A multi-agent movie plot generation method based on capability boundary constraint

Abstract

The application discloses a kind of multi-agent movie plot generation methods based on ability boundary constraint, it is related to artificial intelligence inference technical field, including, collect movie plot creation demand, and execute plot semantic extraction, output plot task data;According to the plot task of narrative undertaking attribute, plot task data is matched with agent, determines plot agent, and according to plot responsibility calibration allows inference section and limit inference section and configures out-of-bounds limit, output ability boundary data;According to ability boundary data, plot task is bound with plot agent with boundary, according to the plot of plot task in succession plot generation order arrangement and handover information limit of boundary binding relationship, output boundary scheduling data;Through neural network, plot agent is driven to boundary scheduling data and executes constraint inference, output movie plot data.The application improves the boundary controllability of multi-agent movie plot segment by boundary constraint inference and segment ownership mark.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence reasoning technology, and in particular to a multi-agent film and television plot generation method based on capability boundary constraints. Background Technology

[0002] In recent years, with the development of generative neural networks, natural language generation, semantic representation learning, narrative structure modeling, and intelligent collaborative creation technologies, film and television plot generation has gradually shifted from single text continuation to plot semantic understanding, narrative task decomposition, character behavior deduction, scene relationship organization, and fragmented plot compilation. In film and television plot creation scenarios, character relationships, scene transitions, event causality, conflict progression, plot rhythm, and dialogue style together constitute a complex narrative structure. The plot content generation process needs to maintain semantic consistency, character responsibility consistency, and plot progression continuity simultaneously. Multi-agent collaborative methods enable different plot creation responsibilities to be divided and operated within the same plot task chain, and have become an important research direction for neural network-assisted film and television content generation.

[0003] However, existing methods for generating film and television plots typically focus on generating text content directly based on plot prompts, or using multiple agents to complete local creation tasks separately. They lack continuous constraints on the capability boundaries, overstepping limits, and ownership relationships of generated segments for different plot agents. When multiple plot agents participate in plot compilation, problems such as unclear plot task ownership, rewritten handover information, and difficulty in tracing the source of segments can easily arise, affecting the boundary consistency of film and television plot generation results and the accuracy of subsequent corrections. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a multi-agent film and television plot generation method based on capability boundary constraints to solve the problem of lack of capability boundary constraints and ownership traceability in the field of film and television plot generation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides a multi-agent film and television plot generation method based on capability boundary constraints, which includes:

[0008] Collect film and television plot creation needs, perform plot semantic extraction, obtain plot semantic elements, divide the plot semantic elements into task granularities according to the narrative function and plot progression relationship of the plot semantic elements, and output plot task data;

[0009] The plot task data is matched with intelligent agents according to the narrative undertaking attributes of the plot task to determine the plot intelligent agent. Allowable reasoning segments and restricted reasoning segments are marked according to the plot responsibility and cross-boundary restrictions are configured. Capability boundary data is output. The plot task is plot content that can be independently processed by the plot intelligent agent and is divided by plot semantic elements according to the narrative function and plot progression relationship. The plot responsibility is the plot processing range determined according to the plot task, narrative undertaking attributes and task position belonging of the plot intelligent agent.

[0010] Based on the capability boundary data, establish boundary binding relationships between the plot tasks and the plot agents. Arrange the generation order and handover information of the boundary binding relationships according to the plot sequence of the plot tasks, and output boundary scheduling data.

[0011] By driving the plot agent through a neural network to perform constraint reasoning on the boundary scheduling data, film and television plot segments are generated, and ownership tags are configured for the film and television plot segments to form tagged plot segment data. The tagged plot segment data is then bound to the capability boundary data to trace the source relationship, and ownership plot data is output. According to the capability boundary data, the ownership plot data is corrected for overstepping and plot connection and synthesized, and the film and television plot data is output.

[0012] As a preferred embodiment of the multi-agent film and television plot generation method based on capability boundary constraints described in this invention, the specific steps for obtaining plot semantic elements are as follows:

[0013] Semantic layering is performed on plot descriptions in film and television plot creation to generate plot semantic hierarchy data.

[0014] Identify the narrative objects and narrative relationships corresponding to each plot description based on the plot semantic hierarchy data, and perform plot semantic positioning to output plot semantic elements.

[0015] As a preferred embodiment of the multi-agent film and television plot generation method based on capability boundary constraints described in this invention, the specific steps for outputting plot task data are as follows:

[0016] Based on the role and position of plot semantic elements in film and television plots, the narrative function of plot semantic elements is positioned to form narrative function data;

[0017] The narrative function data contains plot semantic elements that have foreshadowing, causal continuity, and scene connection, which are linked together to form a plot progression link;

[0018] The semantic elements of the plot are transformed into task-level data according to the plot progression chain, and the plot task data is output.

[0019] As a preferred embodiment of the multi-agent film and television plot generation method based on capability boundary constraints described in this invention, the specific steps for determining the plot agent are as follows:

[0020] The narrative responsibility attribute is extracted from the plot task data, and the plot tasks in the plot task data are assigned to tasks according to the narrative responsibility attribute to generate plot task position data.

[0021] The plot task data is matched with the specific responsibilities of each agent, and the matching relationship is bound to determine the plot agent for the film and television plot.

[0022] As a preferred embodiment of the multi-agent film and television plot generation method based on capability boundary constraints described in this invention, the specific steps for outputting capability boundary data are as follows:

[0023] Divide the narrative agent into narrative responsibilities and output agent responsibility data;

[0024] Based on the agent's responsibility data, the plot tasks corresponding to the plot responsibilities are classified as allowed reasoning segments, and the plot tasks and plot sequence relationships that do not correspond to the plot responsibilities are classified as restricted reasoning segments.

[0025] To meet the needs of film and television plot creation, configure outbound entry restrictions and outbound rollback restrictions for allowed and restricted reasoning segments, and output capability boundary data.

[0026] As a preferred embodiment of the multi-agent film and television plot generation method based on capability boundary constraints described in this invention, the boundary binding relationship is formed by performing boundary positioning processing on the capability boundary data, obtaining the capability boundaries and out-of-bounds restrictions of each plot agent, and solidifying the boundary adaptation relationship between the plot tasks in the plot task data and the plot agents with corresponding capability boundaries and out-of-bounds restrictions.

[0027] As a preferred embodiment of the multi-agent film and television plot generation method based on capability boundary constraints described in this invention, the specific steps for outputting boundary scheduling data are as follows:

[0028] According to the plot sequence of the plot tasks, the plot tasks and plot agents in the boundary binding relationship are arranged in the order of plot progression to generate binding order data;

[0029] Based on the binding order data, handover information confirmation and prohibition of modification information are performed on adjacent plot agents to form handover restriction data;

[0030] The binding sequence data and handover constraint data are scheduled and associated according to the plot sequence of the story mission, and the boundary scheduling data is output.

[0031] As a preferred embodiment of the multi-agent film and television plot generation method based on capability boundary constraints described in this invention, the specific steps for forming labeled plot segment data are as follows:

[0032] Semantic representation transformation is performed on the plot tasks, plot agents, and boundary binding relationships in the boundary scheduling data to form boundary compilation data;

[0033] The narrative agent determines the narrative tasks, receiveable handover information, and unchangeable boundary information from the boundary compilation data, and integrates them into agent compilation data according to the corresponding arrangement order of the narrative agents.

[0034] Using a neural network, the plot task is used as the reasoning object, the receivable handover information is used as the connection and maintenance condition, and the unalterable boundary information is used as the constraint condition. The data compiled by the intelligent agent is used to perform plot text reasoning, output plot text, and the plot intelligent agent classifies the plot text into the corresponding plot task to form candidate plot fragment data.

[0035] The candidate plot segment data is checked for boundary compliance based on the boundary scheduling data. Candidate plot segment data that meets the boundary binding relationship is retained to generate film and television plot segments.

[0036] Configure segment ownership tags for the plot agents, plot tasks, and boundary binding relationships corresponding to film and television plot segments, and attach the segment ownership tags to the film and television plot segments to form tagged plot segment data.

[0037] As a preferred embodiment of the multi-agent film and television plot generation method based on capability boundary constraints described in this invention, the specific steps for outputting ownership plot data are as follows:

[0038] Boundary indexing is performed on the ownership markers of the marked story segments to locate the segment boundary index data.

[0039] Align and verify the fragment boundary index data with the capability boundary positions and out-of-bounds restriction positions in the capability boundary data to form traceability alignment data;

[0040] The source alignment data and the marked plot fragment data are encapsulated with ownership rights and output ownership plot data.

[0041] As a preferred embodiment of the multi-agent film and television plot generation method based on capability boundary constraints described in this invention, the specific steps for outputting film and television plot data are as follows:

[0042] According to the capability boundary data, the ownership plot data is traced back to the out-of-bounds position, the ownership plot fragments that exceed the capability boundary data are located, and the ownership plot fragments are returned to the corresponding plot intelligence for limited reasoning correction, forming corrected plot data;

[0043] By utilizing the sequence of plot tasks and fragment ownership markers in the ownership plot data, adjacent plot fragments in the corrected plot data are connected and supplemented to output film and television plot data.

