

[0032] A first determination module is configured to determine, for each plot unit, a first contribution degree of the plot unit to a core plot element according to a plot abstract of the plot unit and a plot abstract of the whole script;

[0033] A second determination module is configured to determine, as a potential waterlogging plot, a plot unit corresponding to a first contribution degree lower than a first preset threshold;

[0034] A third determination module is configured to determine, according to a script original text of the potential waterlogging plot and the plot abstract of the whole script, a second contribution degree of the potential waterlogging plot to the core plot element;

[0035] A fourth determination module is configured to determine, as a target waterlogging plot, a potential waterlogging plot corresponding to a second contribution degree lower than a second preset threshold.

[0036] In one possible implementation, the first determination module is specifically configured to:

[0037] input the plot abstract and the plot abstract of the whole script into a pre-trained large language model to perform multi-dimensional analysis through the large language model, including: analyzing a driving effect of the plot abstract on the main plot development; analyzing a characterization effect of the plot abstract on the main character's personality characteristics; analyzing an interpretation effect of the plot abstract on the expression of the work theme;

[0038] perform weighted summation operation on each dimension analysis result output by the large language model to obtain the first contribution degree.

[0039] In one possible implementation, the third determination module is specifically configured to:

[0040] input the script original text and the full-episode plot summary into a pre-trained large language model to perform multi-dimensional analysis by the large language model, including: analyzing the driving effect of the script original text on the main plot development; analyzing the characterization effect of the script original text on the main character's personality characteristics; analyzing the interpretation effect of the script original text on the expression of the work theme;

[0041] performing weighted summation operation on the multi-dimensional analysis results output by the large language model to obtain the second contribution degree.

[0042] In one possible implementation, the device further includes a branch line determination module configured to:

[0043] analyze the plot content of each target water injection plot to identify target characters or target branch plots involved in each target water injection plot;

[0044] cluster and integrate multiple target water injection plots involving the same target character or the same target branch plot to form a corresponding water injection branch.

[0045] In one possible implementation, the device further includes a generation module configured to:

[0046] For each water injection branch, calculate the ratio of the total number of words of the water injection branch to the total number of words of the to-be-detected script to obtain a branch full-episode water injection proportion;

[0047] calculate the ratio of the total number of words of the water injection branch in each episode to the total number of words of each episode to obtain a branch single-episode water injection proportion;

[0048] generate a water injection branch distribution heat map according to the branch full-episode water injection proportion and the branch single-episode water injection proportion.

[0049] In one possible implementation, the device further includes a first calculation module configured to:

[0050] divide the to-be-detected script by episode to obtain episode texts;

[0051] For each episode text, calculate the ratio of the total number of words of the target water injection plot in the episode text to the total number of words of the episode text to obtain a plot single-episode water injection proportion.

[0052] In one possible implementation, the device further includes a second calculation module configured to:

[0053] count the full-episode plot total number of words of all target water injection plots in the to-be-detected script;

[0054] calculate the ratio of the full-episode plot total number of words to the total number of words of the to-be-detected script to obtain a plot full-episode water injection proportion.

[0055] In a third aspect, an electronic device is provided, comprising a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other through the communication bus;

[0056] The memory is configured to store a computer program.

[0057] The processor is configured to execute the program stored in the memory to implement the method steps of any one of the first aspect.

[0058] In a fourth aspect, a computer readable storage medium is provided, characterized in that the computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the method steps of any one of the first aspect.

[0059] In a fifth aspect, a computer program product comprising instructions which, when executed on a computer, cause the computer to carry out the script waterlogging plot detection method of any one of the above aspects.

