Novel conflict modeling method and system based on dynamic binary tree and feature driving

Abstract

The invention discloses a novel conflict modeling method and system based on a dynamic binary tree and feature driving, and belongs to the technical field of natural language processing and literature creation. According to the method, an S-C-E triple structure is defined, a recursive conflict tree model is constructed, dynamic generation of conflict nodes is driven by feature vectors, and quantitative management of conflict intensity, narrative tension and event density is realized through an emotional intensity equation, logic reliability constraint, a suspension entropy equation and a rhythm control function. The system comprises a structure definition module, a feature engine module, a neural network execution unit module and the like. Therefore, through a dynamic binary tree recursive structure and a feature vector driving mechanism, the technical defects of existing novel conflict modeling are overcome, hierarchical generation of conflict events, quantitative constraint of a causal relationship, scientific regulation and control of emotional intensity and dynamic management of a narrative rhythm are realized, and the method is suitable for outline construction of various virtual literature works.

Description

Technical Field

[0001] This application relates to the technical fields of natural language processing and literary creation, and in particular to a novel conflict modeling method and system based on dynamic binary trees and feature-driven methods. Background Technology

[0002] In traditional novel writing, conflict construction is one of the core elements determining a story's appeal. Currently, existing technologies have the following main shortcomings:

[0003] 1. Conflict building lacks a systematic recursive framework.

[0004] Traditional methods often rely on creators' experience to manually design conflicts, employing linear or simple branching structures, making it difficult to form hierarchical conflict networks. For example, common binary tree conflict models in existing technologies are mostly static symmetrical structures (such as simple left-right branching), which can only achieve basic conflict layering and cannot dynamically expand conflict nodes according to story logic, resulting in a single conflict level that is difficult to support complex narrative needs.

[0005] 2. Insufficient coherence in causal logic

[0006] Existing conflict modeling methods lack a quantitative description of the causal relationships between events, and the dependencies between conflict nodes are determined solely by experience, which can easily lead to logical gaps. For example, traditional models cannot guarantee the strength of the dependence of subsequent conflicts on preceding conflicts, resulting in a lack of rationality in the story's development. This is especially true in complex genres such as fantasy and martial arts, where it is difficult to meet the needs of conflict evolution under the constraints of cultivation systems, power relationships, and other rules.

[0007] 3. The control over emotional intensity and narrative rhythm is rough.

[0008] Current technologies lack scientific models for quantifying emotions, and adjustments to conflict intensity rely on subjective judgment, failing to dynamically adjust according to the story's progress. For example, traditional methods lack features such as I... e The emotional intensity equation, = tanh(ω·c+θ·t+κ), is insufficient for precise control by combining conflict complexity (c), event location (t), and genre characteristics (κ); furthermore, narrative pacing management lacks mathematical model support and cannot be achieved through T... m The function dynamically adjusts the event density, causing the story's pace to fluctuate, which affects the reader's experience.

[0009] 4. Lack of suspense entropy management

[0010] Current conflict modeling does not incorporate information entropy theory for quantifiable management of suspense; suspense setup relies on the creator's intuition, making it difficult to guarantee the sustainability of narrative tension. For example, traditional methods cannot utilize H... s =H0·(1-∑p i ·logpi The equation η+1 calculates the probability distribution of unresolved conflicts, resulting in insufficient suspense or excessive accumulation of key plot points, thus disrupting the balance of the story's pacing.

[0011] 5. Lack of feature-driven mechanisms

[0012] Current technologies fail to abstract core story elements (such as causal relationships, character emotions, and worldview rules) into feature vectors to drive conflict generation, resulting in a lack of interpretability and computability in conflict node generation. For example, traditional models do not employ β=(β1,β2,…,β…) m ) and γ=(γ1,γ2,…,γ k The feature vectors are fused with causal chain weights, emotional transfer probabilities, and worldview parameters, resulting in the conflict generation process being unable to be deeply coupled with the underlying logic of the story.

[0013] In summary, existing novel conflict modeling techniques have significant shortcomings in terms of systematicity, logical coherence, emotional control, and feature-driven aspects. There is an urgent need for a conflict modeling method based on dynamic recursive structure and quantitative feature-driven approach to meet the needs of automated outline generation in complex literary creation. Summary of the Invention

[0014] This application aims to at least partially address one of the technical problems in the related art.

[0015] Therefore, one objective of this application is to provide a novel conflict modeling method and system based on dynamic binary trees and feature-driven approaches. By using the recursive structure of dynamic binary trees and the feature vector-driven mechanism, this method solves the technical defects of existing novel conflict modeling, realizes the hierarchical generation of conflict events, quantitative constraints on causal relationships, scientific regulation of emotional intensity, and dynamic management of narrative rhythm, and is applicable to the outline construction of various fictional literary works.

[0016] To achieve the above objectives, the first aspect of this application proposes a novel conflict modeling method based on dynamic binary trees and feature-driven methods, comprising the following steps:

[0017] S1. Define the story triplet structure M = SCE, where S is the initiating conflict event, E is the ending conflict event, and C is the core conflict event;

[0018] S2. Construct a recursive conflict tree model:

[0019] Basic structure: When N=1, M1=SCE;

[0020] Recursive structure (N≥2): Where F m (β) represents the pre-conflict subforum, G n (γ) represents the post-conflict subforest, β = (β1, β2, ..., β)m ) and γ = (γ1, γ2, …, γ k ) are respectively the pre - and post - feature vectors containing the story causality strength index, the character emotional state transition probability, and the worldview rule constraint parameters;

[0021] S3. Dynamically create conflict nodes through the branch generation function:

[0022] Left - branch node: L n (α) = σ(Ψ·α + b), where α = (S, C, β), Ψ is the conflict weight matrix, b is the emotional bias term, and σ is the non - linear activation function;

[0023] Right - branch node: R n (δ) = φ(Θ·δ + d), where δ = (C, E, γ), Θ is the conflict weight matrix, d is the emotional bias term, and φ is the non - linear activation function;

[0024] S4. Apply the emotional intensity equation to control the conflict intensity: I e = tanh(ω·c + θ·t + κ), where c is the number of branches, t is the event position parameter (0 < S → E < 1), ω and θ are the emotional weight coefficients, and κ is the genre adjustment factor;

[0025] S5. Execute the logical reliability constraint: To ensure the causal coherence of conflict events.