[0044] The beneficial effects of this invention are as follows: By using boundary constraint reasoning and fragment ownership marking, the boundary controllability of multi-agent film and television plot fragments is improved. After semantic representation transformation, the plot task, plot agent, and boundary binding relationship in the boundary scheduling data enable the plot agent to complete plot text reasoning under the constraints of the plot task to be compiled, the acceptable handover information, and the unchangeable boundary information. After the film and television plot fragments are generated, fragment ownership marking is configured to ensure that the plot fragments maintain a clear correspondence with the corresponding plot agent, plot task, and boundary binding relationship. The marked plot fragment data is then bound to the capability boundary data for tracing back the source relationship, correcting the constraint reasoning, and supplementing the connection of adjacent plot fragments, providing a continuous basis for backtracking out of bounds, correcting the constraint reasoning, and supplementing the connection of adjacent plot fragments. This ensures that the film and television plot data generated by multi-agents maintains clear fragment sources, clear boundary constraints, and stable plot connections. Attached Figure Description

[0045] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 This is a flowchart of a multi-agent film and television plot generation method based on capability boundary constraints.

[0047] Figure 2 A flowchart for generating story missions and defining ability boundaries.

[0048] Figure 3 This is a flowchart of boundary binding relationships and boundary scheduling data.

[0049] Figure 4 This is a flowchart for generating ownership plot data and synthesizing film and television plots. Detailed Implementation

[0050] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0051] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0052] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0053] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a multi-agent film and television plot generation method based on capability boundary constraints, including the following steps:

[0054] S1. Collect the needs for film and television plot creation, and perform plot semantic extraction to obtain plot semantic elements. According to the narrative function and plot progression relationship of the plot semantic elements, perform task granularity division on the plot semantic elements and output plot task data.

[0055] S1.1 Perform semantic layering on the plot descriptions in film and television plot creation requirements to generate plot semantic layer data.

[0056] Specifically, the plot descriptions are extracted from the needs of film and television plot creation and segmented according to their narrative meaning into plot theme descriptions, character relationship descriptions, scene state descriptions, event development descriptions, conflict change descriptions, and dialogue intention descriptions. The plot theme descriptions are placed at the overall narrative layer, the character relationship descriptions, scene state descriptions, and event development descriptions are placed at the plot composition layer, and the conflict change descriptions and dialogue intention descriptions are placed at the plot progression layer. Plot descriptions within the same layer are arranged according to their order of appearance and semantic dependency. Hierarchical connections are established for plot descriptions that have hierarchical relationships between different layers, so that the plot descriptions have a clear hierarchical position, category, and sequence relationship in the semantic hierarchy, generating plot semantic hierarchy data.

[0057] Furthermore, when collecting film and television plot creation requirements, we read the plot outline, genre, character settings, character relationships, scene settings, main event lines, conflict direction, narrative style, dialogue style, target length, and creative constraints submitted by the creators. We then organize the text descriptions, segmented outlines, and character settings according to the plot's purpose into film and television plot creation requirements. These requirements include plot themes, character participation, scene locations, event start and end, conflict changes, dialogue intentions, plot sequence, and boundary constraints, providing a source of content for subsequently extracting plot descriptions from these requirements.

[0058] It should be noted that narrative semantic rules are used to classify plot descriptions. Input fields include plot description, plot outline, character settings, character relationships, scene settings, main event line, conflict direction, dialogue style, and creative constraints. When a plot description includes thematic direction, core proposition, story goal, or overall narrative purpose, it is classified as a plot theme description. When a plot description includes character names, character identities, character relationships, character goals, or the basis for character actions, it is classified as a character relationship description. When a plot description includes scene names, spatial locations, temporal states, environmental states, or scene transitions, it is classified as a scene state description. When a plot description includes the occurrence of an event, the progression of the event, the result of the event, or the objects affected by the event, it is classified as an event development description. When a plot description includes the object of conflict, the cause of conflict, the escalation of conflict, the mitigation of conflict, or an opposing relationship, it is classified as a conflict change description. When a plot description includes dialogue objects, dialogue purpose, dialogue emotion, expressive intent, or dialogue function, it is classified as a dialogue intent description.

[0059] Semantic dependency relationships are determined based on shared narrative objects, action sequence, causal orientation, scene continuation, and dialogue sequence. When plot descriptions contain the same characters, scenes, events, conflict objects, or dialogue objects, and one plot description requires the other plot description to provide information on character identities, scene locations, event causes, conflict motives, or dialogue purposes, a semantic dependency relationship is determined to exist between the plot descriptions.

[0060] S1.2 Identify the narrative objects and narrative relationships corresponding to each plot description according to the plot semantic hierarchy data, and perform plot semantic positioning to output plot semantic elements.

[0061] Specifically, according to the hierarchical position, category, and context of each plot description, the core phrases, action phrases, and relational phrases are read item by item from the plot semantic hierarchy data. The parts of the core phrases that point to characters, scenes, events, conflicts, and dialogue are identified as narrative objects, and the parts of the action phrases and relational phrases that reflect participation, occurrence, progression, turning point, cause and effect, connection, and direction are identified as narrative relationships. The narrative objects and narrative relationships are respectively assigned to the corresponding overall narrative layer, plot composition layer, and plot progression layer in the plot semantic hierarchy data. The corresponding positions between narrative objects, between narrative relationships, and between narrative objects and narrative relationships are preserved according to the context of the plot description in the plot semantic hierarchy data, and the plot semantic elements are output.

[0062] Furthermore, core word segments, action word segments, and relational word segments are obtained consecutively based on part-of-speech tagging, syntactic dependency relations, and plot semantic role determination. First, the plot description in the plot semantic level data is divided into text content according to sentence boundaries. Part-of-speech tagging is performed on the divided text content, marking noun word segments and proper noun word segments as candidate core word segments, and verb word segments as candidate action word segments. Using the candidate action word segments as the central position of syntactic dependency relations, candidate core word segments with subject-predicate, verb-object, attributive-head, adverbial-head, and coordinate relationships with the candidate action word segments are searched along the syntactic dependency relations to determine the action initiator and the action receiver. The text describes the process of identifying and associating action-related content. It then details how to classify action initiators, recipients, scenes, causes, results, and dialogue objects according to the semantic hierarchy data. The text also assigns character participation, event occurrence, causal relationships, scene transitions, conflict changes, and dialogue continuity as semantic roles. When a single word segment corresponds to multiple semantic roles, the text identifies the semantic roles that align with the narrative meaning of the plot description, ensuring that core word segments correspond to clear narrative objects and action and relationship word segments correspond to clear narrative relationships, thus outputting the semantic elements of the plot.

[0063] S1.3. Based on the role and position of plot semantic elements in the film and television plot, locate the narrative function of plot semantic elements and form narrative function data.

[0064] Specifically, after the plot semantic elements are output, their hierarchical position, contextual relationship, and corresponding position within the film and television plot are checked to verify their role in plot theme development, character relationship establishment, scene state transitions, event progression, conflict escalation, and dialogue intent expression. Plot semantic elements used to define the plot theme and narrative direction are positioned as having a theme-leading function; those used to define character identities, relationships, and the basis for character actions are positioned as having a character establishment function; those used to define scene entry, scene stay, and scene transition are positioned as having a scene connection function; those used to define the occurrence, progression, and outcome of events are positioned as having an event progression function; those used to define the emergence, escalation, and resolution of conflicts are positioned as having a conflict escalation function; and those used to define the dialogue audience, purpose, and emotion are positioned as having a dialogue expression function. Based on the positioning results between plot semantic elements and narrative functions, the functional affiliation, position, and contextual relationship of the plot semantic elements are recorded to form narrative function data.

[0065] S1.4 Connect the narrative semantic elements in the narrative function data that have forward and backward linkage, causal continuity and scene connection into a plot progression link.

[0066] Specifically, the data is read to determine the functional affiliation, position, and contextual relationships of plot semantic elements in the narrative function data. Relationships where adjacent plot semantic elements require the preceding plot semantic element to provide character objectives, conflict motives, or event conditions are labeled as "preceding traction." Relationships where cause and effect content are linked and the plot theme remains consistent in the event progression function are labeled as "causal continuation." Relationships where scene entry, scene stay, and scene transition are linked and character participation is continuous in the scene connection function are labeled as "scene connection." Following the sequence of events corresponding to preceding traction, causal continuation, and scene connection, the related plot semantic elements are sequentially linked, retaining their functional affiliation and position at the linking points to obtain the plot progression chain.

[0067] S1.5. Perform task-level conversion on the semantic elements of the plot according to the plot progression chain, and output plot task data.

[0068] Specifically, the connection position, functional affiliation, and role of plot semantic elements are read along the plot progression chain. Plot semantic elements that serve the functions of character development, event progression, scene connection, conflict elicitation, or dialogue expression within the same connection segment are merged into plot tasks. Plot semantic elements that span different functional affiliations and play different roles in causal connection, scene continuity, and plot progression are transformed into adjacent plot tasks. According to the plot sequence of plot semantic elements in the plot progression chain, the progression order and correlation basis between adjacent plot tasks are preserved, resulting in plot task data that includes plot tasks, functional affiliation, role position, and plot sequence relationship.