[0060] The embodiments of the present application have the following beneficial effects:

[0061] The embodiments of the present application provide a script waterlogging plot detection method and device, an electronic device and a storage medium. In the present application, the script content is first structured and processed, divided into a plurality of plot units and corresponding abstracts are generated, thereby establishing a unified basis for subsequent analysis. Then, by comparing the abstracts of each plot unit with the abstract of the whole plot, the first contribution degree of each plot unit to the core plot elements is calculated, and preliminary quantitative screening is completed. Finally, for the potential waterlogging plot screened out, the second contribution degree is calculated based on the original text content, and the accuracy of the evaluation result is ensured through double threshold determination. Through the layer-by-layer evaluation process, the technical scheme realizes accurate quantification of the actual narrative utility of each plot unit, significantly improves the reliability of waterlogging detection, and provides a scientific decision basis for script quality control.

[0062] Of course, implementing any product or method of the present application does not necessarily require all the advantages described above to be achieved at the same time. BRIEF DESCRIPTION OF DRAWINGS

[0063] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and serve to explain the principles of the present application together with the specification.

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced as follows. Obviously, for those skilled in the art, other drawings can also be obtained based on these drawings without any creative effort.

[0065] One or more embodiments are illustrated by way of example in the figures that are part of this disclosure, which illustrate the general manner of an implementation, which should not be construed as limiting. Similar elements in the drawings are denoted by like reference numerals for consistency throughout the various figures. The figures are not necessarily drawn to scale, and the dimensions of the various features can have been exaggerated or minimized for the sake of clarity.

[0066] Figure 1 A flow chart of a script waterlogging plot detection method provided for an embodiment of the present application;

[0067] Figure 2 A flow chart of another script waterlogging plot detection method provided for an embodiment of the present application;

[0068] Figure 3 A structural schematic diagram of a script waterlogging plot detection device provided for an embodiment of the present application;

[0069] Figure 4 A structural schematic diagram of an electronic device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0070] In order to make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in connection with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of protection of the present application.

[0071] The following disclosure provides many different embodiments, or examples, for implementing different structures of the present application. For the purpose of simplicity, the elements and settings of the particular examples in the following description are depicted in the drawings. Of course, they are merely examples and are presented to provide an illustrative discussion of the application. Therefore, the purpose is not to limit the present application. Furthermore, the present application can refer to a reference numeral and / or letter in different examples. Such repetition is for the sake of simplicity and clarity and does not indicate a relationship between the various embodiments and / or settings discussed.

[0072] Figure 1A flowchart of a script water injection plot detection method provided by an embodiment of the present application is shown. The method can be applied to one or more electronic devices such as a smart phone, a notebook computer, a desktop computer, a portable computer, and a server. In addition, the execution subject of the method can be hardware or software. When the execution subject is hardware, the execution subject can be one or more of the electronic devices. For example, a single electronic device can execute the method, or multiple electronic devices can cooperate with each other to execute the method. When the execution subject is software, the method can be implemented as multiple software or software modules, or as a single software or software module. No specific limitation is imposed herein.

[0073] As shown in Figure 1 , the method specifically includes:

[0074] S101, dividing a script to be detected into a plurality of episode units, and generating an episode abstract for each episode unit.

[0075] An episode unit refers to a plot segment composed of a plurality of script scenes that are continuous in time and consistent in theme, and the division basis includes time continuity and event theme consistency.

[0076] An episode abstract is a concise expression of the core content of an episode unit, and needs to include the main events and key information of the unit.

[0077] In the embodiments of the present application, the following technical solutions are adopted to divide the plot units: (1) time continuity determination: by analyzing the time sequence marks (such as "next day", "same day evening") or scene transition descriptions between episodes, the time-continuous episodes are classified into the same unit; (2) theme consistency determination: based on natural language processing technology, the keywords (such as characters, places, events) of each episode are extracted, and the theme-related continuous episodes are clustered by semantic similarity calculation (such as cosine similarity > 0.7) to be classified into the same unit. The generation of plot abstracts is realized by the following scheme: first, the core sentences of subject-predicate-object structure are extracted from the dialogue and description text of each episode; then, the NER (Named Entity Recognition) technology and TF-IDF (term frequency-inverse document frequency) algorithm are applied to count the entities such as characters, objects, and places in each episode that appear more than a preset threshold (such as 3 times). Finally, the core sentences and key entities obtained in the previous two steps are input into the pre-trained BART (Bidirectional and Auto-Regressive Transformer) abstract model (which adopts an encoder-decoder architecture to understand the context through bidirectional encoding and generate an abstract through self-regressive decoding), and a standardized abstract of 50-100 words is output. The model is fine-tuned with script field text during training to ensure that the generated abstract contains the necessary plot elements.