[0026] According to the novel conflict modeling method and system based on dynamic binary tree and feature - driven of the embodiment of the present application, through the dynamic binary tree recursive structure and the feature vector driving mechanism, it solves the technical defects of the existing novel conflict modeling, realizes the hierarchical generation of conflict events, the quantitative constraint of causal relationships, the scientific regulation of emotional intensity, and the dynamic management of narrative rhythm, and is applicable to the outline construction of various fictional literary works.

[0027] In addition, the novel conflict modeling method and system based on dynamic binary tree and feature - driven proposed above in the present application may also have the following additional technical features:

[0028] In an embodiment of the present application, the construction of the pre - conflict sub - forest satisfies:

[0029]

[0030] The construction of the post - conflict sub - forest satisfies:

[0031]

[0032] In an embodiment of the present application, the generation of the feature vectors β and γ includes:

[0033] Based on the causal chain weights of the initial conflict S and the core conflict C, the influence factor of historical events on the current conflict is calculated.

[0034] Extract the transition probability matrix of the character's emotional state in the evolution of the conflict;

[0035] It incorporates the rules and constraints of the novel's worldview (such as cultivation system and power relationships).

[0036] In one embodiment of this application, a depth-first traversal order is used to generate a conflict sequence, where the number of layers N and the number of nodes satisfy the following:

[0037] Number of new nodes per layer = 2 N-2 (N≥2);

[0038] Total number of nodes = 2 N -2 (excluding S, C, and E; 0 when N = 1).

[0039] In one embodiment of this application, narrative tension is managed through a suspense entropy equation:

[0040]

[0041] Where H0 is the basic suspense value, p i Let η be the probability distribution of unresolved conflicts, and η be the random perturbation factor.

[0042] In one embodiment of this application, a rhythm control function is used to adjust the event density:

[0043]

[0044] Where β is the attenuation coefficient, λ is the fluctuation amplitude, and rand(-0.2,0.2) is a random number within the interval.

[0045] In one embodiment of this application, a conflict modeling system includes:

[0046] Structure definition module: Configured to create SCE triplet structure and store recursive rule base (conflict tree construction logic for N=1 and N≥2);

[0047] Feature Engine: Dynamically generates β and γ feature vectors for verification. and The causal constraints are determined, and the event impact factor is calculated based on the correlation between S and C;

[0048] Neural network execution unit: built-in L n (α), R n(δ) The generating function performs a linear transformation through the weight matrix Ψ / Θ and the bias term b / d, and generates left / right branch conflict nodes through the σ / φ activation function (either ReLU or Sigmoid).

[0049] Entropy controller: Real-time calculation of H s The value is adjusted based on the suspense entropy feedback to maintain narrative tension;

[0050] Rhythm adjustment module: based on T n The function dynamically controls the event density and optimizes the conflict rhythm by combining the genre adjustment factor κ.

[0051] The advantages of this application compared to existing technologies are:

[0052] (1) A hierarchical organization of conflict events is achieved through a dynamic binary tree recursive structure (N-level expansion), with the number of nodes at each level set to 2. N-2 The number of nodes increases regularly, reaching a total of 2. N -2, supporting the structured generation of complex story outlines.

[0053] (2) Through The equal partial derivative constraint ensures the strength of conflict dependency, and the causal chain weights in the β / γ eigenvectors are combined to achieve the logical coherence of conflict events.

[0054] (3) Based on I e The equation quantifies conflict emotions and combines conflict complexity (such as the number of branches) with story position (such as the emotional difference between early struggles and later counterattacks) to achieve dynamic adjustment of emotional intensity, adapting to the needs of different genres such as fantasy and xianxia.

[0055] (4) Through T n Functions and Suspense Entropy H s The synergistic effect of these factors dynamically controls the density of events and the tension of suspense, avoiding a loss of rhythm (such as increasing the probability of unresolved conflicts pi at key plot points to enhance suspense).

[0056] (5) The story elements are abstracted into β / γ feature vectors, and the conflict generation is driven by input vectors such as α=(S,C,β), so that the generation of conflict nodes has a clear causal basis and worldview relevance (such as the generation of breakthrough events under the constraints of cultivation system rules).

[0057] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0058] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0059] Figure 1 This is a schematic diagram of a recursive conflict tree model structure of a novel conflict modeling method and system based on dynamic binary trees and feature-driven methods according to an embodiment of this application.

[0060] Figure 2 This is a schematic diagram illustrating the logic of constructing pre- and post-conflict subforests according to an embodiment of the novel conflict modeling method and system based on dynamic binary trees and features, in accordance with this application.

[0061] Figure 3 This is a schematic diagram illustrating the generation process of feature vectors β and γ in a novel conflict modeling method and system based on dynamic binary trees and feature-driven methods according to an embodiment of this application.

[0062] Figure 4 This is a schematic diagram of the conflict tree node distribution and traversal order of a novel conflict modeling method and system based on dynamic binary trees and feature-driven methods according to an embodiment of this application.