[0069] Furthermore, when converting task granularity, the smallest plot content that can be independently processed by a plot agent in the plot progression chain is used as the benchmark for dividing plot tasks. When multiple plot semantic elements jointly describe the same character's actions, the same scene changes, the same event node, or the same conflict changes, the multiple plot semantic elements are merged into one plot task. When a plot semantic element is associated with character development, event progression, and scene connection functions, it is split into multiple adjacent plot tasks according to its functional affiliation and position in the plot progression chain. Each plot task retains its corresponding plot semantic elements, functional affiliation, position, preceding and following connections, causal continuity, and scene connection, so that the plot task data can clearly reflect the content source, narrative responsibility, and plot sequence of the plot task.

[0070] S2. Match the plot task data with agents according to the narrative responsibility attributes of the plot task, determine the plot agent, mark the allowed reasoning segment and restricted reasoning segment according to the plot responsibility and configure the boundary restriction, and output the capability boundary data. The plot task is the plot content that can be independently handled by the plot agent and formed by plot semantic elements according to the narrative function and plot progression relationship. The plot responsibility is the plot processing scope determined according to the plot task, narrative responsibility attributes and task position belonging to the plot agent.

[0071] S2.1 Extract the narrative responsibility attribute from the plot task data, and determine the task position attribution of the plot tasks in the plot task data according to the narrative responsibility attribute to generate plot task position data.

[0072] Specifically, when the plot task data enters the task position assignment determination, the plot task, function assignment, position of action, and plot sequence relationship in the plot task data are read. Plot tasks whose function is assigned to character development and whose position of action points to character identity, character relationship, and the basis of character action are extracted as character narrative responsibility attributes. Plot tasks whose function is assigned to event progression and whose position of action points to event occurrence, event progression, and event result are extracted as event narrative responsibility attributes. Plot tasks whose function is assigned to scene connection and whose position of action points to scene entry, scene stay, and scene transition are extracted as scene narrative responsibility attributes. Plot tasks whose function is assigned to conflict traction and whose position of action points to conflict emergence, conflict escalation, and conflict relief are extracted as conflict narrative responsibility attributes. Plot tasks whose function is assigned to dialogue expression and whose position of action points to dialogue object, dialogue purpose, and dialogue emotion are extracted as dialogue narrative responsibility attributes. According to the narrative responsibility attributes, the plot tasks in the plot task data are respectively determined to character task positions, event task positions, scene task positions, conflict task positions, and dialogue task positions, and the plot task, narrative responsibility attributes, task position assignment, and plot sequence relationship are recorded to generate plot task position data.

[0073] S2.2. Pair the plot task data with the specific responsibilities of each agent, bind the pairing relationships, and determine the plot agents for the film and television plot.

[0074] Specifically, the matching criteria are based on the plot tasks, narrative attributes, task affiliation, and plot sequence in the plot task data. The names, contents, and available task positions registered by each agent are verified. Character task positions correspond to character setup responsibilities, event task positions to event progression responsibilities, scene task positions to scene connection responsibilities, conflict task positions to conflict facilitation responsibilities, and dialogue task positions to dialogue expression responsibilities. When the name, content, and available task position all match the task affiliation, a unique pairing result is established between the plot tasks in the plot task data and the corresponding agents. The plot tasks, narrative attributes, task affiliation, agents, and task names are recorded in the same pairing relationship to complete the pairing binding and determine the plot agents for the film / TV storyline.

[0075] It should be noted that when two or more narrative agents meet the criteria of consistent responsibility item name, responsibility item content, and available task position with corresponding task position affiliation, priority pairing is performed based on the narrative responsibility attribute, role position, and plot sequence relationship in the narrative task position data. Narrative agents whose responsibility item content and narrative responsibility attribute correspond completely, whose available task position and task position affiliation are stable, and whose narrative responsibilities and plot sequence relationships are continuously connected are identified as priority pairing agents. If the task position affiliation in the narrative task position data cannot completely match the responsibility item name, responsibility item content, and available task position, the narrative agent is not directly determined. Instead, the narrative task is further subdivided into sub-narrative tasks corresponding to character development, event progression, scene transition, conflict elicitation, and dialogue expression according to functional affiliation and role position. These sub-narrative tasks are then re-included in the responsibility item-specific pairing. If the responsibility item-specific pairing still cannot be completed after subdivision, the corresponding narrative task is marked as a narrative task requiring supplementary responsibility item, and the process is reverted to the creative constraints and plot purpose division in the film and television narrative creation requirements for responsibility item supplementation. The narrative agent for the film and television narrative is determined only after the responsibility item supplementation is completed.

[0076] Furthermore, the duty item names, duty item contents, and available task positions registered by each agent are derived from the agent's pre-set duty information. This pre-set duty information is compiled from historical film and television plot creation samples, labeled plot duty types, and agent-verified plot processing records. The duty item names indicate the character setup, event progression, scene transition, conflict facilitation, and dialogue expression undertaken by each agent. The duty item contents limit the types of plot tasks that each agent can handle. Available task positions limit the character task positions, event task positions, scene task positions, conflict task positions, and dialogue task positions that each agent can match. The plot usage classification in the film and television plot creation requirements and plot task data is used for matching and verification with the agent's pre-set duty information and is not used as the source of each agent's capabilities. When the task position attribution in the plot task position data is consistent with the available task positions, and the narrative undertaking attribute is consistent with the duty item contents, the corresponding agent is identified as the plot agent of the film and television plot.

[0077] Furthermore, when two or more narrative agents simultaneously satisfy the conditions of having the same duty item name, duty item content, and available task position, the narrative undertaking attributes corresponding to the duty item content are first checked item by item, and the narrative agent that can completely correspond to all narrative undertaking attributes of the current narrative task is given priority. When multiple narrative agents can correspond to the same narrative undertaking attributes, the number of records with the same task position and processing results consistent with the labeled narrative duty type is counted from the agent's verified narrative processing records, and the number of all verified narrative processing records corresponding to the same task position is counted, and the narrative agent with the higher proportion of the aforementioned consistent records is given priority. When the aforementioned proportions are consistent, the causal relationship between the current narrative task and the adjacent narrative tasks is checked along the narrative sequence relationship in the narrative task data, and the narrative agent whose duty item content can take over the current narrative task's position and maintain the narrative sequence relationship remains unchanged is given priority. When they are still consistent, the narrative agents are determined according to the registration order in the agent's preset duty information, so that the duty item exclusive pairing can form a unique corresponding narrative agent.

[0078] S2.3. Divide the role of the story agent into roles and output the role data of the agent.

[0079] Specifically, for a plot agent related to a film or television plot, the plot tasks, narrative attributes, task affiliations, and responsibility names corresponding to the same pairing relationship are used to define the plot responsibilities undertaken by the plot agent. Character tasks correspond to character development responsibilities, event tasks correspond to event progression responsibilities, scene tasks correspond to scene connection responsibilities, conflict tasks correspond to conflict facilitation responsibilities, and dialogue tasks correspond to dialogue expression responsibilities. When the same plot agent is assigned to multiple task affiliations, multiple plot responsibilities are retained according to the plot sequence, and the division results between plot agents, plot responsibilities, plot tasks, narrative attributes, task affiliations, and responsibility names are recorded, and agent responsibility data is output.

[0080] Furthermore, when the same plot agent corresponds to multiple task positions, the responsibilities of character setup, event progression, scene connection, conflict elicitation, and dialogue expression are not merged into a single plot responsibility. Instead, independent plot responsibilities are retained according to the plot task, narrative attributes, and task position affiliation. When multiple plot responsibilities act on the same plot task segment, the processing order is determined according to the plot sequence. The plot responsibility that comes first in the plot sequence and provides character identity, event cause, scene location, conflict motivation, or dialogue purpose for subsequent plot tasks is the leading responsibility, while other plot responsibilities are supporting responsibilities. When there is content overlap between multiple plot responsibilities, the overlapping content is limited to the corresponding plot task's position of action, and content outside the corresponding position of action is not included in the corresponding plot responsibility.

[0081] S2.4. Based on the agent's responsibility data, classify the plot tasks corresponding to the plot responsibilities as allowed reasoning segments, and classify the plot tasks and plot sequence relationships that do not correspond to the plot responsibilities as restricted reasoning segments.

[0082] Specifically, the system reads the plot responsibilities, plot tasks, narrative attributes, and task positions of each plot agent from the agent responsibility data. Using plot responsibilities as the segmentation benchmark, it checks the narrative attributes and task positions of each plot task item by item. When both the narrative attributes and task positions of a plot task fall within the processing scope of the corresponding plot responsibility, the plot task is assigned to the allowed reasoning segment of the corresponding plot agent, and the original plot sequence relationship of the plot task in the plot task data is preserved. When either the narrative attributes or task position of a plot task exceeds the processing scope of the corresponding plot responsibility, the plot task is assigned to the restricted reasoning segment of the corresponding plot agent, and the plot tasks connecting the restricted reasoning segment and the plot sequence relationships that cause the plot agent to enter non-corresponding plot responsibilities are also assigned to the restricted reasoning segment. The allowed reasoning segment limits the plot tasks that the plot agent can participate in and the plot text content that can be generated around the corresponding plot task. The restricted reasoning segment limits the plot tasks that the plot agent cannot enter and the plot sequence relationships that cannot be changed, so that each plot agent corresponds to a clear allowed reasoning segment and restricted reasoning segment.