[0078] This step realizes the structured processing of script content through the above technical means, establishes a standardized basis for subsequent quantitative analysis, avoids the complexity of directly processing the original script text, and retains the necessary narrative information, thereby providing an operable analysis unit for subsequent water injection detection.

[0079] S102, for each plot unit, determining a first contribution degree of the plot unit to a core plot element according to a plot abstract of the plot unit and a full-plot plot abstract.

[0080] The core plot element refers to a set of key dimensions that constitute the narrative value of the script, including: main plot development dimension: the role of the plot in promoting the development and resolution of the core conflict of the story; main character shaping dimension: the role of the plot in portraying the character traits and growth arc of the main character and key supporting characters; theme expression dimension: the role of the plot in interpreting and deepening the core idea of the work (such as love, justice, and other themes).

[0081] The first contribution degree is a comprehensive evaluation index for quantifying the contribution degree of the plot unit to the above core elements.

[0082] In the embodiments of the present application, the first contribution degree of the plot unit to the core plot element is determined according to the plot summary of the plot unit and the full-plot plot summary, which can include the following steps: inputting the plot summary and the full-plot plot summary into a pre-trained large language model to perform multi-dimensional analysis through the large language model, including: analyzing the driving effect of the plot summary on the main plot development; analyzing the characterization effect of the plot summary on the main character's personality characteristics; analyzing the interpretation effect of the plot summary on the expression of the work theme; and performing weighted summation operation on the multi-dimensional analysis results output by the large language model to obtain the first contribution degree.

[0083] The pre-trained large language model refers to a natural language processing model (such as the GPT (Generative Pre-trained Transformer) series) trained based on massive text data and having deep semantic understanding capability, which captures deep associations between texts through an attention mechanism. The weighted summation operation refers to a calculation method of linear combination of each dimension score according to a preset weight (such as main line driving 0.5, character portrayal 0.3, theme interpretation 0.2).

[0084] In specific implementation, after the plot summary and the full-plot plot summary are structured and spliced, they are input into the large language model, and specific prompt templates (such as "Please evaluate the driving degree of the plot on the main line development (1-5 points)") are designed to guide the model to output quantitative scores of three dimensions, and the first contribution degree is obtained by weighted summation after normalization processing. Through this scheme, the deep semantic analysis capability of the large language model can be used to realize multi-dimensional quantitative evaluation of the plot narrative function, which not only avoids the subjectivity of manual evaluation, but also identifies potential important plots (such as transition plots with hidden clues) that are difficult to find by traditional methods, providing an objective and quantifiable judgment criterion for subsequent water injection plot screening.

[0085] S103, determining the plot unit corresponding to the first contribution degree lower than the first preset threshold as a potential water injection plot.

[0086] The first preset threshold refers to a pre-set contribution degree threshold value for preliminarily distinguishing between effective plots and potential water injection plots.

[0087] In the embodiments of the present application, the first contribution degree calculated is compared with the first preset threshold (such as 0.4 points, full score 1 point), and the plot unit lower than the value is marked as a potential water injection plot. This step realizes preliminary screening of the plot unit by setting an objective quantitative standard, which not only retains possible important plots (contribution degree higher than the threshold), but also effectively reduces the candidate range that needs to be analyzed in depth, significantly improving the subsequent processing efficiency.

[0088] S104, determine the second contribution degree of the potential water injection plot to the core plot element according to the script original of the potential water injection plot and the full-plot plot summary.

[0089] The script original refers to the complete text content contained in the plot unit, including dialogue, action description, and scene details, and other original creative information.

[0090] The second contribution degree refers to a refined contribution degree evaluation index based on the script original.