[0063] Figure 5 This is a schematic diagram illustrating the calculation process of the suspense entropy equation of a novel conflict modeling method and system based on dynamic binary trees and feature-driven methods according to an embodiment of this application.

[0064] Figure 6 A schematic diagram of the dynamic effect of the rhythm control function T_n of a novel conflict modeling method and system based on dynamic binary trees and feature-driven methods according to an embodiment of this application;

[0065] Figure 7 This is a schematic diagram of the conflict modeling system modules of a novel conflict modeling method and system based on dynamic binary trees and feature-driven methods according to an embodiment of this application. Detailed Implementation

[0066] The following describes, with reference to the accompanying drawings, a novel conflict modeling method and system based on dynamic binary trees and feature-driven approaches according to embodiments of this application.

[0067] like Figures 1-7 As shown in the figure, the novel conflict modeling method based on dynamic binary trees and feature-driven methods in this application includes the following steps:

[0068] I. Story Triple Structure (S1)

[0069] Let the set of stories be M, consisting of triples: M = SCE. Where:

[0070] S (Initial Conflict Event): The initial disturbance factor at the beginning of the story, usually an extreme predicament faced by the protagonist (such as the annihilation of his family, loss of abilities, etc.);

[0071] E (End of Conflict Event): The final steady state of the story's ending, i.e., the protagonist's goal is achieved or the conflict is completely resolved (such as successful revenge, world peace, etc.);

[0072] C (Core Conflict Event): The global turning point, the key node connecting the beginning and the end (such as the protagonist obtaining a mysterious inheritance or revealing the truth of a conspiracy).

[0073] Example: In fantasy novels, S can be "the young Lin Fan's family was wiped out", C can be "awakening his bloodline and obtaining the Nine Tribulations Divine Technique", and E can be "destroying the villainous forces and rebuilding his family".

[0074] Sequence connector ( () indicates the sequential connection relationship of conflicting events, such as This indicates that events A, B, and C occur in a chronological order.

[0075] II. Construction of the Recursive Conflict Tree Model (S2)

[0076] 1. Basic Structure (N=1)

[0077] When the conflict tree depth is 1, the model is in its simplest form: M1 = SCE, which only contains the start, core, and end events, forming the main framework of the story.

[0078] 2. Recursive expansion structure (N≥2)

[0079] When the depth is N, the model uses the preceding conflict subforest F. n (β) and post-conflict subforest G n (γ) Conflict of extended details:

[0080]

[0081] Pre-conflict subforum F n (β):

[0082] Includes all the preparatory events before the protagonist reaches the core conflict, denoted by the preceding feature vector β = (β1, β2, ..., β m The driver, with parameters including:

[0083] Causal relationship strength index: quantifies the degree of logical dependence between events (such as the correlation strength between "being betrayed" and "cultivating revenge");

[0084] Character emotional state transition probability: describes the likelihood of the protagonist's emotional state changing (such as the probability of transitioning from "despair" to "resilience");

[0085] Worldview rules and constraints parameters: conform to the rules and restrictions set in the story (such as "forbidden techniques cannot be used during the Qi Refining stage" in fantasy novels).

[0086] Post-conflict subforest G n(γ):

[0087] Including development events following the core conflict, denoted by the post-feature vector γ = (γ1, γ2, ..., γ k Driven by β, with parameters similar to β, but emphasizing the impact of core conflicts on the outcome (such as the promotion of "gaining godhood" on "challenging the immortal realm").

[0088] Recursive generation rules:

[0089] When N = 2, F2(β) = L2(α), G2(γ) = R2(δ), and the model is extended to S-L2(α)-C-R2(δ)-E;

[0090] When N>2 That is, each layer adds left and right branch conflicts based on the previous layer.

[0091] III. Dynamic Conflict Node Generation (S3)

[0092] 1. Left branch node generation function

[0093] L n (α) = σ(Ψ·α+b), used to generate preceding conflict events:

[0094] Input vector α: α = (S, C, β), which integrates initial conflict, core conflict and prior features;

[0095] Conflict weight matrix Ψ: Adjusts the weight of each input parameter on the conflict node (e.g., the weight of S is higher than that of the secondary features in β);

[0096] Emotional bias term b: Introduces basic emotional tendency (e.g., in tragic themes, b is negative, which enhances the sense of oppression);

[0097] Nonlinear activation functions σ, such as tanh or ReLU, transform linear combinations into nonlinear conflict intensities (e.g., tanh(·) limits the result to [-1,1], corresponding to the intensity of the conflict).

[0098] 2. Right branch node generation function

[0099] R n (δ) = φ(Θ·δ+d), used to generate subsequent conflict events:

[0100] Input vector δ: δ = (C, E, γ), which integrates core conflict, termination conflict and post-features;

[0101] Conflict weight matrix Θ, sentiment bias term d, activation function φ: function similarly to the left branch, but parameters are adjusted for subsequent logic (e.g., the weight of E affects the outcome).

[0102] IV. Emotional Intensity Control (S4)

[0103] Emotional intensity equation: I e =tanh(ω·c+θ·t+κ), quantifying the emotional impact of a conflict event:

[0104] c (number of branches): The number of conflicting branches in the current layer. The more branches, the higher the conflict complexity (e.g., when N=3, c=2). 3 -2 = 2);

[0105] t (position parameter): The relative position of the event in the storyline, 0 represents the start S, 1 represents the end E (e.g., t = 0.5 for a mid-stage event);

[0106] ω, θ (emotional weighting coefficients): Adjust the proportion of the influence of complexity and location on emotion (e.g., ω = 0.6 emphasizes conflict complexity, θ = 0.4 emphasizes story stage);

[0107] κ (Genre Adjustment Factor): Adjusts the baseline intensity based on the genre (e.g., κ = 0.3 for fantasy novels, κ = 0.1 for romance novels).