[0083] S2.5 Configure outbound entry restrictions and outbound rollback restrictions for allowed and restricted reasoning segments according to the needs of film and television plot creation, and output capability boundary data.

[0084] Specifically, the process involves reading the plot theme, target length, creative restrictions, plot sequence, and boundary restrictions from the film and television plot creation requirements, and then matching each item with the allowed and restricted reasoning segments formed in S2.4. The plot content corresponding to the plot tasks within the allowed reasoning segments is checked against the plot theme, and the plot sequence of the plot tasks within the allowed reasoning segments is checked against the plot sequence in the film and television plot creation requirements to confirm the range of plot tasks that the plot agent can enter. The plot tasks within the restricted reasoning segments are checked against the creative restrictions and boundary restrictions, and plot tasks that the plot agent cannot enter and plot sequences that cannot be changed are configured as boundary-crossing restrictions. Based on the position of the plot tasks within the restricted reasoning segments in the plot sequence, the plot task position corresponding to when the plot agent enters the restricted reasoning segment is determined, and the position where the plot text generated by the plot agent first appears beyond the allowed reasoning segment content is determined as the plot text position. The plot task position and plot text position are configured as boundary-crossing fallback restrictions. The allowed reasoning segments, restricted reasoning segments, boundary-crossing restrictions, and boundary-crossing fallback restrictions are mapped to the corresponding plot agents and plot responsibilities, and the capability boundary data is output.

[0085] It should be noted that whether the plot text deviates from the allowed reasoning segment is determined jointly by semantic consistency, entity consistency, and event relationship consistency. First, the character names, character identities, scene names, event content, conflict objects, and dialogue objects in the plot text are uniformly named with the corresponding content registered in the film and television plot creation requirements. The plot content objectives, functional affiliations, roles, and plot sequence relationships of the allowed reasoning segment corresponding to the plot task are read from the plot task data. When the plot content objectives, functional affiliations, and roles expressed in the plot text are consistent with the corresponding plot task, semantic consistency is established. When the character names, character identities, character relationships, scene locations, and dialogue objects in the plot text do not exceed the limits of the corresponding plot task, entity consistency is established. When the occurrence, outcome, cause and effect, conflict direction, and plot sequence relationships of events in the plot text maintain the relationship recorded in the corresponding plot task, event relationship consistency is established. When semantic consistency, entity consistency, and event relationship consistency are all established, the plot text is determined to be in the allowed reasoning segment. If any consistency is not established, the plot text is determined to deviate from the allowed reasoning segment, and the position of the plot text that produces inconsistent content is associated with the position of the corresponding plot task and the out-of-bounds rollback limit.

[0086] Among them, the boundary prohibition restriction is used to limit the plot agent from entering the plot task within the restricted reasoning segment during the plot text reasoning process, and from changing the plot sequence relationship corresponding to the restricted reasoning segment; the boundary rollback restriction is used to determine the plot task that needs to be reprocessed according to the position of the plot task corresponding to the deviated content after the plot text is judged to deviate from the allowed reasoning segment, and to determine the range of plot text that needs to be regenerated according to the position of the corresponding plot text; the capability boundary data is formed by the correspondence between the plot agent, plot responsibility, allowed reasoning segment, restricted reasoning segment, boundary prohibition restriction and boundary rollback restriction.

[0087] S3. Based on the capability boundary data, establish boundary binding relationships between the plot tasks and the plot agents. Arrange the generation order and handover information of the boundary binding relationships according to the plot sequence of the plot tasks, and output the boundary scheduling data.

[0088] S3.1 Based on the capability boundary data, establish a boundary binding relationship between the plot tasks in the plot task data and the plot intelligence agents.

[0089] Specifically, capability boundary data is used as the boundary positioning processing object. The corresponding positions between the narrative agent, capability boundary, boundary violation restriction, allowed reasoning segment, and restricted reasoning segment are extracted from the capability boundary data. The narrative task, function attribution, position of action, and narrative sequence relationship in the narrative task data are then assigned to the corresponding allowed reasoning segment and restricted reasoning segment. When the function attribution, position of action, and narrative sequence relationship of a narrative task falls into the allowed reasoning segment of the narrative agent but does not enter the restricted reasoning segment, the narrative task in the narrative task data and the corresponding narrative agent with the capability boundary and boundary violation restriction are identified as boundary adaptation objects. When a narrative task enters the restricted reasoning segment or touches a boundary violation restriction, the boundary adaptation between the narrative task and the narrative agent is blocked. The narrative task, narrative agent, capability boundary, boundary violation restriction, and adaptation result that have passed boundary adaptation are fixed in the same relationship record, forming a boundary binding relationship.

[0090] It should be noted that the correspondence between plot tasks and allowed / restricted reasoning segments is determined jointly based on functional affiliation, position of action, plot sequence, and narrative responsibility attribute. The functional affiliation of a plot task aligns with the plot agent's responsibilities, its position is within the plot task content the agent can participate in, the plot sequence does not alter the progression order defined by the capability boundary data, and when the narrative responsibility attribute corresponds to the allowed reasoning segment, the plot task is matched to the allowed reasoning segment. If the functional affiliation of a plot task does not belong to the plot agent's responsibilities, and its position exceeds the plot task content the agent can participate in, the plot sequence will rewrite the handover information or change the plot progression order. When the narrative responsibility attribute corresponds to the restricted reasoning segment, the plot task is matched to the restricted reasoning segment. When a plot task simultaneously satisfies some conditions of both allowed and restricted reasoning segments, the restricted reasoning segment and the boundary violation restriction are used as the priority criterion, and no boundary adaptation object is established between the plot task and the plot agent.

[0091] The plot task whose boundary adaptation is blocked will not enter the current boundary binding relationship. Instead, it will return to the plot task data and be re-matched with the allowed reasoning segments of other plot agents. If it still cannot be matched, it will be split into sub-plot tasks corresponding to character development, event progression, scene connection, conflict traction and dialogue expression according to the function and position of the role. These sub-plot tasks will be re-included in the exclusive matching of duty items and ability boundary marking, so that the plot generation chain will not be broken due to the boundary adaptation blockage.

[0092] Furthermore, boundary localization processing is used to clarify whether the plot tasks in the plot task data can fall within the capability boundaries of the plot agent; it allows the plot tasks that the plot agent can participate in, the plot semantic content that can be received, and the plot text content that can be generated, corresponding to the inference segment; it restricts the plot tasks that the plot agent cannot enter, the handover information that cannot be rewritten, and the plot sequence that cannot be changed, corresponding to the inference segment; when the functional affiliation, position of action, and plot sequence of a plot task all correspond to the allowed inference segment, but not to the restricted inference segment, the plot task and the plot agent are identified as boundary adaptation objects; when the functional affiliation, position of action, or plot sequence of a plot task corresponds to the restricted inference segment, no boundary adaptation object is established.

[0093] S3.2. Arrange the plot progression order of plot tasks and plot agents in the boundary binding relationship according to the plot sequence, and generate binding order data.

[0094] Specifically, the arrangement is based on the fixed plot tasks, plot agents, capability boundaries, boundary violations, and adaptation results in the boundary binding relationships. The order of plot tasks and plot agents is adjusted according to the plot sequence in the plot task data. Plot tasks with sequential entanglement are arranged in the order of entanglement; plot tasks with causal continuity are arranged according to the relationship of cause content first and result content second; plot tasks with scene connection are arranged according to the continuous relationship of scene entry, scene stay, and scene transition. When there are multiple boundary binding relationships under the same plot sequence, the arrangement order is distinguished according to the functional affiliation and position of the plot tasks. After the plot progression order of plot tasks and plot agents is arranged, the plot tasks, plot agents, boundary binding relationships, plot sequence relationships, and arrangement order are recorded to generate binding order data.

[0095] Furthermore, when multiple boundary binding relationships within the same plot sequence occur at the same plot node, they are not directly arranged according to functional affiliation. Instead, the arrangement is checked to see if there is character status provision, scene location limitation, event cause introduction, conflict triggering, or dialogue purpose connection between plot tasks. Plot tasks that can provide character status, scene location, or event cause for other plot tasks are assigned the base arrangement number, while plot tasks that depend on character status, scene location, event cause, conflict trigger, or event result are assigned the base arrangement number sequentially. When character dialogue and conflict escalation occur simultaneously, if the dialogue purpose is triggered... If a conflict escalates, the plot task corresponding to the dialogue expression function is configured with the base arrangement number, and the plot task corresponding to the conflict traction function is configured with the base arrangement number followed by the next number. If the conflict state determines the dialogue content, the plot task corresponding to the conflict traction function is configured with the base arrangement number, and the plot task corresponding to the dialogue expression function is configured with the base arrangement number followed by the next number. If the plot task corresponding to the dialogue expression function and the plot task corresponding to the conflict traction function do not provide generation conditions for each other, they are configured with the same arrangement number, and the shared character state, scene state, and event result are retained in the handover constraint data.

[0096] S3.3. Based on the binding order data, perform handover information confirmation and prohibition information restriction on adjacent plot agents to form handover restriction data.