[0091] In the embodiments of the present application, S104 can include the following steps: inputting the script original and the full-plot plot summary into a pre-trained large language model to perform multi-dimensional analysis through the large language model, including: analyzing the driving effect of the script original on the main plot development; analyzing the characterization effect of the script original on the main character's personality characteristics; analyzing the interpretation effect of the script original on the expression of the work theme; performing weighted summation operation on the multi-dimensional analysis results output by the large language model to obtain the second contribution degree.

[0092] In this scheme, the complete original text (retaining all narrative details) of the potential water injection plot and the full-plot plot summary are input into a pre-trained large language model together, through the design of detailed prompt instructions (such as "analyze the contribution of the following dialogue to the characterization of the character's personality"), guide the large language model to deeply mine the narrative elements (such as implied character relationship development, key prop foreshadowing) that may be ignored in the summary, and recalculate their scores in the three dimensions of main line driving, character shaping and theme expression, and output the second contribution degree after weighted summation (the weight can be the same as or adjusted from the first contribution degree). This scheme can verify the accuracy of the preliminary screening results through in-depth analysis of the original text, and can also find narrative elements that are not reflected in the summary but are actually important (such as character personality characterization through subtle dialogue), ensuring that the final determined water injection plot has higher reliability.

[0093] S105, determining the potential water injection plot corresponding to the second contribution degree lower than the second preset threshold as the target water injection plot.

[0094] The second preset threshold refers to the contribution degree critical value set for the original text level analysis, which is usually set to a more stringent value (for example, 0.3 points, full score 1 point) than the first preset threshold, to reflect higher requirements for analysis of the original text details.

[0095] In the embodiments of the present application, the system compares the second contribution degree of the potential water injection case with the threshold value, and the case lower than the threshold value is finally determined as the target water injection case. This step implements a double threshold screening mechanism, and its beneficial effects include: 1) the initial screening based on the summary ensures the analysis efficiency, 2) the review based on the original text ensures the determination accuracy, 3) the final output of the target water injection case contains both obvious redundant content (such as irrelevant branches) and seemingly reasonable but actually invalid cases (such as repetitive character interactions), providing accurate modification targets for script optimization. This hierarchical determination method effectively balances the evaluation efficiency and accuracy, solving the double problems of high misjudgment rate and high omission rate in the traditional method of water injection case recognition.

[0096] In the embodiments of the present application, first, the script content is structured and divided into several episode units and corresponding summaries are generated, establishing a unified basis for subsequent analysis; then, by comparing the summaries of each episode unit with the summary of the whole drama plot, the first contribution degree to the core plot elements is calculated, completing the preliminary quantitative screening; finally, for the potential water injection cases screened out, the second contribution degree is calculated based on the original text content, and the accuracy of the evaluation result is ensured through double threshold determination. This technical solution realizes the accurate quantification of the actual narrative utility of each episode unit through a progressive evaluation process, significantly improving the reliability of water injection detection and providing a scientific basis for script quality control.

[0097] Referring to Figure 2 An embodiment flowchart of another script water injection plot detection method provided by the embodiments of the present application is shown. As Figure 2 shown, the flowchart can include the following steps:

[0098] S201, analyze the plot content of each target water injection case, and identify the target characters or target branch plots involved in each target water injection case.

[0099] S202, cluster and integrate multiple target water injection cases involving the same target character or the same target branch plot to form a corresponding water injection branch.

[0100] For ease of understanding, S201 and S202 are described together as follows:

[0101] The target character refers to a non-main character (such as a supporting actor name extracted by named entity recognition) that repeatedly appears in the target water injection case.

[0102] The target branch plot refers to an independent story unit composed of multiple related water injection cases (such as a love line identified by event keywords such as "competition" and "love").

