Scenario generation method and system for novel reorganization

Abstract

The invention discloses a script generation method and system for novel reorganization. The method comprises the steps of obtaining a novel text to be reorganized; performing element extraction based on the novel text to be recompiled to obtain target element information; abstract refining is carried out based on the target element information, and a refined narrative unit library is obtained; performing plot extraction based on the refined narrative unit library to obtain a climax plot position for the novel text to be recompiled; and based on the target element information, the refined narrative unit library and the climax plot position, carrying out structured sticking point design to obtain a target script for the novel text to be recompiled. The problems that a traditional novel reorganization process highly depends on manpower, logic consistency of front and back plots is difficult to guarantee, and the quality of a generated script is low are solved.

Description

Technical Field

[0001] This application relates to the field of text processing technology, and in particular to a method and system for generating scripts for novel adaptation. Background Technology

[0002] Currently, adapting long-form online novels and other IPs into short dramas has become an important business model in the digital content industry. Such adaptations not only effectively utilize the reader base accumulated by the original work, but also bring considerable content traffic and commercial returns to the platform.

[0003] Currently, novel adaptations are often done manually. However, due to the original work's length, complex plot, and numerous characters, the screenwriting team needs to invest a significant amount of time and energy in plot extraction, structural reorganization, and dialogue design. This is not only inefficient but also prone to human oversight, leading to missing key plot points or deviations in character settings. Furthermore, without systematic tools to assist in episode segmentation and pacing control, problems such as plot inconsistencies, lost foreshadowing, or emotional discontinuities are easily encountered, severely impacting the narrative coherence and audience experience of the short drama. Clearly, the traditional novel adaptation process relies heavily on manual intervention, making it difficult to guarantee logical consistency throughout the plot and resulting in low-quality scripts. Summary of the Invention

[0004] To overcome the problems of traditional novel adaptation processes that rely heavily on manual labor, making it difficult to ensure logical consistency between plot points and resulting in low-quality scripts, this application provides a script generation method and system for novel adaptation.

[0005] Firstly, in order to solve the aforementioned technical problems, this application provides a script generation method for novel adaptation, comprising: Obtain the text of the novel to be adapted; Based on the text of the novel to be adapted, the target element information is obtained by extracting elements. Based on the target element information, a refined narrative unit library is obtained by summarizing and refining the information. Plot extraction is performed based on a refined narrative unit library to obtain the climax plot positions for the novel text to be adapted; Based on target element information, a refined narrative unit library, and the location of climax plot points, a structured checkpoint design is performed to obtain a target script for the novel text to be adapted.

[0006] Secondly, this application also provides a script generation system for novel adaptation, comprising: The acquisition module is used to acquire the text of the novel to be adapted. The element extraction module is used to extract elements from the text of the novel to be adapted, and obtain target element information. The summary refinement module is used to refine summaries based on target element information, resulting in a refined narrative unit library. The plot extraction module is used to extract plots based on a refined narrative unit library to obtain the climax plot positions for the novel text to be adapted. The script generation module is used to design structured checkpoints based on target element information, a refined narrative unit library, and the location of climax plot points, to obtain the target script for the novel text to be adapted.

[0007] Thirdly, this application also provides a computing device, including a memory, a processor, and a program stored in the memory and running on the processor, wherein the processor executes the program to implement the steps of the script generation method for novel adaptation described above.

[0008] Fourthly, this application also provides a computer-readable storage medium storing instructions that, when executed on a terminal device, cause the terminal device to perform steps of a script generation method for novel adaptation.

[0009] The beneficial effects of this application are as follows: First, based on the novel text to be adapted, element extraction is performed to obtain target element information, and then the summary is refined based on the target element information to preserve the overall narrative logic of the novel text to be adapted, resulting in a refined narrative unit library. Second, based on the refined narrative unit library, plot extraction is performed to obtain the climax plot positions for the novel text to be adapted, and based on the target element information, the refined narrative unit library, and the climax plot positions, a structured checkpoint design is performed to obtain the target script for the novel text to be adapted, thereby achieving automated adaptation from novel to script. In this way, through element extraction and summary refinement, the overall narrative logic of the novel text to be adapted can be preserved, so that the target script obtained by automated adaptation can maintain the logical consistency of the plot, thereby improving the script quality of the target script. Attached Figure Description

[0010] Figure 1 This is a schematic flowchart illustrating an exemplary embodiment of the present application for a script generation method for novel adaptation; Figure 2 This is a schematic diagram illustrating the process of determining the location of a climax scene in an exemplary embodiment of this application; Figure 3 This is a schematic flowchart of the script generation method for novel adaptation provided by the application in an exemplary embodiment of this application; Figure 4 for Figure 3 The corresponding further detailed flowchart; Figure 5 This is a schematic diagram illustrating the structure of a script generation system for novel adaptation, as shown in an exemplary embodiment of this application. Detailed Implementation

[0011] The following embodiments are further explanations and supplements to this application and do not constitute any limitation on this application.

[0012] Traditional adaptation processes rely heavily on manual labor, resulting in a series of serious technical challenges: Poor robustness of preprocessing: Novel texts come from diverse sources, have varying chapter formats, and even contain a large number of irrelevant characters (such as advertisements and author notes). Traditional text segmentation methods based on fixed rules are ill-suited to this complexity, resulting in costly and inefficient data cleaning and formatting work in the early stages.

[0013] Efficiency and scalability bottlenecks: A novel of one million words requires a writing team to spend weeks or even months reading, understanding, refining, and planning. According to industry statistics, the average adaptation cycle is as long as 2-4 months. This labor-intensive model severely limits the large-scale production capacity of IP adaptations.

[0014] Logical consistency is difficult to guarantee: When dealing with long narratives, human memory and processing capacity are limited, which can easily lead to logical flaws such as inconsistencies in the plot and inconsistencies in character settings (OOC), which greatly affect the quality of the final work.

[0015] Business decisions are highly subjective: the judgment of key plot points such as "exciting moments" and "exciting moments," as well as how to set commercial "hook points," largely depends on the screenwriter's personal experience. This decision-making method lacks quantitative basis, has an unstable success rate, and carries a high market risk.

[0016] While existing technologies have proposed using AI for text summarization, most solutions are limited to localized and isolated summarization, failing to handle the complex contextual dependencies within long narratives of millions of words. These methods are essentially "bottom-up," lacking a macro-level understanding of the overall story structure, and therefore cannot solve the core challenges of adaptation: ensuring global logical consistency and conducting business structure planning based on the overall plot. Therefore, there is an urgent need in this field for a systematic, automated adaptation solution that takes a "macro-first, micro-second" approach.

[0017] To address the aforementioned issues, embodiments of this application provide a method and system for generating scripts for novel adaptations, which will be described in detail below.

[0018] This application effectively solves the problem of fragmented contextual logic caused by text segmentation in the processing of long novels, as well as network jitter and data consistency issues during concurrent calls to large-scale pre-trained language models (or servers), by constructing a globally unique internal key (Aegis Key) management mechanism and an adaptive API calling strategy. This enables the industrialized and stable production of million-word-level long novel adaptations. The core idea of ​​this application is "deconstruction first, then parallel synthesis, followed by correction, and finally planning," supplemented by three supporting modules: global data consistency management, adaptive API communication management, and dynamic strategy management. It can achieve the following objectives: Robust handling: Achieve universality and high fault tolerance for source text of any format.