[0097] Specifically, the plot agents corresponding to adjacent sequences in the binding sequence data are identified as adjacent plot agents. The plot tasks, plot agents, and boundary binding relationships with the earlier sequence numbers are used as the handover source, and the plot tasks, plot agents, and boundary binding relationships with the later sequence numbers are used as the receiving objects. Fixed character states, scene states, event results, conflict states, and dialogue directions are extracted from the handover source as handover information to be transmitted between adjacent plot agents. Plot task content, plot sequence relationships, and boundary binding relationships that are limited by ability boundaries and overstepping restrictions are selected from the handover source as unalterable information that the receiving objects cannot modify. The adjacent plot agents, handover information, unalterable information, and sequence numbers are recorded together to form handover constraint data.

[0098] S3.4. Arrange and associate the binding sequence data and handover constraint data according to the plot sequence of the story mission, and output the boundary scheduling data.

[0099] Specifically, when binding sequence data and handover constraint data enter the scheduling association, the plot sequence and arrangement order number of the plot tasks are used as the association benchmark. The plot tasks, plot agents, boundary binding relationships, plot sequence relationships, and arrangement order in the binding sequence data are matched one by one with the adjacent plot agents, handover information, unmodifiable information, and arrangement order number in the handover constraint data. Plot tasks, plot agents, boundary binding relationships, handover information, and unmodifiable information under the same arrangement order number are grouped into the same scheduling position. The scheduling flow between different arrangement order numbers is preserved according to the plot sequence of the plot tasks, forming boundary scheduling data that includes plot tasks, plot agents, boundary binding relationships, plot sequence relationships, arrangement order, handover information, and unmodifiable information.

[0100] It should be noted that the boundary binding relationship fixes the plot tasks, plot agents, capability boundaries, and out-of-bounds restrictions into the same relationship record, so that multi-agent plot generation is no longer just task allocation, but boundary adaptation is completed before the plot tasks enter the constraint reasoning; the binding sequence data further organizes the arrangement order of plot tasks and plot agents according to the plot sequence, and the handover constraint data synchronously restricts the handover information and the information that prohibits modification, so that the boundary scheduling data simultaneously has task ownership, generation order, handover content, and prohibition of modification constraints, providing a clear boundary scheduling basis for the subsequent generation of film and television plot segments.

[0101] S4. Drive the plot agent through a neural network to perform constraint reasoning on the boundary scheduling data, generate film and television plot segments, configure segment ownership tags for the film and television plot segments to form tagged plot segment data, bind the tagged plot segment data with the capability boundary data to trace the source relationship, output ownership plot data, and perform over-boundary backtracking correction and plot connection synthesis on the ownership plot data according to the capability boundary data to output film and television plot data.

[0102] S4.1 Perform semantic representation transformation on the plot tasks, plot agents and boundary binding relationships in the boundary scheduling data to form boundary compilation data.

[0103] Specifically, using the plot tasks, plot agents, and boundary binding relationships in the boundary scheduling data as semantic representation transformation objects, the plot content objectives, functional affiliations, roles, and plot sequence relationships are extracted from the plot tasks and organized into plot task representation content. Plot responsibilities, available plot tasks, and corresponding order are extracted from the plot agents and organized into plot agent representation content. Capability boundaries, boundary crossing restrictions, handover information, and prohibited modification information are extracted from the boundary binding relationships and organized into boundary binding relationship representation content. The plot task representation content, plot agent representation content, and boundary binding relationship representation content are arranged according to the order in the boundary scheduling data, so that plot tasks can correspond to plot agents, and plot agents can correspond to boundary binding relationships, forming boundary compilation data.

[0104] Furthermore, semantic representation conversion is accomplished using field-based text encoding. Following the sequence numbering in the boundary scheduling data, the content representing the plot tasks is arranged in a fixed order based on plot content objectives, functional affiliation, role position, and plot sequence. The content representing the plot agents is arranged in a fixed order based on plot responsibilities, available plot tasks, and corresponding sequence. The content representing boundary binding relationships is arranged in a fixed order based on capability boundaries, boundary crossing restrictions, handover information, and prohibited modification information. Corresponding field identifiers are configured between different contents to ensure that the plot task representations, plot agent representations, and boundary binding relationship representations under the same sequence number maintain field correspondence and do not change the original plot meaning recorded in the fields. Multiple sequence numbers are arranged consecutively according to the plot sequence, enabling the plot agents to read plot tasks, plot responsibilities, capability boundaries, boundary crossing restrictions, handover information, and prohibited modification information item by item according to field identifiers and sequence, forming boundary compilation data with definite field positions and sequence relationships.

[0105] S4.2. The plot task, the receivable handover information and the unchangeable boundary information are determined from the boundary compilation data by the plot intelligence agent, and integrated into the intelligence agent compilation data according to the corresponding arrangement order of the plot intelligence agents.

[0106] Specifically, according to the arrangement order in the boundary compilation data, the plot task representation content, plot agent representation content, and boundary binding relationship representation content corresponding to the same arrangement order are read, and the handover information and prohibition information associated with the corresponding arrangement order in the boundary scheduling data are read; the plot task, plot content objective, functional affiliation, role position, and plot sequence relationship are extracted from the plot task representation content, and the functional affiliation and role position are mapped to the plot responsibilities in the plot agent representation content. When both the functional affiliation and role position fall within the processing scope of the corresponding plot responsibility, the plot task is determined as the compilation object of plot text reasoning; the character status, scene status, event result, conflict status, and dialogue direction are extracted from the corresponding handover information. Each handover message is checked against the capability boundaries and out-of-bounds restrictions in the boundary binding relationship representation. Handover messages that do not cause the plot agent to enter the restricted reasoning segment and do not trigger out-of-bounds restrictions are identified as acceptable handover messages. Capability boundaries and out-of-bounds restrictions are extracted from the boundary binding relationship representation, and the plot task content, plot sequence, and boundary binding relationship that cannot be modified are extracted from the corresponding unmodifiable information. The capability boundaries, out-of-bounds restrictions, and unmodifiable information corresponding to the current plot task are jointly identified as unmodifiable boundary information. The plot task, acceptable handover messages, and unmodifiable boundary information are recorded in the same order according to the corresponding arrangement in the plot agent representation to form agent compilation data.

[0107] Furthermore, when the function or role of a plot task exceeds the scope of its corresponding plot responsibility, the corresponding plot task will not be included in the agent compilation data of the current plot agent; when handover information causes the current plot task to enter a restricted reasoning segment or touches the boundary prohibition restriction, the corresponding handover information will not be determined as receivable handover information; the unmodifiable boundary information only retains the capability boundary, boundary restriction and unmodifiable information corresponding to the current plot task, so that the plot task, receivable handover information and unmodifiable boundary information in the agent compilation data maintain the same order and correspond, and are respectively used as the compilation object, connection maintenance content and boundary restriction content of subsequent plot text reasoning.

[0108] S4.3. Using a neural network, the plot task is used as the reasoning object, the receivable handover information is used as the connection and maintenance condition, and the unalterable boundary information is used as the constraint condition. The data compiled by the intelligent agent is used to perform plot text reasoning, output plot text, and the plot intelligent agent classifies the plot text into the corresponding plot task to form candidate plot fragment data.

[0109] Specifically, the system reads the plot tasks, receivable handover information, and unalterable boundary information according to the order in the agent's compiled data; it reads the plot content objectives, functional affiliations, roles, and plot sequence from the plot tasks as the main conditions for plot text inference; it reads the character states, scene states, event outcomes, conflict states, and dialogue directions from the receivable handover information as conditions for maintaining plot continuity; and it reads the capability boundaries, boundary crossing restrictions, and unalterable information corresponding to the current plot task from the unalterable boundary information. Among these, the allowed and restricted reasoning segments in the capability boundaries are used to limit the generation range of the plot text, the boundary crossing restrictions in the boundary crossing restrictions are used to prevent the plot text from entering the restricted reasoning segments, and the plot task content, plot sequence, and boundary binding relationships in the unalterable information are used to limit the content of the plot text that cannot be rewritten, and together they serve as the limiting conditions for plot text inference. Under the constraints of the main conditions for plot text reasoning, the neural network organizes character actions, scene changes, event progression, conflict changes, and dialogue content corresponding to the current plot task. Under the constraints of plot continuity, it continues the character state, scene state, event outcome, conflict state, and dialogue direction. Under the constraints of plot text reasoning restrictions, it excludes generated content that enters restricted reasoning segments, violates boundary restrictions, or rewrites prohibited information, and outputs plot text. After the plot text is generated, the plot content goals, functional affiliations, and positions expressed in the plot text are checked against the current plot task item by item. Plot texts that match the checks are assigned to the corresponding plot task. The plot task, plot agent, plot text, receivable handover information, and unchangeable boundary information are kept in their original order to form candidate plot fragment data.