[0108] Function: To ensure that the intensity of conflict changes dynamically as the story progresses. For example, when the number of branches increases in the middle stage (t=0.5), the emotional intensity I... e Significant improvement.

[0109] V. Logical Reliability Constraints (S5)

[0110] 1. Causal coherence constraint:

[0111] The partial derivative of the preceding conflict subforum F with respect to the initial conflict S is positive, meaning that an increase in the intensity of S will inevitably lead to an increase in the intensity of events in F (e.g., when S is "extermination of a family", the motivation for "cultivation and revenge" in F is stronger).

[0112] The partial derivative of the post-conflict subforum G with respect to the ending event E is negative, meaning that the higher the difficulty of achieving E, the stronger the obstruction event in G (e.g., if E is "challenging the Immortal Emperor", the "Immortal Realm Siege" event in G is more intense).

[0113] 2. Increasing Conflict Constraint:

[0114] The partial derivative of the current branch conflict with respect to the upper branch must be greater than 0.6 to ensure that the conflict intensity increases by at least 60% at each level (e.g., the intensity of L(3,1) with N=3 must be more than 60% higher than that of L(2,1) with N=2) to avoid weak plot.

[0115] VI. Summary of Technical Solution Workflow

[0116] 1. Initialization: Define the S, C, E triplet to determine the core framework of the story;

[0117] 2. Recursive expansion: Following the rule of N≥2, through F n (β) and G n (γ) Generates a multi-level conflict tree, with nodes dynamically created by left and right branch functions at each level;

[0118] 3. Emotional and logical control: utilizing I e The equation adjusts the intensity of conflict emotions, and the partial derivative constraint ensures the causal coherence and increasing intensity of events.

[0119] 4. Generate a complete conflict chain: starting from S, through F n (β), C, G n (γ) to E, forming a story outline containing multiple layers of detailed conflict.

[0120] In one embodiment of this application, such as Figures 1-7 As shown, the pre-conflict subforest construction satisfies:

[0121]

[0122] The post-conflict subforest construction satisfies:

[0123]

[0124] Understandably, the following section covers: I. Basic Definitions and Symbol Explanation

[0125] 1. Core Concepts of Conflict Subforest

[0126] Pre-conflict subforum F n (β): refers to the set of all preparatory conflicts located between the initial conflict event S and the core conflict event C, defined by the preceding feature vector β = (β1, β2, ..., β). m The driving force includes parameters such as the strength of causal relationships in the story and the probability of character emotional shifts;

[0127] Post-conflict subforest G n (γ): refers to the set of all developmental conflicts located between the core conflict event C and the terminating conflict event E, defined by the posterior feature vector γ = (γ1, γ2, ..., γ). k Driven by parameters that emphasize the impact of core conflicts on the outcome;

[0128] Sequence connector : Indicates the sequential arrangement of conflicting events, such as This indicates that event A occurred before event B.

[0129] 2. Number of recursion levels n

[0130] n represents the depth of the conflict tree. n=1 represents the simplest structure, and when n≥2, multi-level conflict nodes are generated through recursive expansion.

[0131] II. Pre-conflict subforum F n (β) Construction rules

[0132] 1. Basic Case (when n=1)

[0133] (Empty set), that is, when the conflict tree depth is 1, there are no additional preparatory events between S and C, and they are directly connected as SCE.

[0134] Example: If n=1, the outline of a fantasy novel can be simplified to "S (family massacre and awakening of inheritance) - C (obtaining divine techniques) - E (successful revenge)", without any intermediate conflicts.

[0135] 2. Recursive expansion (when n≥2)

[0136] Construction formula:

[0137] Workflow:

[0138] ①Preorder structure inheritance: First, completely preserve the preceding conflicting subforum F at level n-1. n-1 (β), ensuring that the order of historical conflict events remains unchanged;

[0139] ② Add a new left branch for the current layer: Generate function L from the left branch n (α) Create the pre-conflict event of the nth layer, where α=(S,C,β) is the input vector, and fuse the initial conflict, core conflict and pre-conflict features;

[0140] ③ Sequence concatenation: This will concatenate the newly added event L. n (α) connected to F n-1 (β) then forms a deeper chain of preceding conflicts.

[0141] Example of recursive expansion:

[0142] When n=2: That is, S-L2(α)-CE;

[0143] When n=3: That is, S-L3(α)-L2(α)-CE;

[0144] When n=4: And so on, adding nodes on the left side for each layer and arranging them in recursive order.

[0145] III. Post-conflict subforest G n (γ) Construction rules

[0146] 1. Basic Case (when n=1)

[0147] That is, there are no additional development events between C and E, and they are directly connected as CE.

[0148] Example: When n=1, the outline is "SC (obtain the divine secret)-E", omitting the challenge process after the core conflict.

[0149] 2. Recursive expansion (when n≥2)

[0150] Construction formula:

[0151] Workflow:

[0152] ① Add a right branch to the current layer: Generate function R through the right branch n (δ) Create the post-conflict event of the nth layer, where δ=(C,E,γ) is the input vector, and fuse the core conflict, the ending conflict and the post-conflict features;

[0153] ② Post-order structure inheritance: inherit the post-conflict subforest G at level n-1. n-1 (γ) Following the newly added event R n (δ) After that, ensure the sequential continuation of subsequent conflict events;

[0154] ③ Sequence splicing: forming The subsequent conflict chain connects C and E.

[0155] Example of recursive expansion:

[0156] When n=2: That is, C-R2(δ)-E;

[0157] When n=3: That is, C-R3(δ)-R2(δ)-E;

[0158] When n=4: Each layer adds a new node on the right and arranges them in recursive order, forming a symmetrical expansion with the previous subforum (but logically driven by δ).