[0103] In the embodiments of the present application, the water injection branch line can be obtained through the following steps: (1) performing semantic analysis on each target water injection scenario to extract the character name, key action (such as "confession" and "duel") and scene elements; (2) clustering and integrating scenarios involving the same character or similar events (such as "matchmaking by martial arts" and "treasure hunting") based on a semantic similarity algorithm to form a structured water injection branch line set, such as based on a hierarchical clustering algorithm, scenarios with a character overlap of more than 70% or an event similarity of more than 0.6 are classified into the same branch line; and (3) performing manual readability annotation on the clustering results (such as "Shi Kuan revenge branch line").

[0104] Figure 2 As shown in the flow, the algorithm automatically identifies water injection scenario groups with commonalities (such as merging eight tailing-related scenarios scattered in five episodes into a complete branch line), provides a clear water injection problem distribution map for the production party, and solves the defect that manual review cannot systematically find dispersed water injection modes.

[0105] In another embodiment of the present application, the method can further include the following steps: for each water injection branch line, calculating the ratio of the total number of words of the water injection branch line to the total number of words of the to-be-detected script to obtain a branch line full-episode water injection proportion; calculating the ratio of the total number of words of the water injection branch line in each episode to the total number of words of each episode to obtain a branch line single-episode water injection proportion; and generating a water injection branch line distribution heat map according to the branch line full-episode water injection proportion and the branch line single-episode water injection proportion.

[0106] The branch line full-episode water injection proportion refers to the percentage of the total number of words of all scenarios included in a specific water injection branch line to the total number of words of the script. The branch line single-episode water injection proportion refers to the percentage of the number of words of scenarios in a single episode. The distribution heat map is a visualization chart that visually displays the distribution of water injection branch lines in each episode through a color gradient.

[0107] The specific implementation includes: (1) counting the total number of words of target water injection scenarios of each water injection branch line in the full episode and each single episode; (2) calculating the ratio to the total number of words in the corresponding range (such as branch line A accounts for 5% in the full episode and 15% in the third episode); and (3) generating a two-dimensional heat map using a visualization tool (such as Matplotlib), with the horizontal axis representing the episode number and the vertical axis representing the branch line name, and the color depth of the color block representing the high and low of the water injection proportion. This scheme accurately identifies the concentration of water injection problems (such as the abnormal peak value of a branch line in a specific episode) through the combination of quantitative indicators (such as revealing that the "tailing branch line" accounts for 4.2% in the full episode and suddenly increases to 18% in the eighth episode) and visual presentation, provides data support for targeted reduction or reconstruction, and solves the defect that traditional experience-based modification lacks quantitative basis.

[0108] In still another embodiment of the present application, the method can further include the following steps: dividing the script to be detected according to episodes to obtain episode texts; for each episode text, calculating the ratio of the total number of target waterlogging plots in the episode text to the total number of words in the episode text to obtain a plot single-episode waterlogging proportion.

[0109] The episode text refers to the complete script content divided according to the broadcast unit, and includes the dialogue and description of all scenes in the episode. The plot single-episode waterlogging proportion refers to the proportion of the number of words of a target waterlogging plot in a single episode.

[0110] Specific implementation includes: (1) dividing the episode text according to the script format mark (such as "Xth episode"); (2) counting the total number of words of all target waterlogging plots contained in each episode text (such as 1500 words of 3 waterlogging plots in the 3rd episode); (3) calculating the percentage of the number of words to the total number of words in the episode (such as 1500 / 10000=15%). This scheme quantifies the waterlogging degree of each episode (such as finding that the waterlogging proportion of the 12th episode is 22%, which is significantly higher than the average of 8%), accurately locates the single-episode unit that needs to be optimized, solves the problem that the traditional overall evaluation cannot identify abnormal fluctuations in single episodes, and provides objective adjustment basis for the inter-episode rhythm balance of the script.

[0111] In addition, in still another embodiment of the present application, the method can further include the following steps: counting the total number of words of all target waterlogging plots in the script to be detected; calculating the ratio of the total number of words of all target waterlogging plots to the total number of words of the script to be detected to obtain a plot overall script waterlogging proportion.