[0110] It should be noted that the main conditions for plot text reasoning, the conditions for maintaining plot continuity, and the conditions for restricting plot text reasoning are organized into a condition input template. The condition input template includes a task field, a connection field, and a restriction field. The task field contains the plot content objectives, functional affiliation, position of action, and plot sequence of the current plot task. The connection field contains the character status, scene status, event results, conflict status, and dialogue direction in the information that can be received. The restriction field contains the allowed reasoning segments, restricted reasoning segments, boundary prohibition restrictions, and plot task content, plot sequence, and boundary binding relationships in the information that cannot be modified, as well as the information that can be obtained from the current plot task. The semantic input layer of the neural network reads the task field, connection field, and restriction field in the same order to form the plot text reasoning content corresponding to the current plot task. The condition constraint layer determines the task field and connection field as the content that needs to be satisfied when generating plot text, and determines the restricted reasoning segment, boundary prohibition restriction, and prohibition information recorded in the restriction field as restricted content. When the text generation layer generates plot text step by step, it checks the currently generated content with the restricted content corresponding to the restriction field. When the currently generated content enters the restricted reasoning segment, touches the boundary prohibition restriction, or replaces, negates, or changes the order of plot task content, plot sequence relationship, and boundary binding relationship, a restriction mask is applied to the corresponding generated content to prevent the corresponding generated content from entering the plot text, and the subsequent plot text continues to be generated under the constraints of the task field and connection field.

[0111] After the plot text is output, the verification output layer identifies the plot content objectives, functional affiliations, roles, character states, scene states, event outcomes, conflict states, and dialogue directions from the plot text. It then checks the plot content objectives, functional affiliations, and roles against the task fields; it checks the character states, scene states, event outcomes, conflict states, and dialogue directions against the connection fields; and it checks the plot task content, plot sequence, and boundary binding relationships involved in the plot text against the restriction fields. If the plot text conforms to the current plot task, continues to receive acceptable handover information, and has not entered a restricted reasoning segment, violated boundary restrictions, or rewritten prohibited information, the plot text is retained. If the plot text contains content that does not conform to the task fields, connection fields, or restriction fields, the corresponding plot text position is located. Plot texts that conform to the main conditions for plot text reasoning, plot connection maintenance conditions, and plot text reasoning restriction conditions before the located position are retained. The corresponding non-conforming content is deleted, and the original task fields, connection fields, and restriction fields remain unchanged. The neural network then regenerates the content at the corresponding plot text position to complete the plot text reasoning.

[0112] Furthermore, during neural network training, film and television plot samples with conditional input templates and boundary annotations are used as training sources. The film and television plot samples include no less than 5,000 fully annotated plot samples, with no less than 1,000 plot tasks in each of the five categories: character development, event progression, scene connection, conflict escalation, and dialogue expression. Each film and television plot sample records the plot task, the information that can be received for handover, the boundary information that cannot be modified, and the plot text that has been manually confirmed. The plot content objectives, functional affiliation, position of action, and plot sequence corresponding to the plot task are written into the task field. The character status, scene status, event result, conflict status, and dialogue direction corresponding to the information that can be received for handover are written into the connection field. The allowed reasoning segment, restricted reasoning segment, boundary violation restriction, and unmodifiable information corresponding to the boundary information that cannot be modified are written into the restriction field. The manually confirmed plot text is used as the corresponding training output, so that each film and television plot sample forms a training correspondence between the conditional input template and the manually confirmed plot text.

[0113] The neural network adopts a text generation network structure including a semantic input layer, a conditional constraint layer, a text generation layer, and a verification output layer. During training, the semantic input layer reads the task field, connection field, and restriction field of each film and television plot sample. The conditional constraint layer determines the restriction mask based on the restriction field. The text generation layer generates plot text under the constraints of the task field and connection field. The verification output layer compares the generated plot text with the corresponding manually confirmed plot text, acceptable handover information, and unalterable boundary information. The text generation difference is determined based on the content difference between the generated plot text and the manually confirmed plot text. The connection preservation difference is determined based on the content in the generated plot text where the character state, scene state, event result, conflict state, and dialogue direction are not kept continuous. The boundary rewriting difference is determined based on the content in the generated plot text that enters the restricted reasoning section, touches the boundary prohibition restriction, and rewrites the prohibited information. The text generation difference, connection preservation difference, and boundary rewriting difference are used together as the training loss, and the neural network parameters are continuously adjusted according to the changes in the training loss, so that the generated plot text gradually fits the plot task, maintains acceptable handover information, and avoids unalterable boundary information.

[0114] After each training round, independent verification samples are used to verify the plot task consistency rate, handover information continuity rate, prohibited information non-rewriting rate, and boundary compliance rate. The plot task consistency rate is calculated based on the number of samples in the generated plot text where the plot content objectives, functional affiliations, and positions of action are consistent with the corresponding plot task. The handover information continuity rate is calculated based on the number of samples where the character state, scene state, event outcome, conflict state, and dialogue direction all maintain a corresponding relationship. The prohibited information non-rewriting rate is calculated based on the number of samples where no plot task content, plot sequence, or boundary binding relationship has been rewritten. The boundary compliance rate is calculated based on the number of samples where the generated plot text is located in the allowed reasoning segment but has not entered the restricted reasoning segment and has not violated the boundary prohibition restriction. The training rounds are no less than thirty rounds. When the boundary compliance rate corresponding to the independent verification samples is no less than 95% and the prohibited information non-rewriting rate is no less than 98%, the neural network training is completed. The trained neural network is used to perform the aforementioned plot text reasoning on the data compiled by the agent.

[0115] The sample size, training rounds, boundary compliance rate, and rate of non-modification of prohibited information were determined based on the offline training and calibration results of the neural network. Before training the neural network, film and television plot samples were organized according to five categories of plot tasks: character development, event progression, scene connection, conflict escalation, and dialogue expression. Independent validation samples were used to continuously validate the plot task consistency rate, handover information continuity rate, rate of non-modification of prohibited information, and boundary compliance rate corresponding to different training sample sizes and training rounds. When the increase in the number of training samples stabilized and the improvement rate of plot task consistency and boundary compliance reached the boundary constraint requirements, the corresponding sample size was determined as the training sample size. When the decrease in training loss stabilized and the plot task consistency rate, handover information continuity rate, rate of non-modification of prohibited information, and boundary compliance rate corresponding to the independent validation samples no longer continued to improve, the corresponding training round was determined as the training completion round. The boundary compliance rate and the rate of non-modification of prohibited information were determined based on the number of plot texts that met the capability boundaries and the number of plot texts that did not modify prohibited information in the independent validation samples, respectively.

[0116] S4.4. Perform boundary compliance verification on the candidate plot segment data according to the boundary scheduling data, retain the candidate plot segment data that conforms to the boundary binding relationship, and generate film and television plot segments.

[0117] Specifically, when candidate plot segment data enters the boundary compliance verification, the plot task, plot agent, boundary binding relationship, plot sequence, handover information, and unalterable information in the boundary scheduling data are used as the verification basis. The plot task, plot agent, plot text, receivable handover information, and unalterable boundary information in the candidate plot segment data are checked item by item. If the plot content goal, functional affiliation, and position of the plot text are consistent with the plot task in the boundary scheduling data, the plot agent to which the plot text belongs is consistent with the plot agent in the boundary scheduling data, the plot text does not exceed the capability boundary and overstepping limit defined by the boundary binding relationship, the plot text can continue the handover information in the boundary scheduling data, and the plot text does not rewrite the unalterable information in the boundary scheduling data, the candidate plot segment data is determined to comply with the boundary binding relationship. If the plot text deviates from the plot task, the plot agent affiliation is inconsistent, the capability boundary is crossed, the overstepping limit is touched, the handover information is broken, or the unalterable information is rewritten, the candidate plot segment data is excluded from the film and television plot segments. After the boundary compliance verification is completed, the candidate plot segment data that comply with the boundary binding relationship is retained and film and television plot segments are generated.

[0118] Furthermore, when determining whether the plot text has exceeded the capability boundaries and overstepping limits, the plot content objectives, functional affiliations, positions of action, character states, scene states, event outcomes, conflict states, and dialogue directions are extracted from the plot text and checked against the capability boundaries, overstepping limits, allowed reasoning segments, and restricted reasoning segments in the boundary binding relationship. If the plot content objectives, functional affiliations, and positions of action in the plot text all fall within the allowed reasoning segments, and the character states, scene states, event outcomes, conflict states, and dialogue directions all belong to acceptable handover information, and the plot text does not contain plot tasks in the restricted reasoning segments, has not rewritten prohibited information, and has not touched the plot sequence relationships, handover information, and boundary binding relationships corresponding to overstepping limits, then the plot text is determined not to have exceeded the capability boundaries and overstepping limits defined by the boundary binding relationship.

[0119] Furthermore, before performing boundary compliance verification on the plot text in the candidate plot fragment data, the plot semantic elements in the plot text are identified according to part-of-speech tagging, syntactic dependency relations, and plot semantic role determination. The plot content objectives, functional attribution, position of action, character status, scene status, event results, conflict status, and dialogue direction are obtained from the plot text. According to the verification method of semantic consistency, entity consistency, and event relationship consistency, the plot content objectives, functional attribution, and position of action are checked against the plot tasks in the boundary scheduling data item by item. The character status, scene status, event results, conflict status, and dialogue direction are checked against the acceptable handover information item by item. The plot sequence relationship, handover information, and boundary binding relationship are checked against the unalterable boundary information item by item. If all verifications are consistent and the plot text has not entered the restricted reasoning section or touched the boundary restriction, the corresponding candidate plot fragment data is retained. If any verification result is inconsistent, the corresponding candidate plot fragment data is excluded from the film and television plot fragments.

[0120] S4.5 Configure segment ownership tags for the plot agents, plot tasks and boundary binding relationships corresponding to the film and television plot segments, and attach the segment ownership tags to the film and television plot segments to form tagged plot segment data.