[0159] IV. Explanation of the logical coherence of recursive construction

[0160] 1. The causal progression of the preceding sub-forest

[0161] F n The recursive construction of (β) ensures that the preceding conflict events are arranged in the order of "near core conflict → far initiation conflict" (e.g., when n=3). L2 is closer to C, and L3 is closer to S, which conforms to the narrative logic of "resolving recent conflicts first and then tracing back to earlier groundwork".

[0162] For example, in the fantasy synopsis, when n=5, F5 contains events from L(5,1) to L(5,8), which are arranged according to the recursive rule as progressive conflicts from S to C (such as "being demoted to a slave" → "being betrayed by a friend" → "being bullied by the outer sect of the sect").

[0163] 2. Results-oriented approach of post-planting subforests

[0164] G n The recursive construction of (γ) is arranged in the order of "near core conflict → far termination conflict" (e.g., when n=3). R3 is closer to C, and R2 is closer to E, reflecting how the core conflict gradually drives the outcome.

[0165] Example: After the core conflict C is "obtaining the divine technique", G3 may include "attempting to create a body refining technique" (R3) → "breaking through the peak of Qi refining" (R2), ultimately leading to E (revenge).

[0166] 3. Constraint Support

[0167] The construction process is subject to logical reliability constraints (such as...) This ensures that the intensity of each new conflict level is higher than the previous one, avoiding plot gaps. For example, the conflict intensity of L3 needs to be more than 60% higher than that of L2, ensuring that the challenges faced by the protagonist gradually escalate.

[0168] In one embodiment of this application, such as Figures 1-7 The following is a description of the implementation method for generating feature vectors β and γ:

[0169] I. Basic Definition of Eigenvectors

[0170] Pre-feature vector β = (β1, β2, ..., β) m ): Used to drive the preceding conflicting subforest F n The generation of (β) is related to the initial conflict S and the core conflict C, reflecting the preparatory event logic from S to C.

[0171] Post-feature vector γ = (γ1, γ2, ..., γ k ): Used to drive the subsequent conflicting subforest G n The generation of (γ) is related to the core conflict C and the ending conflict E, reflecting the developmental event logic from C to E.

[0172] II. Three Core Steps in Feature Vector Generation

[0173] 1. Calculation of causal chain weights and impact factors

[0174] Input: Initial conflict S, core conflict C, and a sequence of historical conflict events (e.g., F) n-1 (Events in the middle).

[0175] Processing flow:

[0176] ① Construct a causal dependency graph from S to C, and label the direct / indirect impact of each intermediate event on C (e.g., in the sequence "family extermination" → "cultivation for revenge" → "obtaining divine techniques", the impact weight of "cultivation for revenge" on C is higher than that of other events).

[0177] ② The influence factor is calculated using a weighted directed edge model:

[0178] The weight of a direct causal event is α1 = 0.7 (e.g., S directly causes L). n (α)), the weight of the indirect causal event α2 = 0.3 (e.g., L) n-1 Impact L n );

[0179] The formula for calculating the impact factor is: f_i = α1·w_i + α2·∑f_j (w_i is the inherent weight of event i, and f_j is the impact factor of the preceding event).

[0180] Output: Causal chain weight parameters, which constitute the basic logical dimensions (such as β1 and γ1) in β and γ.

[0181] 2. Extraction of the probability matrix for character emotional state transition

[0182] Input: A collection of the protagonist's emotional states during conflict events (such as "despair", "anger", "resilience"), and a record of historical emotional changes.

[0183] Processing flow:

[0184] ① Define the emotional state space S_emotion={e1,e2,…,e_n} (e.g. e1=fear, e2=determination);

[0185] ② Statistically analyze the frequency of emotional transfer between adjacent conflict events to generate a transfer matrix P:

[0186] P[i][j] represents the probability of transitioning from state e_i to e_j. For example, P[“despair”][“resilience”] = 0.6 means that there is a 60% probability of transitioning from despair to resilience.

[0187] ③ Adjust the probability based on the story stage: the probability of negative emotions shifting is higher in the early stages of conflict (e.g., P[“despair”][“anger”] = 0.7 in β), and shifts to positive emotions in the later stages (e.g., P[“confidence”][“victory”] = 0.8 in γ).

[0188] Output: Emotional transfer probability parameters, which constitute the emotional dimensions (e.g., β2, γ2) in β and γ.

[0189] 3. Integration of Worldview, Rule, and Constraint Parameters

[0190] Input: Novel world-building document (such as cultivation system, power relationships, regional rules).

[0191] Processing flow:

[0192] ① Extract rule constraints into quantifiable parameters:

[0193] Cultivation system: Breakthrough conditions for stages such as Qi Refining → Foundation Establishment → Golden Core (e.g., the constraint parameter β3 = 0.9 corresponding to "Forbidden techniques cannot be used during the Qi Refining stage" in β, prohibiting high-difficulty events);

[0194] Factional Relationships: Hatred Value of Hostile Sects (e.g., the "Sect Pursuit" event in γ is driven by the faction hatred parameter γ3 = 0.8);

[0195] ② Convert to binary or numerical constraints:

[0196] Allowed events: parameter = 1 (e.g., "hunting monsters" in the mid-stage of cultivation, γ4 = 1);

[0197] Forbidden events: parameter = 0 (e.g., "using the Immortal Emperor's technique" in the Foundation Establishment stage β4 = 0).

[0198] Output: Worldview constraint parameters, constituting the rule dimensions in β and γ (e.g., β) m γ k ).