[0112] The total number of words of all target waterlogging plots refers to the total number of words of all target waterlogging plots after multi-level judgment. The plot overall script waterlogging proportion refers to the percentage of the total number of words of all target waterlogging plots to the total number of words of the script.

[0113] Specific implementation includes: (1) aggregating the original text content marked as target waterlogging plots; (2) counting the cumulative number of words of these plots (such as 32,000 words of 85 waterlogging plots in the overall script); (3) calculating the ratio of the cumulative number of words to the total number of words of the script (3.2 / 40=8%). This scheme generates an objective overall script waterlogging quantitative indicator (such as "waterlogging rate 8%"), provides a benchmark evaluation data of the overall quality of the script for the production party, solves the defect that the traditional subjective evaluation cannot be quantified, and can scientifically judge whether the script needs to be modified overall by comparing with the industry benchmark value (such as 5%), thereby realizing the whole-process detection from micro-plot recognition to macro-quality evaluation.

[0114] Based on the same technical concept, the embodiments of the present application also provide a script waterlogging plot detection device, as shown in Figure 3 The device includes:

[0115] The dividing module 31 is configured to divide the script to be detected into a plurality of episode units, and generate an episode summary for each episode unit;

[0116] The first determining module 32 is configured to determine, for each episode unit, a first contribution degree of the episode unit to a core plot element according to the episode summary of the episode unit and the full-plot summary;

[0117] The second determining module 33 is configured to determine, as a potential watered episode, an episode unit corresponding to a first contribution degree lower than a first preset threshold;

[0118] The third determining module 34 is configured to determine, according to the script original text of the potential watered episode and the full-plot summary, a second contribution degree of the potential watered episode to the core plot element;

[0119] The fourth determining module 35 is configured to determine, as a target watered episode, a potential watered episode corresponding to a second contribution degree lower than a second preset threshold.

[0120] In one possible implementation, the first determining module is specifically configured to:

[0121] input the episode summary and the full-plot summary into a pre-trained large language model to perform multi-dimensional analysis by the large language model, including: analyzing a driving effect of the episode summary on the main plot development; analyzing a characterization effect of the episode summary on the main character's personality characteristics; analyzing an interpretation effect of the episode summary on the expression of the work theme;

[0122] perform weighted summation operation on each dimension analysis result output by the large language model to obtain the first contribution degree.

[0123] In one possible implementation, the third determining module is specifically configured to:

[0124] input the script original text and the full-plot summary into a pre-trained large language model to perform multi-dimensional analysis by the large language model, including: analyzing a driving effect of the script original text on the main plot development; analyzing a characterization effect of the script original text on the main character's personality characteristics; analyzing an interpretation effect of the script original text on the expression of the work theme;

[0125] perform weighted summation operation on each dimension analysis result output by the large language model to obtain the second contribution degree.

[0126] In one possible implementation, the apparatus further includes a branch determining module configured to:

[0127] analyze the plot content of each target watered episode to identify target roles or target branch plots involved in each target watered episode;

[0128] The multiple target water injection plots related to the same target role or the same target branch plot are clustered and integrated to form a corresponding water injection branch.

[0129] In a possible implementation, the apparatus further includes a generating module configured to:

[0130] For each water injection branch, a ratio of a total number of words of the water injection branch to a total number of words of the to-be-detected script is calculated to obtain a branch full-episode water injection proportion.

[0131] A ratio of a total number of words of the water injection branch in each episode to a total number of words of each episode is calculated to obtain a branch single-episode water injection proportion.

[0132] According to the branch full-episode water injection proportion and the branch single-episode water injection proportion, a water injection branch distribution heat map is generated.

[0133] In a possible implementation, the apparatus further includes a first calculating module configured to:

[0134] The to-be-detected script is divided by episode to obtain episode texts;

[0135] For each episode text, a ratio of a total number of words of a target water injection plot in the episode text to a total number of words of the episode text is calculated to obtain a plot single-episode water injection proportion.

[0136] In a possible implementation, the apparatus further includes a second calculating module configured to:

[0137] A full-episode plot total number of words of all target water injection plots in the to-be-detected script is counted.