[0121] Specifically, after a film or television plot segment is generated, the corresponding plot text, plot agent, plot task, and boundary binding relationship are read. Plot agent tags are configured according to the names and order of plot agents in the boundary scheduling data; plot task tags are configured according to the plot content objectives, functional affiliations, and roles of plot tasks in the boundary scheduling data; and boundary binding relationship tags are configured according to the capability boundaries, boundary crossing restrictions, handover information, and prohibition-of-modification information in the boundary binding relationship. The plot agent tags, plot task tags, and boundary binding relationship tags are collectively used as segment ownership tags and written into the corresponding film or television plot segment, ensuring a one-to-one correspondence between the film or television plot segment and the plot agent, plot task, and boundary binding relationship, thus forming tagged plot segment data.

[0122] S4.6. Perform boundary indexing on the ownership markers of the marked story segments to form segment boundary index data.

[0123] Specifically, the ownership markers of the plot segments in the data are used as the entry point for boundary indexing. The plot agent markers, plot task markers, and boundary binding relationship markers in the ownership markers are read. The plot agent corresponding to the plot segment is located according to the plot agent marker, the plot task corresponding to the plot segment is located according to the plot task marker, and the boundary binding relationship corresponding to the plot segment is located according to the boundary binding relationship marker. The associated positions of capability boundaries, boundary restrictions, handover information, and prohibition information are extracted from the boundary binding relationships. The plot agent, plot task, boundary binding relationship, capability boundaries, boundary restrictions, handover information, and prohibition information are then indexed and correlated with the plot segments to form the segment boundary index data.

[0124] S4.7 Align and verify the fragment boundary index data with the capability boundary positions and out-of-bounds restriction positions in the capability boundary data to form traceability alignment data.

[0125] Specifically, when the fragment boundary index data enters the alignment verification, the plot agent, plot task, boundary binding relationship, capability boundary, and overstepping limit associated with the film and television plot fragment in the fragment boundary index data are used as the verification objects. The capability boundary data is used to find the capability boundary position that matches the plot agent, and the plot task is checked to see if it falls into the allowed inference segment, whether the boundary binding relationship corresponds to the capability boundary, and whether the overstepping limit corresponds to the overstepping limit position. When the plot task falls into the allowed inference segment and the boundary binding relationship, capability boundary, and overstepping limit are all consistent with the capability boundary position and overstepping limit position in the capability boundary data, the alignment result between the fragment boundary index data and the capability boundary data is recorded. When the plot task does not fall into the allowed inference segment, or the boundary binding relationship, capability boundary, and overstepping limit are inconsistent with the capability boundary position and overstepping limit position, the deviation result between the fragment boundary index data and the capability boundary data is recorded. The alignment result, deviation result, capability boundary position, and overstepping limit position are summarized to form the source alignment data.

[0126] S4.8. Encapsulate the source alignment data and marked plot fragment data into ownership-based data and output ownership-based plot data.

[0127] Specifically, the ownership encapsulation uses fragment ownership tags as the basis for correspondence. It maps the film and television plot fragments and fragment ownership tags in the marked plot fragment data to the alignment results, deviation results, capability boundary positions, and out-of-bounds restriction positions in the source alignment data. The plot agent tags in the fragment ownership tags correspond to the capability boundary positions, the plot task tags in the fragment ownership tags correspond to the alignment results or deviation results, and the boundary binding relationship tags in the fragment ownership tags correspond to the out-of-bounds restriction positions. When the correspondence is consistent, the film and television plot fragments, fragment ownership tags, alignment results, capability boundary positions, and out-of-bounds restriction positions are associated and saved. When the correspondence is inconsistent, the film and television plot fragments, fragment ownership tags, deviation results, capability boundary positions, and out-of-bounds restriction positions are associated and saved. The output is ownership plot data that can be used to look up capability boundary positions and out-of-bounds restriction positions through fragment ownership tags.

[0128] Furthermore, the film and television plot segments in the ownership plot data maintain a one-to-one correspondence with the segment ownership tags; the segment ownership tags are used to backtrack the plot agents, plot tasks, and boundary binding relationships; the alignment results are used to indicate that the film and television plot segments are consistent with the capability boundary data, and the deviation results are used to indicate that the film and television plot segments are inconsistent with the capability boundary data; the capability boundary position and the over-limit position are used for subsequent over-limit position backtracking, so that the ownership plot data can directly support the formation of the corrected plot data.

[0129] S4.9. Based on the capability boundary data, backtrack the ownership plot data to identify the ownership plot fragments that exceed the capability boundary data, and return the ownership plot fragments to the corresponding plot agent for limited reasoning correction to form corrected plot data.

[0130] Specifically, when ownership plot data enters a boundary-crossing position for backtracking, the capability boundary position and boundary-crossing restriction position retrieved from the fragment ownership marker are used as the backtracking basis. The film and television plot fragments in the ownership plot data are compared with the allowed reasoning segments, restricted reasoning segments, and boundary-crossing back-off restrictions in the capability boundary data. If the plot text of the film and television plot fragment falls into the allowed reasoning segment and does not touch the boundary-crossing restriction position, the film and television plot fragment is retained. If the plot text of the film and television plot fragment enters the restricted reasoning segment, or if the plot text rewrites the handover information, plot sequence, and boundary binding relationship corresponding to the boundary-crossing restriction position, the film and television plot fragment is positioned as an ownership plot fragment exceeding the capability boundary data. The corresponding plot agent is determined according to the fragment ownership marker in the ownership plot fragment, and the plot task position, handover information position, and plot text position that need to be corrected are determined according to the boundary-crossing back-off restrictions in the capability boundary data. The corresponding plot agent retains the plot text that conforms to the capability boundary data according to the allowed reasoning segment, deletes or rewrites the plot text that exceeds the capability boundary data according to the restricted reasoning segment, and restores the rewritten handover information, plot sequence, and boundary binding relationship according to the boundary-crossing back-off restrictions, thus forming corrected plot data.

[0131] Furthermore, the restricted reasoning correction adopts the neural network execution method used for plot text reasoning. After determining the plot task position, handover information position, and plot text position that need to be corrected according to the out-of-bounds backtracking restrictions, the plot text in the ownership plot segment that is located before the backtracking position and meets the allowed reasoning segment is retained as a fixed context. The plot task and plot content goal that need to be corrected are written into the task field, the unmodified character status, scene status, event result, conflict status, and dialogue direction are written into the connection field, and the ability boundary, out-of-bounds restrictions, prohibited modification information, and the restored plot sequence relationship are written into the restriction field, thus reforming the process. Data is compiled by an agent matched with the corresponding plot agent; the corresponding plot agent uses a neural network to re-execute plot text reasoning only for plot text positions located by the boundary fallback restriction, generates corrected text, replaces the out-of-bounds content in the original ownership plot segment, and re-verifies the corrected text according to the boundary compliance check; if the corrected text still enters the restricted reasoning segment or still touches the boundary fallback restriction, the fallback position is determined according to the boundary fallback restriction and the restricted reasoning correction is re-executed until the corrected text falls into the allowed reasoning segment and does not touch the boundary fallback restriction, the original fragment ownership mark and corresponding boundary binding relationship are retained, and the corrected plot data is formed.

[0132] S4.10. Using the sequence of plot tasks and fragment ownership markers in the ownership plot data, connect and supplement adjacent plot fragments in the corrected plot data, and output the film and television plot data.

[0133] Specifically, after correcting out-of-bounds content in the revised plot data, based on the plot task sequence and segment ownership markers in the ownership plot data, consecutively numbered film and television plot segments in the revised plot data are identified as adjacent plot segments. The plot agents, plot tasks, and boundary binding relationships corresponding to adjacent plot segments are checked according to the segment ownership markers, and the continuity of character states, scene states, event results, conflict states, and dialogue directions between adjacent plot segments is compared. If there are breaks in character states, scene states, missing event results, jumps in conflict states, or incoherent dialogue directions, connecting text that conforms to the plot tasks, plot agents, and boundary binding relationships is added according to the plot task sequence. The connecting text is only used to supplement the transitions between adjacent plot segments in terms of character actions, scene changes, event impact, conflict changes, and dialogue intent; it does not rewrite the boundary binding relationships, handover information, or prohibited information corresponding to the segment ownership markers. After completing the connecting and supplementing of adjacent plot segments, the film and television plot data is output.

[0134] Furthermore, the connecting text is generated by the plot agent corresponding to the plot task to which the connecting text belongs, through a neural network used for plot text reasoning. It obtains character states, scene states, event results, conflict states, and dialogue directions from the ends of adjacent plot segments with earlier sequence numbers, and obtains corresponding content from the beginning positions of adjacent plot segments with later sequence numbers. It then identifies the missing content corresponding to character state breaks, scene state breaks, missing event results, conflict state transitions, and discontinuous dialogue directions. The plot task and plot content target corresponding to the missing content are written into the task field, and the end states of adjacent plot segments with earlier sequence numbers and the existing beginning states of adjacent plot segments with later sequence numbers are written into the connecting text field. The connection field is used to write the capability boundaries, boundary restrictions, handover information, and prohibited modification information corresponding to the fragment ownership markers into the restriction field, forming the intelligent agent compilation data for connecting and supplementing the text. The plot intelligent agent to which the connecting text belongs generates the connecting text through a neural network. The generated connecting text is checked according to the boundary compliance verification. If it conforms to the boundary binding relationship and has not rewritten the handover information and prohibited modification information, the connecting text is inserted between two adjacent plot fragments, and the fragment ownership markers are configured according to the plot task, plot intelligent agent, and boundary binding relationship to which the connecting text belongs. If the connecting text touches the boundary restriction, the corresponding connecting text is regenerated according to the limited reasoning. After completing the connecting and supplementing of all adjacent plot fragments, the film and television plot data is output.