[0019] Logical self-consistency: By constructing global knowledge, a high degree of logical consistency between micro-plots and macro-setting is ensured during the adaptation of long texts. Furthermore, it eliminates logical fragmentation caused by technical failures in the data processing flow (such as communication interruptions or inconsistent data identification) at the system level.

[0020] Structural Reconstruction: Efficiently and intelligently reconstructing linear literary narratives into non-linear dramatic structures that meet commercial needs.

[0021] Scientific decision-making: Providing closed-loop support for key business decisions (such as checkpoint setting), from quantitative data analysis to AI-driven qualitative creative decision-making.

[0022] The script generation method for novel adaptation provided in this application can be specifically executed by a server. It should be noted that the server can be a standalone server, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms; no limitation is imposed here.

[0023] It should be noted that the "pre-trained language model" mentioned in this specification refers to a neural network model that has been pre-trained on a large amount of text data and possesses the ability to understand, generate, or represent natural language. In specific implementations, this model can be a model based on the Transformer architecture (such as BERT, GPT series, T5, etc.), or a model with equivalent functionality based on other architectures that will emerge in the future as technology develops. This application does not limit the specific model architecture.

[0024] Please see Figure 1 , Figure 1 An exemplary embodiment of this application illustrates a script generation method for novel adaptation, such as... Figure 1 As shown, this application provides a script generation method for novel adaptation, including: S11, Obtain the text of the novel to be adapted; S12, Extract elements from the novel text to be adapted to obtain target element information; S13, Based on the target element information, the summary is refined to obtain a refined narrative unit library; S14, based on the refined narrative unit library, extract the plot points to obtain the climax plot positions for the novel text to be adapted; S15 involves designing structured checkpoints based on target element information, a refined narrative unit library, and the location of climax scenes to obtain the target script for the novel text to be adapted.

[0025] The script generation method for novel adaptation provided in this application first extracts elements from the novel text to be adapted, obtaining target element information. Then, it refines the summary based on the target element information to preserve the overall narrative logic of the novel text, resulting in a refined narrative unit library. Second, it extracts plot points from the refined narrative unit library to obtain the climax plot positions for the novel text. Finally, it designs structured checkpoints based on the target element information, the refined narrative unit library, and the climax plot positions to obtain the target script for the novel text, thus achieving automated adaptation from novel to script. In this way, through element extraction and summary refinement, the overall narrative logic of the novel text to be adapted can be preserved, ensuring that the automatically adapted target script maintains logical consistency throughout the plot, thereby improving the script quality. Simultaneously, because the climax plot positions are designed during script generation, each episode of the target script can form strong memorable and communicative points, attracting audience attention and increasing audience curiosity, thereby enhancing the commercial value of the target script.

[0026] Optionally, based on the text of the novel to be adapted, element extraction is performed to obtain target element information, including: The adapted novel text is deconstructed to generate multiple text blocks with overlapping areas; Chapter information is extracted based on text blocks to obtain micro-level element information for the text blocks; Text extraction is performed based on multiple micro-level information elements to obtain macro-level narrative information for the novel text to be adapted. Target element information is formed based on micro-level element information and macro-level narrative information.

[0027] In the embodiment provided in this application, the text structure of the novel to be adapted is performed, and chapter information is extracted from multiple text blocks with overlapping areas generated by the structure to obtain corresponding micro-element information. Based on multiple micro-element information, text refinement is performed to obtain macro-narrative information for the novel to be adapted, forming target element information containing micro-element information and macro-narrative information. This facilitates subsequent adaptation of the novel from both macro and micro perspectives, thereby preserving the overall narrative logic of the novel to be adapted, ensuring the integrity of the adapted target script, and thus improving the script quality of the target script.

[0028] In an exemplary embodiment provided in this application, the specific steps of text deconstruction are as follows: receiving a complete long novel text (the novel text to be adapted), executing the Indiscriminate Sliding Window Chunking algorithm to generate multiple text chunks with overlapping areas, which can transform physical segmentation into subsequent logical recognition, ensuring the integrity and high fault tolerance of information input.

[0029] The specific steps for extracting micro-level information are as follows: all text blocks are fed into the first pre-trained language model in parallel, which autonomously identifies and extracts the chapter information of the novel text to be adapted, and generates structured data containing atomic chapter_number (chapter number), chapter_title (chapter title), chapter summary, characters, and other micro-level information, thus obtaining the micro-level information of each text block.

[0030] Optionally, text extraction is performed based on multiple micro-level information elements to obtain macro-level narrative information for the novel text to be adapted, including: Align multiple micro-level information elements; Using preset text deduplication rules, text deduplication is performed based on the aligned information of multiple micro-elements to obtain multiple target micro-element units. By piecing together multiple target micro-element units, global narrative information is obtained; By extracting the core elements from the overall narrative information, we obtain macro-narrative information specific to the novel text to be adapted.

[0031] In the embodiment provided in this application, multiple micro-element information is aligned, text is deduplicated, and spliced ​​to obtain global narrative information containing the overall narrative content and framework of the novel to be adapted. The core elements of the global narrative information are extracted, and the filmable content in the novel is extracted to obtain macro-narrative information for the novel text to be adapted. This facilitates subsequent adaptation of the novel from a macro perspective based on the macro-narrative information, so as to preserve the overall narrative framework of the novel text to be adapted, ensure the integrity of the adapted target script, and thus improve the script quality of the target script.

[0032] In one exemplary embodiment provided in this application, text extraction includes alignment, deduplication, splicing, and extraction of core elements.

[0033] The text deduplication rule can be a later-chunk-override rule. The specific steps for deduplication are as follows: based on the unique internal key generated by the global Aegis (protection shield) internal key management module, and combined with the later-chunk-override rule, deduplication is performed. That is, when two overlapping text blocks contain the same chapter information, the information extracted from the text block that is later in the original text is adopted first to ensure the integrity of the information, solve the problem of duplicate information caused by overlapping blocks, and splice them into a preliminary and complete chapter element library to obtain global narrative information.

[0034] Macro-narrative synthesis (assembly + core element extraction): All chapter summary texts within the global narrative information are assembled, then progressively refined using a multi-level pre-trained language model, ultimately synthesizing a highly condensed "global outline," providing macro-narrative information specific to the novel text to be adapted. This "global outline" includes the core story outline, main character biographies, world-building, and other macro-narrative information, serving as the highest guiding principle and consistency constraint for all subsequent steps in scriptwriting. Alternatively, the multi-level pre-trained language model used for progressively refining and synthesizing the "global outline" can be replaced with a one-time long text summarization model.

[0035] Optionally, the target element information includes micro-level element information and macro-level narrative information. The micro-level element information includes summary information from multiple chapters. Based on the target element information, the summary is refined to obtain a refined narrative unit library, including: Based on the preset summary elements and macro-narrative information, the chapter summary information is refined and extracted to obtain refined narrative units; A refined narrative unit library is formed based on multiple refined narrative units.

[0036] In the embodiment provided in this application, chapter summary information is refined and extracted based on macro narrative information according to preset summary elements, so as to fully extract highly representative summary information, obtain refined narrative units, and form a refined narrative unit library. In this way, subsequent script writing only needs to focus on these highly representative summary information, which can shorten the script content while ensuring the integrity of the final target script, thereby improving the script writing efficiency and the script quality of the target script.