[0199] III. Logical Integration of Feature Vector Generation

[0200] 1. Generation of β:

[0201] With S and C as the core, the causal chain weights from S to C (which determine the logical order of events), the probability of early emotional transfer (such as the "determination of revenge" after "extermination of the family"), and the initial rules of the worldview (such as the restrictions on low-level cultivation) are integrated to form β = (causal parameters, emotional parameters, rule parameters).

[0202] 2. Generation of γ:

[0203] Centered on C and E, the causal chain weights from C to E are integrated (such as the impetus of "obtaining the divine technique" to "challenging the sect"), the probability of emotional transfer in the later stages (such as the confidence of "victory is in sight"), and the high-level rules of the worldview (such as the intervention of the immortal forces) to form γ = (causal parameter, emotional parameter, rule parameter).

[0204] In one embodiment of this application, such as Figures 1-7 As shown, the implementation method for depth-first traversal and calculation of conflict node count is explained below:

[0205] I. Application of Depth-First Search (DFS) in Conflict Trees

[0206] 1. Definition of traversal

[0207] The conflict event sequence is generated in an in-order recursive order of "left branch → core conflict → right branch", ensuring that deep child nodes are processed first and then the upper level is returned, which conforms to the gradual development of the narrative logic.

[0208] Example: When N=3, the traversal order is L(3,1)→L(2,1)→L(3,2)→C→R(3,1)→R(2,1)→R(3,2), that is, first all nodes of the left branch, then the core conflict, and finally all nodes of the right branch.

[0209] 2. Recursive implementation logic

[0210] For the left branch F n (β) and right branch G n (γ) Employ a depth-first strategy:

[0211] ① First, recursively process the nth level node L of the left branch. n (α), then process n-1 layers of nodes until the innermost layer;

[0212] ② Resolving core conflict C;

[0213] ③ Recursively process the right branch node R in a mirror-symmetric manner. n (δ), to ensure that the left and right structures correspond.

[0214] II. Calculation of the number of new nodes per layer (N≥2)

[0215] 1. Formula: Number of new nodes per layer = 2 N-2

[0216] 2. Derivation and Examples

[0217] When N=2: Number of new nodes = 2 2-2 =1, corresponding to the left branch L(2,1) and the right branch R(2,1), a total of 2 nodes (1 on each side);

[0218] When N=3: Number of new nodes = 2 3-2 =2, the left branch adds L(3,1) and L(3,2), the right branch adds R(3,1) and R(3,2), for a total of 4 nodes;

[0219] The rule is that the number of new nodes doubles with each additional layer (exponential growth), ensuring that the complexity of conflicts increases with the number of layers.

[0220] III. Calculation of Total Number of Nodes (excluding S, C, and E)

[0221] 1. Formula: Total number of nodes = 2 N -2

[0222] 2. Derivation and Examples

[0223] When N=1: Total number of nodes = 2 1 -2 = 0, meaning there is only SCE and no additional conflicting nodes;

[0224] When N=2: Total number of nodes = 2 2 -2 = 2, corresponding to L(2,1) and R(2,1);

[0225] When N=3: Total number of nodes = 2 3 -2 = 6, corresponding to the left branch [L(3,1),L(2,1),L(3,2)] and the right branch [R(3,1),R(2,1),R(3,2)];

[0226] In one embodiment of this application, such as Figures 1-7 As shown, the implementation method of suspense entropy value equation and narrative tension management is explained below:

[0227] I. Basic Definition of Suspense Entropy Equation

[0228] Suspense Entropy H_S: Quantifies the degree of uncertainty of unresolved conflicts in a novel's narrative. The higher the value, the stronger the narrative tension and the higher the reader's expectation for subsequent developments.

[0229] Mathematical expression:

[0230]

[0231] Among them: H0 is the basic suspense value, which is determined by the novel's genre;

[0232] p i Let be the probability distribution of the i-th unresolved conflict;

[0233] η is a random perturbation factor to avoid rigidity in the intensity of suspense.

[0234] II. Parameter Analysis and Calculation Methods

[0235] 1. Basic suspense value H0

[0236] Definition: The baseline suspense intensity based on the novel genre.

[0237] Example:

[0238] Fantasy / Xianxia genre: H0 = 0.8 (requires maintaining long-term suspense, such as "the sealing of the godhood");

[0239] Short story: H0 = 0.5 (short suspense period, low intensity).

[0240] 2. Unresolved conflict probability distribution pi

[0241] Calculation logic:

[0242] ① Calculate the set of currently unresolved conflict events {E1, E2, ..., E...} m}, such as "the truth behind the massacre of a family".

[0243] "Mysterious origins";

[0244] ② Assign probabilities based on the weighted impact of conflicts on the main storyline:

[0245] Where w i For conflict E i Weights (e.g., core conspiracy w) i =0.4, secondary branch w i =0.1).

[0246] Example: If there are two unresolved conflicts with weights of 0.6 and 0.4 respectively, then p1 = 0.6 and p2 = 0.4.

[0247] 3. Information entropy calculation term ∑pi·logpi

[0248] Function: To measure the uncertainty of unresolved conflicts. The more uniform the probability distribution (e.g., p...), the more effective it becomes. i (Approaching), the larger the entropy value, the more dispersed the suspense; when a certain conflict probability is prominent (e.g., p), i =0.8), the smaller the entropy value, the more concentrated the suspense.

[0249] Example:

[0250] Uniform distribution (p1=p2=0.5): 0.5·log0.5+0.5·log0.5≈-0.693, then 1-(-0.693)=1.693;

[0251] Centralized distribution (p1 = 0.8, p2 = 0.2): 0.8·log0.8 + 0.2·log0.2 ≈ -0.503, then 1 - (-0.503) = 1.503.

[0252] 4. Random disturbance factor η

[0253] Values: random numbers η∈[-0.1,0.1]) to avoid the mechanical nature of the formula.