[0138] A ratio of the full-episode plot total number of words to a total number of words of the to-be-detected script is calculated to obtain a plot full-episode water injection proportion.

[0139] In the embodiments of the present application, first, the script content is structured and processed, divided into a plurality of plot units and corresponding abstracts are generated, to establish a unified basis for subsequent analysis; then, by comparing the abstracts of the plot units with the full-episode plot abstract, a first contribution degree of the plot units to the core plot elements is calculated, to complete preliminary quantitative screening; finally, for the potential water injection plots screened out, a second contribution degree is calculated based on the original content of the plots, to ensure the accuracy of the evaluation result through double threshold determination. Through the layer-by-layer evaluation process, the technical solution realizes accurate quantification of the actual narrative utility of each plot unit, significantly improves the reliability of water injection detection, and provides a scientific decision basis for script quality control.

[0140] Based on the same technical concept, the embodiments of the present application also provide an electronic device, such as Figure 4As shown, the electronic device includes a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112 and the memory 113 communicate with each other through the communication bus 114,

[0141] The memory 113 is configured to store a computer program.

[0142] The processor 111 is configured to execute the program stored in the memory 113 to implement the following steps:

[0143] Divide the to-be-detected script into a plurality of episode units, and generate an episode summary for each episode unit;

[0144] For each episode unit, determine a first contribution degree of the episode unit to a core plot element according to the episode summary of the episode unit and a full-plot plot summary;

[0145] Determine an episode unit corresponding to a first contribution degree lower than a first preset threshold as a potential water injection episode;

[0146] Determine a second contribution degree of the potential water injection episode to the core plot element according to a script original text of the potential water injection episode and the full-plot plot summary;

[0147] Determine a potential water injection episode corresponding to a second contribution degree lower than a second preset threshold as a target water injection episode.

[0148] The communication bus mentioned in the above electronic device can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The communication bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or only one type of bus.

[0149] The communication interface is configured to communicate between the above electronic device and other devices.

[0150] The memory can include a random access memory (RAM) and can also include a non-volatile memory (NVM), such as at least one disk memory. Optionally, the memory can also be at least one storage device located away from the aforementioned processor.

[0151] The processor described above can be a general processor, including a central processing unit (CPU), a network processor (NP), etc.; can also be a digital signal processor (DSP), an application specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component.

[0152] In another embodiment provided in the present application, a computer readable storage medium is also provided, and the computer readable storage medium stores a computer program. The computer program is executed by a processor to implement the steps of any of the script water plot detection methods.

[0153] In another embodiment provided in the present application, a computer program product containing instructions is also provided, and when the computer program product is run on a computer, the computer is caused to execute any of the script water plot detection methods in the above embodiments.

[0154] The device embodiments described above are only schematic, wherein the units shown as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e., can be located in one place, or can be distributed on multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the present embodiment scheme.

[0155] From the above description of the embodiments, those skilled in the art can clearly understand that the embodiments can be implemented by means of software plus a general hardware platform, and of course can also be implemented by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in the various embodiments or some parts of the embodiments.

[0156] It is to be understood that the terminology used herein is for the purpose of describing particular example embodiments only and is not intended to be limiting. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. The terms "comprises", "comprising", "includes", "including" and "has" are inclusive and therefore specify the presence of stated features, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, steps, operations, elements, components, and / or groups thereof. The method steps, processes, and operations described herein are not to be construed as necessarily requiring their performance in the particular order

[0157] The above description is merely that of the specific embodiments of the application and as such is not to be taken in a limiting sense. Various modifications and alterations of the embodiments described herein will become apparent to those skilled in the art from the foregoing description, which does not limit the generic principles of the application and the present application is intended to encompass all such modifications and alternative constructions. The scope of the application should be determined by the following claims:

Claims

1. A method for detecting water-injected plots in scripts, characterized in that: The method comprises: Divide the script to be tested into several plot units and generate a plot summary for each plot unit; For each plot unit, determining the first contribution of the plot unit to the core plot elements based on the plot summary of the plot unit and the plot summary of the entire play; Determining a plot unit corresponding to a first contribution degree lower than a first preset threshold as a potential water injection plot; Determining a second contribution of the potential watered-down plot to the core plot elements based on the original script of the potential watered-down plot and the plot summary of the entire play; The potential water injection scenario corresponding to the second contribution degree lower than the second preset threshold is determined as the target water injection scenario.