[0037] In an exemplary embodiment provided in this application, the specific steps of refinement extraction (summary refinement and scripting) are as follows: instruct a pre-trained language model to use the macro narrative information as a "global outline" to review and rewrite the chapter summary information (literary summary) including micro element information into a "filmable" scripted summary containing five elements such as scene, characters, actions, dialogue, and time, thereby obtaining refined narrative units, ensuring logical consistency with the global settings, and outputting a refined narrative unit library formed by multiple refined narrative units.

[0038] Optionally, the refined narrative unit library includes multiple refined narrative units; plot extraction is performed based on the refined narrative unit library to obtain the climax plot positions for the novel text to be adapted, including: Narrative tension is calculated based on refined narrative units to obtain the corresponding narrative tension score; A narrative tension curve is formed based on multiple narrative tension scores; Peak detection is performed based on the narrative tension curve to obtain the location of the climax plot for the novel text to be adapted.

[0039] In the embodiment provided in this application, firstly, the narrative tension score of each refined narrative unit is calculated to form a narrative tension curve. This curve is used to rhythmically arrange narrative elements such as "expectation and fulfillment," "conflict and resolution," and "unknown and revealed" in the novel text to be adapted, thus constructing an emotional and plot-driving force that runs through the work. Secondly, peak detection is performed based on the narrative tension curve to extract the core nodes that represent the most concentrated tension, the most intense conflict, and the strongest emotional impact in the narrative. This yields the climax plot positions for the novel text to be adapted, allowing subsequent scriptwriting to use these climax plot positions as key points. This enables the target script to form strong memorability and dissemination points, improving the script quality, attracting audience attention, increasing audience curiosity, and enhancing the commercial value of the target script.

[0040] Optionally, narrative tension is calculated based on refined narrative units to obtain the corresponding narrative tension score, including: The intensity of conflict, semantic mutation rate, and emotional entropy of the refined narrative units were calculated. Based on preset weight parameters, the narrative tension score of the refined narrative unit is obtained by weighting the conflict intensity, semantic mutation degree and emotional entropy.

[0041] In the embodiment provided in this application, the conflict intensity, semantic mutation degree, and emotional entropy of the refined narrative unit are calculated. This allows for the parameterization of narrative elements such as "expectation and fulfillment," "conflict and resolution," and "unknown and revealed" within the refined narrative unit. Furthermore, based on preset weight parameters, a weighted calculation is performed using conflict intensity, semantic mutation degree, and emotional entropy to obtain the narrative tension score of the refined narrative unit. This parameterization of the narrative tension of the refined narrative unit facilitates the understanding of the core nodes in the narrative—where tension is most concentrated, conflict is most intense, and emotional impact is strongest—based on the narrative tension curve formed by multiple narrative tension scores. This reduces the difficulty of extracting the climax plot points from the novel text to be adapted, thereby improving the script quality of the target script obtained by using the climax plot points as key points for scriptwriting.

[0042] In one exemplary embodiment provided in this application, each refined narrative unit (chapter) generates a narrative tension score, which is a weighted sum of three dimensions: conflict intensity, semantic mutation degree, and emotional entropy.

[0043] First, conflict intensity is an indicator that represents the degree of intensity of conflict between the main characters within a chapter. The calculation formula is: in, This represents the total number of valid tokens or words in the current chapter (refined narrative unit) after cleaning. It is used as the denominator for normalization to eliminate statistical bias in total frequency caused by varying chapter lengths (e.g., longer chapters naturally have more verbs), ensuring that the indicator reflects the "density" of conflict rather than its "total amount."

[0044] This represents the preset strength value of the i-th antagonistic verb. The system can preset a "conflict verb strength dictionary" to map verbs to values ​​in the range [0,1]. For example, the violent verbs "slaughter" and "destroy" are assigned a value of 0.9, the moderate conflict verbs "beat" and "roar" are assigned a value of 0.6, and the mild conflict verbs "argue" and "sarcasm" are assigned a value of 0.3.

[0045] This indicates the number of times the main entity and the opposing entity appear simultaneously within a preset sliding window (e.g., 50 tokens or the same sentence). This metric is used to quantify the "contact density" of the two sides in the core conflict.

[0046] These represent the weights of action description and character interaction, respectively. Preferably, these two coefficients can be obtained through regression training on sample scripts labeled with climax levels, or set based on empirical values ​​(e.g., ).

[0047] This approach first extracts key elements using Named Entity Recognition (NER) and Part-of-Speech Tagging (POS). Then, a pre-defined sentiment intensity dictionary assigns differentiated semantic intensity values ​​to different adversarial verbs (such as "defeat" and "counterattack") to distinguish the weight of different actions' contributions to the conflict. Simultaneously, the co-occurrence frequency of protagonist and adversary entities within the same semantic interval (such as a paragraph or sentence) is statistically analyzed to characterize the contact density between the core conflicting parties. Finally, chapter text length is introduced as a normalization factor to smooth the statistical results, thereby eliminating statistical bias caused by varying chapter lengths and accurately quantifying the intensity of conflict per unit text length.

[0048] Specifically, the intensity of the conflict The specific calculation and acquisition methods for each parameter in the calculation formula are as follows: (1) The calculation process of text length L includes: first, preprocessing the original text of the chapter to be processed to remove special symbols, HTML tags and meaningless stop words; then, using a word segmentation tool (such as Jieba segmentation) to segment the cleaned text; finally, counting the total number of valid tokens or valid characters after segmentation and using it as the L value. This step aims to eliminate the total frequency statistics deviation caused by the difference in the number of words in the chapters (such as long chapters and short chapters).

[0049] (2) Regarding the preset strength value Acquisition: The acquisition relies on a pre-built "conflict verb strength dictionary". Construction phase: Common antagonistic verbs from the novel (such as "kill", "hit", "glare", "taunt", etc.) are pre-collected, and each verb is assigned a strength value in the range [[0,1] based on sentiment analysis models or expert scoring (e.g., "destroy" is assigned a much higher value than "argue"). Calculation phase: Verbs within a chapter are traversed, and it is determined whether they are in the dictionary. If they are, their corresponding strength value is directly obtained through a hash lookup; if not, their strength value is defaulted to 0 or matched with the closest dictionary word based on semantic similarity.

[0050] (3) Co-occurrence frequency The calculation uses the "sliding window scanning method": a sliding window of a preset length is set (for example, the window size is 50 tokens, or a natural sentence with a "period" as the boundary).

[0051] Named Entity Recognition (NER) technology is used to identify the sets of "protagonist entities" and "rival entities" in the text. When a sliding window moves over the text, if the window contains at least one "protagonist entity" and at least one "rival entity" simultaneously, it is considered a valid co-occurrence, and a counter is incremented by 1. The total number of implementations for this chapter is then accumulated.

[0052] (4) Weighting coefficients and Determination: These two coefficients are hyperparameters used to balance the weight of "action intensity" and "character interactivity". Initial settings: These can be set based on empirical values ​​(e.g., setting...). =0.6, =0.4), focusing on action description. Dynamic adjustment: In a preferred embodiment, these two weights can be trained using a machine learning model. That is, using sample scripts labeled with "climax level" as ground values, the optimal weight combination is backfitted using linear regression or gradient descent algorithms.