[0254] Function: Temporarily increase or decrease suspense in key plot points (e.g., η = 0.1 when there is a sudden plot twist, and η = -0.1 when the pace is slowed down).

[0255] In one embodiment of this application, such as Figures 1-7 As shown, the implementation method of rhythm control function event density adjustment is explained as follows:

[0256] I. Basic Definition of Rhythm Control Function

[0257] Function expression:

[0258]

[0259] This is used to quantify the density and rhythm of conflict events at the nth level in a novel. The larger the value, the more events there are per unit of text, and the faster the pace.

[0260] II. Parameter Analysis and Function

[0261] 1. Attenuation coefficient β

[0262] Value: Usually β>0, which determines the steepness of the rhythm growth.

[0263] effect:

[0264] The larger β is, the larger the exponential term e -β·n The faster the decay, The faster the speed approaches 1, the greater the acceleration of the rhythm as the number of layers n increases;

[0265] For example, when β = 0.5, the rhythm factor is approximately 0.77 when n = 3; when β = 1, the rhythm factor is approximately 0.95 when n = 3, with the latter showing a more dramatic increase in rhythm.

[0266] 2. Fluctuation amplitude λ

[0267] Values: Usually λ∈[0.1,0.3], controlling the range of random perturbations.

[0268] Function: Generates a random fluctuation of ±20% using rand(-0.2, 0.2), avoiding mechanical rhythm. For example, when λ = 0.2, the fluctuation range is ±4% (0.2 × 0.2), making T... n The calculated value will fluctuate by ±4%.

[0269] 3.

[0270] Properties: The range is [0,1]. When n→∞, it approaches 1, which corresponds to the maximum event density, conforming to the rhythmic pattern of "early setup and later climax" in the story.

[0271] Example: When n=1, The pace is relatively slow; when n=5, The rhythm is approaching its climax.

[0272] III. Event Density Adjustment Workflow

[0273] 1. Parameter initialization

[0274] β and λ are set according to the novel genre:

[0275] Long fantasy / xianxia novels: β = 0.8 (rapid acceleration in the mid-to-late game), λ = 0.2 (moderate fluctuation);

[0276] Short story: β = 0.4 (gradual growth), λ = 0.1 (small fluctuations).

[0277] 2. Calculate T layer by layer n

[0278] For the nth level conflict, first calculate the basic rhythm factor:

[0279] Superimposed random perturbation: T n = f(n)·(1+λ·r), where r∈(-0.2,0.2) is a random number.

[0280] 3. Event density mapping

[0281] According to T n Adjust the number of conflict events at this layer:

[0282] Number of events = Base number of events × T n

[0283] Example: When n=2, the base number of events is 2, T2=0.6, so approximately 2×0.6≈1 event is actually generated (rounded down); when n=5, T5=0.98, 2 events are generated. 5-2 =8 events (because the number of new nodes per layer = 2) N-2 (This needs to be combined with the formula for the number of nodes).

[0284] 4. Dynamic adjustment of rhythm

[0285] When T n <0.5 hours (slow pace), arranging foreshadowing, flashbacks, and other plot points;

[0286] When T n >0.8 (fast pace), with frequent occurrences of conflict outbreaks, battles, and other events;

[0287] The random term λ·r is used for unexpected events (such as when n=3, r=0.15, T3 is temporarily increased by 15%, and an unexpected turning point is inserted).

[0288] In one embodiment of this application, such as Figures 1-7 The following is a description of the implementation method of the conflict modeling system:

[0289] I. Structure Definition Module

[0290] 1. Core Function: Establish the basic framework of the story and store the recursive construction rules of the conflict tree.

[0291] 2. Workflow:

[0292] Triad initialization: Define the initial conflict S, core conflict C, and ending conflict E to form the basic structure M=SCE (e.g., S=“family extermination”, C=“awakening divine technique”, E=“successful revenge”).

[0293] Recursive rule base storage:

[0294] When N=1, the conflict tree is M1=SCE;

[0295] When N≥2, according to Recursive expansion, where F n / G n These are the pre- and post-conflict subforests, respectively.

[0296] II. Feature Engine

[0297] 1. Core functions: Generate feature vectors β / γ, verify causal constraints, and calculate event impact factors.

[0298] 2. Workflow:

[0299] Feature vector generation:

[0300] ① Based on the causal chain of S and C, calculate the influence factor of historical events on the current conflict (such as the influence weight of "extermination of the family" on "cultivation");

[0301] ② Extract the probability matrix of character emotion transfer (such as the transfer probability of "despair → resilience");

[0302] ③ Integrate worldview rules (such as cultivation system stage restrictions) to generate β = (causal parameters, emotional parameters, rule parameters), and γ is generated in the same way.

[0303] Causal constraint verification:

[0304] make sure (When the intensity of S increases, the intensity of the preceding conflict F must also increase);

[0305] make sure (When the difficulty of E increases, the intensity of the subsequent conflict G will necessarily increase).

[0306] III. Neural Network Execution Unit

[0307] 1. Core function: Generate conflict nodes between left and right branches through neural network functions.

[0308] 2. Workflow:

[0309] Left branch generation:

[0310] Given α = (S, C, β), an intermediate value is generated through the linear transformation Ψ·α + b.

[0311] After processing by an activation function σ (such as ReLU / Sigmoid), the output L is obtained. n (α).

[0312] Example: When σ = ReLU, if Ψ·α + b = -2, then L n =0 (filtering invalid conflicts); if the result =3, then L n =3 (High-intensity conflict).

[0313] Right branch generation:

[0314] Given input δ = (C, E, γ), perform a linear transformation using Θ·δ+d;

[0315] The output R is activated by the φ function. n (δ), logically symmetric with the left branch.