[0135] It should be noted that, compared to the method of directly generating plot text from multiple plot agents and simply splicing it together, this solution transforms the plot tasks, plot agents, and boundary binding relationships in the boundary scheduling data into boundary compilation data. This allows plot text reasoning to unfold under the joint constraints of plot tasks, receivable handover information, and unalterable boundary information. After the candidate plot fragment data undergoes boundary compliance verification, only content that conforms to the boundary binding relationship is retained to generate film and television plot fragments, which can reduce problems such as plot agents generating out of bounds, handover information being broken, and unalterable information being rewritten. The fragment ownership tag further establishes a correspondence between film and television plot fragments and plot agents, plot tasks, and boundary binding relationships. The fragment boundary index data and source alignment data can backtrack the tagged plot fragment data to the capability boundary position and the out-of-bounds restriction position. In the out-of-bounds position backtracking and limited reasoning correction, the ownership plot data can locate the ownership plot fragments that exceed the capability boundary data, and complete the connection and supplementation of adjacent plot fragments based on the corrected plot data, thereby improving the boundary controllability of film and television plot data, the traceability of fragment sources, and the stability of plot connection.

[0136] This embodiment also provides a computer device applicable to the multi-agent film and television plot generation method based on capability boundary constraints, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the multi-agent film and television plot generation method based on capability boundary constraints as proposed in the above embodiment.

[0137] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0138] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the multi-agent film and television plot generation method based on capability boundary constraints as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0139] In summary, this invention improves the boundary controllability of multi-agent film and television plot segments through boundary constraint reasoning and segment ownership labeling. After semantic representation transformation, the plot tasks, plot agents, and boundary binding relationships in the boundary scheduling data enable plot agents to complete plot text reasoning within the constraints of the plot tasks to be compiled, acceptable handover information, and unalterable boundary information. After the film and television plot segments are generated, segment ownership labels are configured to ensure a clear correspondence between the plot segments and their corresponding plot agents, plot tasks, and boundary binding relationships. The labeled plot segment data is then bound to the capability boundary data for source tracing, providing continuous evidence for backtracking out-of-bounds locations, correcting constraint reasoning, and supplementing adjacent plot segments. This ensures that the film and television plot data generated by multi-agent agents maintains clear segment origins, explicit boundary constraints, and stable plot connections.

[0140] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A multi-agent film and television plot generation method based on capability boundary constraints, characterized in that: include, Collect film and television plot creation needs, perform plot semantic extraction, obtain plot semantic elements, divide the plot semantic elements into task granularities according to the narrative function and plot progression relationship of the plot semantic elements, and output plot task data; The plot task data is matched with intelligent agents according to the narrative undertaking attributes of the plot task to determine the plot intelligent agent. Allowable reasoning segments and restricted reasoning segments are marked according to the plot responsibility and cross-boundary restrictions are configured. Capability boundary data is output. The plot task is plot content that can be independently processed by the plot intelligent agent and is divided by plot semantic elements according to the narrative function and plot progression relationship. The plot responsibility is the plot processing range determined according to the plot task, narrative undertaking attributes and task position belonging of the plot intelligent agent. Based on the capability boundary data, establish boundary binding relationships between the plot tasks and the plot agents. Arrange the generation order and handover information of the boundary binding relationships according to the plot sequence of the plot tasks, and output boundary scheduling data. By driving the plot agent through a neural network to perform constraint reasoning on the boundary scheduling data, film and television plot segments are generated, and ownership tags are configured for the film and television plot segments to form tagged plot segment data. The tagged plot segment data is then bound to the capability boundary data to trace the source relationship, and ownership plot data is output. According to the capability boundary data, the ownership plot data is corrected for overstepping and plot connection and synthesized, and the film and television plot data is output.

2. The multi-agent film and television plot generation method based on capability boundary constraints as described in claim 1, characterized in that: The specific steps for obtaining the plot semantic elements are as follows: Semantic layering is performed on plot descriptions in film and television plot creation to generate plot semantic hierarchy data. Identify the narrative objects and narrative relationships corresponding to each plot description based on the plot semantic hierarchy data, and perform plot semantic positioning to output plot semantic elements.

3. The multi-agent film and television plot generation method based on capability boundary constraints as described in claim 2, characterized in that: The specific steps for outputting the story mission data are as follows: Based on the role and position of plot semantic elements in film and television plots, the narrative function of plot semantic elements is positioned to form narrative function data; The narrative function data contains plot semantic elements that have foreshadowing, causal continuity, and scene connection, which are linked together to form a plot progression link; The semantic elements of the plot are transformed into task-level data according to the plot progression chain, and the plot task data is output.

4. The multi-agent film and television plot generation method based on capability boundary constraints as described in claim 3, characterized in that: The specific steps for determining the plot intelligence agent are as follows: The narrative responsibility attribute is extracted from the plot task data, and the plot tasks in the plot task data are assigned to tasks according to the narrative responsibility attribute to generate plot task position data. The plot task data is matched with the specific responsibilities of each agent, and the matching relationship is bound to determine the plot agent for the film and television plot.

5. The multi-agent film and television plot generation method based on capability boundary constraints as described in claim 4, characterized in that: The specific steps for outputting the capability boundary data are as follows: Divide the narrative agent into narrative responsibilities and output agent responsibility data; Based on the agent's responsibility data, the plot tasks corresponding to the plot responsibilities are classified as allowed reasoning segments, and the plot tasks and plot sequence relationships that do not correspond to the plot responsibilities are classified as restricted reasoning segments. To meet the needs of film and television plot creation, configure outbound entry restrictions and outbound rollback restrictions for allowed and restricted reasoning segments, and output capability boundary data.

6. The multi-agent film and television plot generation method based on capability boundary constraints as described in claim 5, characterized in that: The boundary binding relationship involves performing boundary positioning processing on the capability boundary data to obtain the capability boundaries and over-boundary restrictions of each plot agent, and solidifying the boundary adaptation relationship between the plot tasks in the plot task data and the plot agents with corresponding capability boundaries and over-boundary restrictions.

7. The multi-agent film and television plot generation method based on capability boundary constraints as described in claim 6, characterized in that: The specific steps for generating the output boundary scheduling data are as follows: According to the plot sequence of the plot tasks, the plot tasks and plot agents in the boundary binding relationship are arranged in the order of plot progression to generate binding order data; Based on the binding order data, handover information confirmation and prohibition of modification information are performed on adjacent plot agents to form handover restriction data; The binding sequence data and handover constraint data are scheduled and associated according to the plot sequence of the story mission, and the boundary scheduling data is output.

8. The multi-agent film and television plot generation method based on capability boundary constraints as described in claim 7, characterized in that: The specific steps for forming the marked plot segment data are as follows: Semantic representation transformation is performed on the plot tasks, plot agents, and boundary binding relationships in the boundary scheduling data to form boundary compilation data; The narrative agent determines the narrative tasks, receiveable handover information, and unchangeable boundary information from the boundary compilation data, and integrates them into agent compilation data according to the corresponding arrangement order of the narrative agents. Using a neural network, the plot task is used as the reasoning object, the receivable handover information is used as the connection and maintenance condition, and the unalterable boundary information is used as the constraint condition. The data compiled by the intelligent agent is used to perform plot text reasoning, output plot text, and the plot intelligent agent classifies the plot text into the corresponding plot task to form candidate plot fragment data. The candidate plot segment data is checked for boundary compliance based on the boundary scheduling data. Candidate plot segment data that meets the boundary binding relationship is retained to generate film and television plot segments. Configure segment ownership tags for the plot agents, plot tasks, and boundary binding relationships corresponding to film and television plot segments, and attach the segment ownership tags to the film and television plot segments to form tagged plot segment data.

9. The multi-agent film and television plot generation method based on capability boundary constraints as described in claim 8, characterized in that: The specific steps for outputting the ownership storyline data are as follows: Boundary indexing is performed on the ownership markers of the marked story segments to locate the segment boundary index data. Align and verify the fragment boundary index data with the capability boundary positions and out-of-bounds restriction positions in the capability boundary data to form traceability alignment data; The source alignment data and the marked plot fragment data are encapsulated with ownership rights and output ownership plot data.

10. The multi-agent film and television plot generation method based on capability boundary constraints as described in claim 9, characterized in that: The specific steps for outputting the film and television plot data are as follows: According to the capability boundary data, the ownership plot data is traced back to the out-of-bounds position, the ownership plot fragments that exceed the capability boundary data are located, and the ownership plot fragments are returned to the corresponding plot intelligence for limited reasoning correction, forming corrected plot data; By utilizing the sequence of plot tasks and fragment ownership markers in the ownership plot data, adjacent plot fragments in the corrected plot data are connected and supplemented to output film and television plot data.

Images