[0053] Second, the Semantic Change Rate is an indicator that represents the intensity of plot twists between adjacent chapters. The calculation formula is: in, This represents a text vectorization function used to map variable-length text data into high-dimensional dense semantic vectors based on semantic encoding models (such as pre-trained language models based on the Transformer architecture). This represents the cosine similarity calculation function, used to measure the directional similarity between two vectors; Let represent the core plot text of chapter i (refined narrative unit) and chapter i-1 (refined narrative unit) after compression using the summarization model. Then, by calculating the cosine similarity of the summary vectors of adjacent chapters, the semantic similarity between them is obtained; subsequently, their complement (1 - similarity) is taken as the difference index. A larger difference indicates a stronger leap in the core events or logical direction between adjacent chapters, thus quantifying the intensity of plot twists. .

[0054] Specifically, semantic mutation degree The implementation process for each parameter in the calculation formula is as follows: (1) Chapter Summary Generation (data preprocessing): This is not the original text of the chapter, but rather compressed text processed by a generative summarization model. The specific steps include: inputting the complete text of chapter i into a pre-trained text summarization model (e.g., BART, T5, or other models based on the Transformer architecture). The model extracts key events, core character actions, and causal logic from the chapter using a self-attention mechanism, filtering out simple environmental descriptions and redundant dialogue, and generating a short text (e.g., a summary of 200-500 words) containing the core plot, denoted as [summary name missing]. The purpose of this approach is to eliminate the interference of rhetorical devices and environmental descriptions, ensuring that the subsequent comparison is based on "plot logic" rather than "word choice and sentence structure."

[0055] (2) Vectorization function Implementation: A deep learning-based sentence encoder is employed. The specific implementation steps include: using a pre-trained semantic encoding model (e.g., a pre-trained language model based on the Transformer architecture); and processing the summarized text. The input model is transformed into a token sequence. The output of the last hidden layer or the vector corresponding to the [CLS] token is extracted. Through mean pooling, the variable-length text is mapped into a dense real-valued vector of fixed dimensions (e.g., a 768-dimensional or 1024-dimensional vector). The effect of this implementation is that the vector can capture the deep semantic features of the text, rather than simple keyword matching.

[0056] (3) Calculation of cosine similarity: This function is used to calculate the cosine value of the angle between two vectors in a multidimensional space.

[0057] (4) Physical meaning of the final score S: The value range of the calculated result S is usually between [0,2] (in practical applications, it is mostly between 0 and 1). When S approaches 0, it means that the two chapter summaries are highly similar in semantics and the plot development is smooth or natural; when S approaches 1 (or even higher), it means that the two chapter summaries have huge semantic differences, which means that a sudden change in the scene, reversal or the intervention of a new core event has occurred, which is judged as "drastic plot twist".

[0058] Third, emotional entropy is an indicator that characterizes the complexity of emotional distribution within a chapter. The formula for calculating emotional entropy E is: E=-Σ( *log( )) in, This represents the frequency or probability distribution of the k-th emotion category (such as "anger", "sadness", "joy", etc., based on a pre-defined fine-grained emotion classification system) in the text of this chapter. In this way, by analyzing the words within a chapter using an emotion lexicon, and calculating the complexity and intertwining of emotions using the Shannon entropy formula, the higher the emotion entropy value, the more complex and entangled the emotional tension within the chapter.

[0059] Narrative Tension Score The calculation formula is: ,in, This represents the weighting coefficient of conflict intensity (C) in the narrative tension evaluation system. This represents the weighting coefficient of semantic abruptness (S) in the narrative tension evaluation system. This represents the weighting coefficient of emotional entropy (E) in the narrative tension evaluation system.

[0060] Specifically, to ensure the accuracy and generalization ability of the narrative tension evaluation model, the specific determination and calculation methods of each parameter in the formula for calculating the narrative tension score are as follows: (1) Weighting coefficients ( The method for determining the genre is as follows: A dynamic, adaptive configuration based on genre is adopted. The system pre-sets multiple weight templates and automatically calls one based on the novel's genre classification (obtained through metadata or identified by a text classification model). For example, for "fantasy or action" novels, the core appeal lies in combat, and the system automatically assigns weights based on these categories. (Intensity of conflict) carries higher weight; for "suspense / mystery" novels, the core appeal lies in plot twists, which the system automatically assigns. (Semantic mutation degree) has a higher weight; for "romance" novels, the core appeal lies in emotional turmoil, and the system automatically assigns... (Emotional entropy) has a higher weight.

[0061] (2) Weight normalization constraint: Before calculation, the weights are usually normalized, and the following constraints must be met: This ensures that the final calculated T value will not be distorted due to the overall scaling of the weights, thus guaranteeing the relative stability of the score.

[0062] (3) Final post-processing (numerical mapping) of the narrative tension score T: The T value calculated by the original formula is a dimensionless value. In order to facilitate users' intuitive understanding or cross-work horizontal comparison, the system will perform Min-Max normalization or Sigmoid mapping on T to map it to a standard interval (for example, the standard interval can be [0,10] points).

[0063] The formula for calculating the numerical mapping of T can be: in, This is the new narrative tension score obtained by numerically mapping T. and These are the maximum and minimum values ​​of tension theory obtained based on statistical analysis of large-scale historical corpora.

[0064] As shown above, the narrative tension scores of all refined narrative units (chaps) are connected to form a visual "narrative tension curve". The weight parameters of these three dimensions can be preset or dynamically adjusted according to the type of adaptation (such as suspense, romance) or market preferences. The peak detection algorithm automatically marks the candidate points of the "climax" to obtain the position of the climax plot for the novel text to be adapted. These candidate points will be used as "data suggestions" for the subsequent script design, so that the target script obtained by the decision can combine creative intuition and data science.

[0065] In one exemplary embodiment provided in this application, the dimensions for calculating the narrative tension score can be added, removed, or modified according to the adaptation target (such as comedy or tragedy). (For example, when the novel text to be adapted is a comedy, a "humor index" dimension is added to calculate the corresponding narrative tension score.) The method of this application is also applicable to other long-form narrative genres, such as film and television adaptations of historical documents and long-form reports, adaptations of animated series, and adaptations of motion comics.

[0066] Please see Figure 2 , Figure 2 This is a flowchart illustrating the process of determining the location of a climax scene in an exemplary embodiment of this application, as shown below. Figure 2 As shown, the process of determining the location of the climax plot is also the algorithm process of "Narrative Tension Curve Generation (NTC-PD)". It shows how the three dimensions of conflict intensity, semantic mutation degree and emotional entropy are integrated into a single point tension score through a weighted formula, and finally form a curve and peak detection.

[0067] Optionally, a method for generating scripts for novel adaptations further includes: Obtain market feedback data for the target script; The weighting parameters are adjusted based on market feedback data to obtain new weighting parameters, which are then used to generate scripts for new novel texts to be adapted.

[0068] In the embodiment provided in this application, the weighting parameters used to calculate the narrative tension score are adjusted based on market feedback data for the target script, resulting in new weighting parameters. This allows subsequent scriptwriting to adapt to the current market situation, thereby increasing the commercial value of the new target script generated from the new novel text to be adapted based on the new weighting parameters. The specific steps for generating a script from the new novel text to be adapted based on the new weighting parameters are consistent with the steps for generating a target script from the novel text to be adapted.