[0316] IV. Entropy Controller

[0317] 1. Core Function: Utilizing the suspense entropy value H s Managing narrative tension and providing feedback to regulate conflict generation.

[0318] 2. Workflow:

[0319] Real-time calculation of H s：

[0320] Where H0 is the subject baseline value, pi is the probability of unresolved conflicts, and η is the random perturbation.

[0321] Feedback adjustment:

[0322] When H s When the value is too high (e.g., >1.0), a conflict resolution event is generated (e.g., revealing part of the truth).

[0323] When H s When the value is too low (e.g., <0.6), new branch conflicts (e.g., villain raids) will occur.

[0324] V. System Collaboration Process

[0325] 1. The structure definition module constructs the SCE framework and recursion rules;

[0326] 2. The feature engine generates β / γ and verifies causal constraints, then inputs them into the neural network;

[0327] 3. The neural network execution unit uses L... n / R n The function generates conflicting nodes and constructs F according to recursive rules. n / G n ;

[0328] 4. Entropy controller calculates H sAdjust conflict generation strategies in real time to ensure narrative tension.

[0329] This system achieves structured generation of conflicts through modular design, and combines neural networks and entropy models to ensure logical coherence and controllable tension in the plot. It is suitable for the automatic construction of outlines for fantasy, martial arts and other genres.

[0330] In summary, the novel conflict modeling method and system based on dynamic binary trees and feature-driven approaches of this application solves the technical defects of existing novel conflict modeling through the dynamic binary tree recursive structure and feature vector-driven mechanism. It realizes hierarchical generation of conflict events, quantitative constraints on causal relationships, scientific regulation of emotional intensity, and dynamic management of narrative rhythm, and is applicable to the outline construction of various fictional literary works.

Claims

1. A novel conflict modeling method based on dynamic binary trees and feature-driven methods, characterized in that, The following steps are involved: S1. Define the story triplet structure M = SCE, where S is the initiating conflict event, E is the ending conflict event, and C is the core conflict event; S2. Construct a recursive conflict tree model: Basic structure: When N=1, M1=SCE; Recursive structure (N≥2): M n =S⊕F n (β)⊕C⊕G n (γ)⊕E, where F n (β) represents the pre-conflict subforum, G n (γ) represents the post-conflict subforest, β = (β1, β2, ..., β) m ) and γ=(γ1,γ2,…,γk) are the pre- and post-feature vectors containing the strength index of causal relationship in the story, the probability of character emotional state transition and the worldview rule constraint parameters, respectively; S3. Dynamically create conflict nodes using a branch generation function: Left branch node: L n (α)=σ(Ψ·α+b), where α=(S,C,β), Ψ is the conflict weight matrix, b is the sentiment bias term, and σ is the nonlinear activation function; Right branch node: R n (δ)=φ(Θ·δ+d), where δ=(C,E,γ), Θ is the conflict weight matrix, d is the sentiment bias term, and φ is the nonlinear activation function; S4. Control the conflict intensity using the emotional intensity equation: I e =tanh(ω·c + θ·t + κ), where c is the number of branches, t is the event position parameter (0 < S→E < 1), ω and θ are emotional weight coefficients, and κ is the genre adjustment factor; S5. Execution logic reliability constraints: To ensure the causal consistency of conflict events.

2. The novel conflict modeling method based on dynamic binary trees and feature-driven approach according to claim 1, characterized in that, The preceding conflict subforest construction satisfies: The post-conflict subforest construction satisfies:

3. The novel conflict modeling method based on dynamic binary trees and feature-driven approach according to claim 1, characterized in that, The generation of the feature vectors β and γ includes: Based on the causal chain weights of the initial conflict S and the core conflict C, the influence factor of historical events on the current conflict is calculated. Extract the transition probability matrix of the character's emotional state in the evolution of the conflict; It incorporates the rules and constraints of the novel's worldview (such as cultivation system and power relationships).

4. The novel conflict modeling method based on dynamic binary trees and feature-driven approach according to claim 1, characterized in that, A collision sequence is generated using a depth-first traversal, where the number of layers N and the number of nodes satisfy the following: Number of new nodes per layer = 2 N-2 (N≥2); Total number of nodes = 2 N -2 (excluding S, C, and E; 0 when N = 1).

5. The novel conflict modeling method based on dynamic binary trees and feature-driven approach according to claim 1, characterized in that, Managing narrative tension through the suspense entropy equation: Where H0 is the basic suspense value, p i Let η be the probability distribution of unresolved conflicts, and η be the random perturbation factor.

6. The novel conflict modeling method based on dynamic binary trees and feature-driven approach according to claim 1, characterized in that, Adjusting event density using a rhythm control function: Where β is the attenuation coefficient, λ is the fluctuation amplitude, and rand(-0.2,0.2) is a random number within the interval.

7. A conflict modeling system implementing any one of the methods of claims 1-6, characterized in that, include: Structure definition module: Configured to create SCE triplet structure and store recursive rule base (conflict tree construction logic for N=1 and N≥2); Feature Engine: Dynamically generates β and γ feature vectors for verification. and The causal constraints are determined, and the event impact factor is calculated based on the correlation between S and C; Neural network execution unit: built-in L n (α), R n (δ) The generating function performs a linear transformation through the weight matrix Ψ / Θ and the bias term b / d, and generates left / right branch conflict nodes through the σ / φ activation function (either ReLU or Sigmoid). Entropy controller: Real-time calculation of H s The value is adjusted based on the suspense entropy feedback to maintain narrative tension; Rhythm adjustment module: based on T n The function dynamically controls the event density and optimizes the conflict rhythm by combining the genre adjustment factor κ.

Images