2. The method according to claim 1, characterized in that Determining the first contribution of the plot unit to the core plot elements based on the plot summary of the plot unit and the plot summary of the entire play includes: Inputting the plot summary and the full play plot summary into a pre-trained large language model to perform multi-dimensional analysis using the large language model, including: analyzing the role of the plot summary in promoting the development of the main plot; analyzing the role of the plot summary in portraying the personality traits of the main characters; and analyzing the role of the plot summary in explaining the theme of the work; A weighted sum operation is performed on the analysis results of each dimension output by the large language model to obtain the first contribution.

3. The method according to claim 1, characterized in that Determining the second contribution of the potential watered-down plot to the core plot elements based on the original script of the potential watered-down plot and the full plot summary includes: The original script and the plot summary of the entire play are input into a pre-trained large language model to perform multi-dimensional analysis through the large language model, including: analyzing the role of the original script in promoting the development of the main plot; analyzing the role of the original script in portraying the personality traits of the main characters; and analyzing the role of the original script in explaining the theme of the work; A weighted sum operation is performed on the analysis results of each dimension output by the large language model to obtain the second contribution.

4. The method according to claim 1, wherein The method further comprises: Analyze the plot content of each target water-injection plot and identify the target characters or target side plots involved in each target water-injection plot; Multiple target water-injection plots involving the same target character or the same target branch plot are clustered and integrated to form corresponding water-injection branches.

5. The method according to claim 4, characterized in that The method further comprises: For each water-injected branch, calculate the ratio of the total number of words in the water-injected branch to the total number of words in the script to be tested, and obtain the water-injected ratio of the branch to the entire script; Calculate the ratio of the total number of words in each episode of the water-injected branch to the total number of words in each episode to obtain the water-injected ratio of each episode of the branch; A heat map of the water injection branch line distribution is generated according to the water injection ratio of the entire branch line and the water injection ratio of a single branch line.

6. The method according to claim 1, characterized in that The method further comprises: Divide the script to be tested into episodes to obtain the text of each episode; For each episode of the drama, the ratio of the total number of words in the target water-injected plot in the drama text to the total number of words in the drama text is calculated to obtain the water-injected ratio of the plot in a single episode.

7. The method according to claim 1, characterized in that The method further comprises: Counting the total number of water-injected plot words in all target water-injected plots in the script to be tested; Calculate the ratio of the total number of words in the full play to the total number of words in the script to be tested to obtain the proportion of the full play to be filled with water.

8. A device for detecting water-injected plots in scripts, characterized in that: The device comprises: A division module is used to divide the script to be tested into several plot units and generate a plot summary for each plot unit; A first determination module is configured to determine, for each plot unit, a first contribution of the plot unit to the core plot elements based on the plot summary of the plot unit and the plot summary of the entire play; A second determining module is configured to determine a plot unit corresponding to a first contribution degree lower than a first preset threshold as a potential water injection plot; A third determining module is configured to determine a second contribution of the potential watered-down plot to the core plot elements based on the original script of the potential watered-down plot and the plot summary of the entire play; The fourth determining module is configured to determine a potential water injection scenario corresponding to a second contribution degree lower than a second preset threshold as a target water injection scenario.

9. An electronic device, characterized in that: It includes a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory communicate with each other via the communication bus; Memory for storing computer programs; The processor is used to implement the script water injection plot detection method described in any one of claims 1-7 when executing the program stored in the memory.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method for detecting water-injected plots in a script according to any one of claims 1 to 7 is implemented.

Images