[0069] In an exemplary embodiment provided in this application, the specific steps for obtaining market feedback data for a target script are as follows: after the adapted work (such as a short drama) filmed according to the target script is launched, the system can receive and analyze user behavior data (such as episode payment rate, drop-off points, etc.) to obtain market feedback data.

[0070] Then, the weighting parameters are adaptively adjusted based on market feedback data: Based on real market feedback data, the system can automatically adjust the weighting parameters for calculating narrative tension. The strategy can be adopted from the DSDP algorithm (Dual-Scaling Direction Point) to make subsequent adaptation decisions more closely reflect real market preferences, forming a self-learning and self-optimizing intelligent closed loop.

[0071] Optionally, based on target element information, a refined narrative unit library, and the location of climax scenes, a structured checkpoint design is performed to obtain a target script for the novel text to be adapted, including: Obtain the current market demand parameters for short dramas, and find the adaptation templates that match the current market demand parameters from the preset script template library; Adapting the novel text to be adapted using the adaptation template and a refined narrative unit library yields the first adapted script. The location of the climax plot is determined as the adaptation checkpoint location. The first adapted script is planned according to the adaptation checkpoint location to obtain the first checkpoint script. The script at the first critical point is mapped onto a pre-defined three-act dramatic structure to obtain the target script for the novel text to be adapted.

[0072] In the embodiment provided in this application, firstly, based on the externally input current market demand parameters for short dramas, an adaptation template matching the current market demand parameters is dynamically loaded to guide the adaptation task. Following this adaptation template, adaptation is carried out based on a refined narrative unit library, and the climax plot points are determined as adaptation checkpoints for checkpoint planning, resulting in a first checkpoint script for the novel text to be adapted. This ensures that each episode of the first checkpoint script forms a strong memorable and communicative point, improving the script quality and commercial value of the first checkpoint script. Secondly, the first checkpoint script is mapped to a preset three-act dramatic structure to structure it into a three-act script that is easy to film, resulting in a target script for the novel text to be adapted. This improves the filming feasibility of the target script while also enhancing its script quality and commercial value. The current market demand parameters for short dramas include the total number of episodes of the target short drama. The three-act dramatic structure is a "setup-confrontation-ending" dramatic structure.

[0073] Optionally, the target element information includes macro-narrative information; based on the target element information, a refined narrative unit library, and the location of climax scenes, a structured checkpoint design is performed to obtain the target script for the novel text to be adapted, including: Based on macro-narrative information, important plot points are extracted from multiple text blocks to obtain a high-value material library; Obtain the current market demand parameters for short dramas, and find the adaptation templates that match the current market demand parameters from the preset script template library; Adapting the novel text to be adapted using the adaptation template and a high-value resource library, a second adapted script is obtained; The location of the climax scene is determined as the adaptation checkpoint location. The new adapted script is planned according to the adaptation checkpoint location to obtain the second checkpoint script. The second checkpoint script is mapped onto a pre-defined three-act dramatic structure to obtain the target script for the novel text to be adapted.

[0074] In the embodiment provided in this application, firstly, the adaptation is carried out according to an adaptation template that matches the current market demand parameters for short dramas. Adaptation is based on a high-value material library extracted from multiple text blocks according to macro-narrative information, and the climax plot points are identified as adaptation checkpoints for checkpoint planning. This allows for checkpoint adaptation of the script from a global perspective, resulting in a second checkpoint script for the novel text to be adapted. This second checkpoint script, while considering the overall plot, forms strong memorable and communicative points within the single episode, thereby improving the script quality and commercial value of the second checkpoint script. Secondly, the second checkpoint script is mapped to a preset three-act dramatic structure, structuring it into a three-act script that is easy to film, resulting in a target script for the novel text to be adapted. This improves the filming feasibility of the target script while also enhancing its script quality and commercial value.

[0075] In an exemplary embodiment provided in this application, the process of adapting based on macro-narrative information is called high-value material exploration (Enrichment): a pre-trained language model is instructed to take the macro-narrative information, which is the "global outline," as high-level knowledge input, return to the original novel text block, and relocate and extract important plots such as the "protagonist's golden quotes" and "famous scenes" that are crucial to the core plot and character arcs from a global perspective. In this way, this "ruminative" exploration ensures the global optimality of the selected materials and outputs a high-value material library.

[0076] In another exemplary embodiment provided in this application, a structured checkpoint design is performed based on target element information, a refined narrative unit library, and the location of climax scenes to obtain a target script for the novel text to be adapted. The specific steps are as follows: First, based on the input parameters of the current market demand for short dramas (commercial parameters, such as the total number of episodes of the target short drama), the most suitable adaptation template is dynamically selected from the preset strategy library (script template library). (For example, when the total number of episodes of the target short drama is 80-100, the adaptation template of "main plot condensation and subplot merging" is selected. For 20-40 episodes, the template framework of "extracting core events and fast-paced progression" is selected.)

[0077] Subsequently, a pre-trained language model is guided to make high-level strategic decisions. Based on these decisions, the refined narrative unit library is adapted according to the adaptation template, and the climax plot points are identified as adaptation checkpoints. These checkpoints are then incorporated into the adaptation process for checkpoint planning, resulting in the first checkpoint script. During checkpoint planning, the adaptation checkpoints are designated as key plot turning points (especially at the boundaries between acts), and are set as "commercial checkpoints" that can stimulate viewers' desire to continue watching (such as the "first checkpoint" at the end of the first act, the "second checkpoint" at the end of the second act, etc.). The rationale for these designs is explained in detail (introducing the climax plot at that location).

[0078] The script for the first critical point is intelligently mapped to the classic three-act dramatic structure of "establishment-confrontation-ending," and the core plots of each act are identified to obtain the target script for the novel text to be adapted.

[0079] Please see Figure 3 , Figure 3 This is a schematic flowchart illustrating the script generation method for novel adaptation provided in an exemplary embodiment of this application. Figure 3 The diagram illustrates the specific steps involved in automatically adapting the original novel text to be adapted into a target script and outputting a report using the script generation method for novel adaptation of this application. The steps are as follows: Step 1: Indiscriminate Information Decomposition (Sliding Window Chunking) S1: Receive the complete novel text to be adapted, execute the Indiscriminate Sliding Window Chunking algorithm to generate multiple text chunks with overlapping areas.

[0080] Step 2: Parallel Information Abstraction (L1-L3 Hierarchical Processing) S2: L1-Micro Element Extraction: Extract chapter information based on text blocks to obtain micro element information for the text blocks; L2 & L3-Macro Narrative Synthesis: Extract text based on multiple micro element information to obtain macro narrative information for the novel text to be adapted; Based on micro element information and macro narrative information, form target element information.

[0081] Step 3: Global Knowledge Guided Refinement & Enrichment (GKGR&E) (L4) S3: This step utilizes the macro-narrative information generated in L3 as a "global outline" as a "global perspective" to conduct a "chief screenwriter-level" in-depth processing of the novel's content. It includes two concurrent sub-tasks: Sub-task A: Summary Refinement and Scripting (Refinement), resulting in a refined narrative unit library; Sub-task B: High-Value Material Exploration (Enrichment), resulting in a high-value material library.

[0082] Step 4 (Optional Enhancement Module): Narrative Tension Curve-Based Peak Detection Optimization S4: Utilizes an optional Narrative Tension Curve - Peak Detection (NTC-PD) module to quantitatively calculate the refined narrative unit library of L4, generating a "narrative tension score" for each chapter (unit).

[0083] Step 5: AI-driven business structure planning (L5, DSDP algorithm) S5: The decision-making center for transforming literary materials into commercial products: Execute the "Dynamic Strategy-Driven Planning (DSDP) algorithm" to design structured checkpoints based on target element information, a refined narrative unit library, and the location of climax plot points, and obtain the target script for the novel text to be adapted.

[0084] Step Six (Optional Enhancement Module): Closed-Loop Optimization Based on Market Feedback: To continuously improve the effectiveness of adaptation strategies, support the generation of A / B test schemes: The system can generate multiple different climax plot points based on the identified locations (for example, Scheme A sets the climax at Chapter 120, and Scheme B sets it at Chapter 135), to generate target scripts with multiple climax schemes.

[0085] In one exemplary embodiment provided in this application, in addition to the method "steps" process, this application also includes three core management modules that run throughout: core management module A, core management module B, and core management module C, which ensure the robustness, consistency, and flexibility of the entire method "steps".

[0086] Core Management Module A: Adaptive API Call Module. This module manages communication between the system and all pre-trained language models, ensuring their robustness in high-concurrency and unstable network environments, including concurrency control, intelligent retry strategies, and response cleaning and validation. This adaptive API call module includes: Concurrency control: Using a semaphore mechanism, the number of concurrent requests sent to the API can be precisely controlled according to a preset concurrency limit (such as ip_analysis_concurrency_limit) to prevent client or server overload caused by too many instantaneous requests.

[0087] Intelligent retry strategy: When encountering API errors (especially HTTP 429 rate limiting errors), an exponential backoff with jitter retry algorithm is automatically triggered. The waiting time is dynamically calculated by a formula, effectively avoiding the "retry storm" problem. The formula for calculating the waiting time can be: In this context, base_delay, attempt, and random_jitter represent the base delay, the current number of retries, and the random jitter value, respectively. Those skilled in the art can adjust these parameters according to the actual network environment and the limitations of the API service provider.

[0088] Response cleaning and validation: Upon receiving the text returned by the model, the built-in JSON cleaner and validator (clean_and_parse_json) automatically executes, including removing code block markers (such as ```json), fixing common trailing comma errors, and attempting parsing. If parsing fails, the call is considered a failure and enters a retry process, thus ensuring the validity of the structured data entering the system from the source.

[0089] Therefore, the adaptive API call management steps corresponding to the above adaptive API call module include: a) controlling concurrent requests through a semaphore mechanism; b) automatically triggering retry logic including exponential backoff and random jitter algorithms when a preset API error is detected; c) automatically cleaning and validating the structured data format of the returned content after receiving the response.

[0090] Core Management Module B: Global Unified Identifier (Key) Management Module (hereinafter referred to as the Aegis internal key management module). This module is the cornerstone for ensuring absolute data consistency during multi-model, multi-stage collaboration. Through atomic key generation and unified normalization functions, it fundamentally eliminates data association errors. The core of the Global Unified Identifier (Key) Management Module lies in: Atomized bond generation: When extracting micro-features in (L1), force the model to return atomized chapter_number (integer) and chapter_title (string).

[0091] Unified Normalization Function: The system uses a unified normalization function (Utils.normalize_chapter_key), which converts chapter names of any format (such as "Chapter 1 Preface", "Chapter 1: Preface") into a unique and predictable internal key by removing spaces, converting to lowercase, and converting full-width characters to half-width characters.

[0092] Global Application: This internal key is used as a unique identifier for all structured data (summary library, quote library, famous scene library). All subsequent data merging, updating, and querying operations (such as enriching to obtain a high-value material library) are based on this key, fundamentally eliminating KeyErrors caused by chapter name mismatches and ensuring lossless data flow and accurate alignment throughout the script generation process.

[0093] Therefore, the core management module described above can execute a global internal key management strategy to ensure data consistency. This strategy includes: a) extracting atomic chapter identifier information in the global knowledge synthesis step; b) using a preset normalization rule to process the chapter identifier information into unique, standardized internal keys; c) aligning and deduplicating duplicate micro-narrative elements generated by block processing based on internal keys before synthesizing the global knowledge base.

[0094] Core Management Module C: Dynamic Prompt and Model Selection Module. This module empowers the system to dynamically adjust its core behaviors based on external input, including dynamic prompt selection and hierarchical model selection, to optimize cost and benefit. This dynamic prompt and model selection module includes: Dynamic Prompt Selection: In step four (DSDP Planning), the system dynamically selects the most suitable prompt from a pre-set prompt template library (script template library) based on the input business parameter of the current short drama market demand (e.g., total_episodes, total number of episodes). For example, if the target number of episodes is 80, the system automatically loads 5_1_Structure Planning_80 episodes.txt as the planning instruction, ensuring that the AI's decision aligns with the business objectives from the outset.

[0095] Layered model selection: This allows for the configuration of pre-trained language models with varying capabilities and costs for tasks of different complexities. For example, a lightweight model (such as gemini-2.5-flash) with faster speed and lower cost can be configured for the initial summary extraction at the L1 layer, while the flagship model with the strongest logical capabilities (such as a model with extremely long context windows and high inference capabilities) can be configured for the synthesis of the macro-narrative information—the "global outline"—at the L3 layer and the final planning at the L5 layer, thus optimizing cost and benefit.

[0096] Please see Figure 4 , Figure 4 for Figure 3 The corresponding further detailed process, such as Figure 4 As shown, the further detailed process of the script generation method for novel adaptation provided by the application is as follows: S100: Perform indifferent information decomposition The system receives a novel (e.g., a 3 million-word TXT file). For example, the chunk size (chunk_size) is set to 5000 characters, and the overlap size (overlap_size) is set to 500 characters. The system executes a sliding window chunking algorithm to divide the file into multiple overlapping text chunks. These values ​​are exemplary and can be adjusted based on factors such as the average paragraph length of the novel text and the size of the context window processed by the language model.

[0097] S200: Parallelized Abstraction and Global Knowledge Synthesis S210-S220 (L1 and Alignment): All text blocks are submitted in parallel to a pre-trained language model (e.g., any commercial or open-source model that supports long text processing, such as the GPT, Gemini, or Llama series) via the adaptive API call module. Each API request instructs the model to extract chapter information within a block and return it in JSON format. The system collects all returned results and performs alignment and deduplication through the global Aegis internal key management module, concatenating them into a preliminary chapter feature library containing micro-feature information of multiple text blocks.

[0098] S230 (L2 & L3 Synthesis): The system extracts summaries from all chapters, splices them together, and submits them to a more powerful pre-trained language model for progressive summarization, ultimately generating a "global outline" of about 2,000 words, thus obtaining macro-narrative information.

[0099] S300: Perform Global Knowledge-Guided Refinement and Enrichment (GKGR&E) The system executes two tasks concurrently: S310 (Refinement): Traverse the initial chapter element library. For each summary, combine it with the "Global Outline". Instruct the pre-trained language model: "Based on the provided "Global Outline", rewrite the following literary summary into a scripted summary containing elements such as scenes, characters, and actions, and correct any logic that does not conform to the global setting", thus obtaining the refined narrative unit library.

[0100] S320 (Enrichment): For each chapter, instruct another pre-trained language model: "Based on the provided 'global outline,' extract the 'golden sentences' and 'famous scenes' from the following original text blocks that best reflect the protagonist's arc or drive the core plot, to obtain a high-value material library."

[0101] S400: Performing Narrative Tension Analysis (NTC-PD) It can perform calculations on the refined narrative unit library output by S310. For example, it can identify the protagonist and antagonist through NER... and calculate the intensity of conflict. Semantic mutation degree is obtained through Sentence-BERT... ; Calculate emotional entropy using an emotional dictionary... Finally, the narrative tension score is calculated based on preset weights. This generates a narrative tension curve, a visual chart, and uses a peak detection algorithm to select climax candidate points from the narrative tension curve to obtain the location of the climax plot.

[0102] S500: Execution of Business Plan (DSDP) The system loads the corresponding script template through dynamic prompts and the model selection module based on external parameters (such as "target_episodes":80). It submits the macro narrative information ("global outline"), the refined narrative unit library, the high-value material library, and (if applicable) the climax plot positions generated by S400 to the pre-trained language model, instructing it to perform three-act structure division and commercial block design to obtain the target script.

[0103] S600: Report Generation The system integrates all outputs, including macro-narrative information (the "overall outline"), structural planning, refined episode summaries, high-value materials, and NTC-PD visualization charts and target scripts, to generate a structured script planning report.

[0104] In summary, the script generation method for novel adaptation proposed in this application can achieve systematization and innovation, efficiency and robustness, adaptation quality and logical consistency, scientific decision-making and commercial value, and system robustness and flexibility.

[0105] The systematization and innovation are specifically reflected in the following: The complete script generation process designed in this application, from deconstruction, global synthesis, guided refinement to planning optimization, is the first systematic solution in the industry that can effectively solve the problems of efficiency, scalability and logical consistency in the adaptation of long novels.

[0106] Efficiency and robustness are specifically reflected in the unique indiscriminate block division and parallel processing mechanism, which can efficiently and stably process source text with any non-standard format, compressing weeks of manual adaptation work into hours, improving efficiency by more than 95%.

[0107] The quality of the adaptation and its logical consistency are specifically reflected in the combination of the "global outline" and the GKGR&E algorithm, which fundamentally ensures that the adapted content has no logical flaws at the macro and micro levels, and achieves the "global optimal" material decision.

[0108] The scientific decision-making and commercial value are specifically reflected in the innovative NTC-PD module, which for the first time quantifies the abstract "plot highlights" into calculable indicators, achieving a combination of data-driven approaches and AI creativity. The optional closed-loop optimization module further incorporates market feedback, enabling the system to self-evolve and significantly improving the accuracy of business checkpoint setting and market success rate.

[0109] The system's robustness and flexibility are specifically reflected in the fact that, through three core management modules (adaptive API calls, Aegis key management, and dynamic strategies), the system achieves stable operation under high concurrency, absolute consistency of multi-stage data flow, and flexible adaptation to different business needs, thus possessing extremely high commercial application value.

[0110] Please see Figure 5 , Figure 5 An exemplary embodiment of this application illustrates a script generation system for novel adaptation, such as... Figure 5 As shown, this application provides a script generation system 500 for novel adaptation, comprising: Module 501 is used to acquire the text of the novel to be adapted. The element extraction module 502 is used to extract elements from the novel text to be adapted and obtain target element information. Abstract Refinement Module 503 is used to refine abstracts based on target element information to obtain a refined narrative unit library; The plot extraction module 504 is used to extract plots based on the refined narrative unit library to obtain the climax plot positions for the novel text to be adapted. The script generation module 505 is used to design structured checkpoints based on target element information, a refined narrative unit library, and the location of climax plot points, to obtain the target script for the novel text to be adapted.

[0111] The script generation system 500 for novel adaptation provided in this application first uses an element extraction module 502 to extract elements from the novel text to be adapted obtained by an acquisition module 501, obtaining target element information. Then, a summary refinement module 503 refines the summary based on the target element information to preserve the overall narrative logic of the novel text, resulting in a refined narrative unit library. Next, a plot extraction module 504 extracts plots based on the refined narrative unit library, obtaining the climax plot positions for the novel text. Finally, a script generation module 505 designs structured plot points based on the target element information, the refined narrative unit library, and the climax plot positions, resulting in a target script for the novel text, thus achieving automated adaptation from novel to script. In this way, through element extraction and summary refinement, the overall narrative logic of the novel text to be adapted can be preserved, ensuring that the automatically adapted target script maintains logical consistency throughout the plot, thereby improving the script quality. At the same time, because the climax plot points were designed during the script generation process, each episode of the target script can form strong memorable and shareable points, which can attract the audience's attention, increase the audience's curiosity, and thus improve the commercial value of the target script.

[0112] Optionally, the element extraction module 502 is specifically used for: The adapted novel text is deconstructed to generate multiple text blocks with overlapping areas; Chapter information is extracted based on text blocks to obtain micro-level element information for the text blocks; Text extraction is performed based on multiple micro-level information elements to obtain macro-level narrative information for the novel text to be adapted. Target element information is formed based on micro-level element information and macro-level narrative information.

[0113] Optionally, the element extraction module 502 is specifically used for: Align multiple micro-level information elements; Using preset text deduplication rules, text deduplication is performed based on the aligned information of multiple micro-elements to obtain multiple target micro-element units. By piecing together multiple target micro-element units, global narrative information is obtained; By extracting the core elements from the overall narrative information, we obtain macro-narrative information specific to the novel text to be adapted.

[0114] Optionally, the target element information includes micro-level element information and macro-level narrative information, with the micro-level element information including multiple chapter summary information; the summary refinement module 503 is used for: Based on the preset summary elements and macro-narrative information, the chapter summary information is refined and extracted to obtain refined narrative units; A refined narrative unit library is formed based on multiple refined narrative units.

[0115] Optionally, the refined narrative unit library includes multiple refined narrative units; the plot extraction module 504 is specifically used for: Narrative tension is calculated based on refined narrative units to obtain the corresponding narrative tension score; A narrative tension curve is formed based on multiple narrative tension scores; Peak detection is performed based on the narrative tension curve to obtain the location of the climax plot for the novel text to be adapted.

[0116] Optionally, the plot extraction module 504 is specifically used for: The intensity of conflict, semantic mutation rate, and emotional entropy of the refined narrative units were calculated. Based on preset weight parameters, the narrative tension score of the refined narrative unit is obtained by weighting the conflict intensity, semantic mutation degree and emotional entropy.

[0117] Optionally, a script generation system for novel adaptations also includes an adjustment module, which is specifically used for: Obtain market feedback data for the target script; The weighting parameters are adjusted based on market feedback data to obtain new weighting parameters, which are then used to generate scripts for new novel texts to be adapted.

[0118] Optionally, the script generation module 505 is specifically used for: Obtain the current market demand parameters for short dramas, and find the adaptation templates that match the current market demand parameters from the preset script template library; Adapting the novel text to be adapted using the adaptation template and a refined narrative unit library yields the first adapted script. The location of the climax plot is determined as the adaptation checkpoint location. The first adapted script is planned according to the adaptation checkpoint location to obtain the first checkpoint script. The script at the first critical point is mapped onto a pre-defined three-act dramatic structure to obtain the target script for the novel text to be adapted.

[0119] Optionally, the target element information includes macro-narrative information; the script generation module 505 is specifically used for: Based on macro-narrative information, important plot points are extracted from multiple text blocks to obtain a high-value material library; Obtain the current market demand parameters for short dramas, and find the adaptation templates that match the current market demand parameters from the preset script template library; Adapting the novel text to be adapted using the adaptation template and a high-value resource library, a second adapted script is obtained; The location of the climax scene is determined as the adaptation checkpoint location. The new adapted script is planned according to the adaptation checkpoint location to obtain the second checkpoint script. The second checkpoint script is mapped onto a pre-defined three-act dramatic structure to obtain the target script for the novel text to be adapted.

[0120] It should be noted that the script generation system for novel adaptation provided in the above embodiments and the script generation method for novel adaptation provided in the above embodiments belong to the same concept. The specific ways in which each module and unit performs its operations have been described in detail in the method embodiments, and will not be repeated here. In practical applications, the script generation system for novel adaptation provided in the above embodiments can be assigned to different functional modules as needed, that is, the internal structure of the system can be divided into different functional modules to complete all or part of the functions described above, and this is not a limitation here.

[0121] A computing device according to an embodiment of this application includes a memory, a processor, and a program stored in the memory and running on the processor. When the processor executes the program, it implements some or all of the steps of the above-described method for generating a script for novel adaptation.

[0122] The computing device can be a computer, and the corresponding program is computer software. The parameters and steps of the computing device described above can be referred to the parameters and steps in the embodiment of the script generation method for novel adaptation described above, and will not be repeated here.

[0123] This application embodiment provides a computer-readable storage medium storing instructions that, when executed, perform the steps of the script generation method for novel adaptation described above.

[0124] The computer-readable storage medium may be a transient computer-readable storage medium or a non-transitory computer-readable storage medium.

[0125] The technical solutions of this disclosure can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes one or more instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of this disclosure. The aforementioned computer-readable storage medium can be a non-transitory computer-readable storage medium, including: USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks, and other media capable of storing program code; it can also be a transient computer-readable storage medium.

[0126] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this application. Each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0127] Those skilled in the art will recognize that this application can be implemented as a system, method, or computer program product. Therefore, this disclosure can be implemented in the following forms: it can be entirely hardware, entirely software (including firmware, resident software, microcode, etc.), or a combination of hardware and software, generally referred to herein as a "module" or "system." Furthermore, in some embodiments, this application can also be implemented as a computer program product contained in one or more computer-readable media, which contains computer-readable program code. Computer-readable storage media can be, for example, but not limited to—electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof.

[0128] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0129] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for generating scripts for novel adaptations, characterized in that, include: Obtain the text of the novel to be adapted; Based on the novel text to be adapted, the target element information is obtained by extracting elements. Based on the target element information, a refined narrative unit library is obtained by summarizing and refining the information. Based on the refined narrative unit library, plot extraction is performed to obtain the climax plot positions for the novel text to be adapted. Based on the target element information, the refined narrative unit library, and the location of the climax plot, a structured checkpoint design is performed to obtain the target script for the novel text to be adapted.

2. The method according to claim 1, characterized in that, The process of extracting elements from the novel text to be adapted to obtain target element information includes: The novel text to be adapted is deconstructed to generate multiple text blocks with overlapping areas; Chapter information is extracted based on the text block to obtain micro-element information for the text block; Text extraction is performed based on multiple micro-element information to obtain macro-narrative information for the novel text to be adapted. Based on the micro-level element information and the macro-level narrative information, target element information is formed.

3. The method according to claim 2, characterized in that, The text extraction based on multiple micro-element information yields macro-narrative information for the novel text to be adapted, including: Align the information of multiple micro-elements; Using a preset text deduplication rule, text deduplication is performed based on the aligned micro-element information to obtain multiple target micro-element units; By splicing together multiple target micro-element units, global narrative information is obtained; The core elements of the global narrative information are extracted to obtain macro-narrative information for the novel text to be adapted.

4. The method according to claim 1, characterized in that, The target element information includes micro-level element information and macro-level narrative information. The micro-level element information includes multiple chapter summary information. The process of refining the summary based on the target element information to obtain a refined narrative unit library includes: Based on the preset summary elements and the macro narrative information, the chapter summary information is refined and extracted to obtain refined narrative units; A refined narrative unit library is formed based on multiple of the aforementioned refined narrative units.

5. The method according to claim 1, characterized in that, The refined narrative unit library includes multiple refined narrative units; the process of extracting plot points based on the refined narrative unit library to obtain the climax plot positions for the novel text to be adapted includes: Narrative tension is calculated based on the refined narrative units to obtain the corresponding narrative tension score; A narrative tension curve is formed based on multiple narrative tension scores; Peak detection is performed based on the narrative tension curve to obtain the location of the climax plot for the novel text to be adapted.

6. The method according to claim 5, characterized in that, The process of calculating narrative tension based on the refined narrative units to obtain the corresponding narrative tension score includes: The conflict intensity, semantic mutation degree, and emotional entropy of the refined narrative unit were calculated. The narrative tension score of the refined narrative unit is obtained by weighting the conflict intensity, semantic mutation degree and emotional entropy according to the preset weight parameters.

7. The method according to claim 6, characterized in that, The method further includes: Obtain market feedback data for the target script; The weight parameters are adjusted based on the market feedback data to obtain new weight parameters, which are then used to generate scripts for new novel texts to be adapted.

8. The method according to any one of claims 1 to 7, characterized in that, The structured timing design based on the target element information, the refined narrative unit library, and the location of the climax plot yields a target script for the novel text to be adapted, including: Obtain the current market demand parameters for short dramas, and find an adaptation template that matches the current market demand parameters from the preset script template library; Adapting the novel text to be adapted using the adaptation template and the refined narrative unit library yields a first adapted script. The location of the climax plot is determined as the adaptation checkpoint location, and the first adapted script is planned according to the adaptation checkpoint location to obtain the first checkpoint script; The first script is mapped to a preset three-act dramatic structure to obtain the target script for the novel text to be adapted.

9. The method according to any one of claims 1 to 7, characterized in that, The target element information includes macro-narrative information; the structured timing design based on the target element information, the refined narrative unit library, and the climax plot positions to obtain the target script for the novel text to be adapted includes: Based on the macro-narrative information, important plot points are extracted from multiple text blocks to obtain a high-value material library; Obtain the current market demand parameters for short dramas, and find an adaptation template that matches the current market demand parameters from the preset script template library; Based on the high-value material library, the adaptation template is used to obtain a second adapted script for the novel text to be adapted. The location of the climax plot is determined as the adaptation checkpoint location. The new adapted script is planned according to the adaptation checkpoint location to obtain the second checkpoint script. The second checkpoint script is mapped onto a preset three-act dramatic structure to obtain the target script for the novel text to be adapted.

10. A script generation system for novel adaptation, characterized in that, include: The acquisition module is used to acquire the text of the novel to be adapted. The element extraction module is used to extract elements based on the novel text to be adapted, and obtain target element information. The summary refinement module is used to refine the summary based on the target element information to obtain a refined narrative unit library; The plot extraction module is used to extract plots based on the refined narrative unit library to obtain the climax plot positions for the novel text to be adapted. The script generation module is used to design structured checkpoints based on the target element information, the refined narrative unit library, and the climax plot positions to obtain a target script for the novel text to be adapted.

Images