Multi-type novel creation platform

Abstract

The invention discloses a multi-type novel creation platform, and relates to the technical field of literature creation. The method comprises the following steps: S1, building a basic function architecture of the multi-type novel creation platform; s2, performing personalized recommendation on the same-style novel according to the novel style of the creator; s3, when the creator creates on the multi-type novel platform, the multi-type novel platform AI performs analysis according to the style architecture of the creator; s4, after the novel created by the author is uploaded, the multi-type novel platform performs auditing; s5, reader feedback collection of the creator's novel on the multi-type novel platform, the method can accurately analyze the key point content of the novel written by the creator by collecting and processing the novel text data and constructing the story line by using the deep learning technology, and can generate a case analysis report; the creator can understand the characteristics and deficiencies of the creation according to the quantitative analysis of the complexity of the story architecture, the compactness of the plot plot and the like presented in the reports.

Description

Technical Field

[0001] The present invention belongs to the technical field of literary creation, and in particular relates to a multi-type novel creation platform. Background Art

[0002] With the application of artificial intelligence technology in the field of literature, some platforms have begun to try to conduct a certain degree of analysis on the novels of creators, but the depth and accuracy of the analysis need to be improved. The data collected for analysis are not processed carefully enough, and there is a lack of systematic data collection, organization, and text cleaning operations, resulting in uneven data quality in subsequent analysis; and in the construction of story lines, more traditional and simple methods are used, which cannot deeply explore the key information such as the distribution of plot themes and plot structure in the novel text, and it is difficult to accurately sort out the complex and diverse story structure patterns and plot development clues. In addition, when conducting structural analysis on different types of novels, there is a lack of a comprehensive and quantitative indicator system, and it is impossible to objectively and accurately measure important characteristics such as the complexity of the story structure, the compactness of the plot, and the richness of the character's personality, thereby failing to provide creators with in-depth and valuable creative reference case analysis reports. Summary of the invention

[0003] In order to solve the above technical problems, the present invention is achieved through the following technical solutions:

[0004] The present invention is a multi-type novel creation platform, comprising the following steps:

[0005] Step S1: Building the basic functional architecture of the multi-type novel creation platform;

[0006] Step S2: Personalized recommendation of novels of the same style based on the author's reading style;

[0007] Step S3: When the creator creates on the multi-genre novel platform, the multi-genre novel platform AI analyzes the author’s style architecture;

[0008] Step S4: After the novel created by the author is uploaded, the multi-type novel platform will review it;

[0009] Step S5: Collecting reader feedback on the creator's novel on multiple types of novel platforms;

[0010] The step S3 includes the following sub-steps:

[0011] Step S31, data collection preprocessing;

[0012] Step S32: The storyline construction process is as follows:

[0013] Step S321, story line construction adopts deep learning: the platform inputs the novel text data into the platform AI, allowing the platform AI to learn and mine the distribution of plot themes in the text, extract the frequency and distribution of different themes in each novel text, and sort out the story structure pattern;

[0014] Step S322, annotation training: train the annotated text to obtain the annotation results of the model for the test set sentences, and then compare them with the manually annotated real labels in the test set to calculate the accuracy, precision, and recall evaluation indicators;

[0015] Step S323, deep learning annotation, through long short-term memory network combined with CRF sequence annotation model, the sentences in the novel text are sequenced and annotated, and the sentences with plot significance, the start, development, climax and ending of the event are marked. The plot chain is constructed by connecting the annotated sentences in chronological order, and the plot development sequence and structure combination in different novels are analyzed; for the part of the novel text involving time description, the time expression recognition technology in natural language processing is used to extract and standardize it, and then the development of the story in the time dimension is sorted out according to the time nodes, and the changes in the plot at different time stages are analyzed;

[0016] The step S33: AI architecture analysis: The platform AI analyzes the story structure, plot development clues and character creation key points. For various types of novels, the platform system generates a case analysis report, and the report content is presented in text and formula numerical structure.

[0017] Furthermore, the specific contents of step S31, data collection preprocessing, are as follows:

[0018] Different types of novel text data are collected through web crawler tools, and the collected novel texts are classified and organized according to different types to establish a labeled novel data set. The collected text formats are cleaned and formatted, and the text processing technology in natural language processing is used to remove special symbols, redundant spaces, and HTML tags in the text, and convert the text into plain text format.

[0019] Furthermore, the step S321, deep learning step is as follows:

[0020] By manually annotating the sentences in the novel text, we set the "Starting Plot" label to mark the sentences where the plot begins, the "Development Plot" label to correspond to the sentences that drive the plot forward, the "Climax Plot" label to be used for the sentences where the plot reaches its climax, and the "Ending Plot" label to mark the sentences where the plot ends.

[0021] Furthermore, the specific labeling training steps in step S322 are as follows:

[0022] The annotated text data is divided into training set, validation set and test set in proportion, and the sentence vectors and corresponding manually annotated labels in the training set are input into the constructed long short-term memory network model in batches. The training is iterated according to the set training rounds. In each round of training, the model calculates the predicted label sequence based on the input data, and then calculates the loss value through the loss function. The optimizer is used to update the model parameters according to the loss value. After the model training is completed, the sentence vectors of the test set are input into the model.

[0023] Furthermore, the evaluation index formulas of the accuracy, precision and recall rate of the model training in step S322 are as follows:

[0024] The calculation formula for the model training accuracy Z is:

[0025]

[0026] Where N is the total number of sentences in the test set, and C is the number of sentences correctly labeled by the model;

[0027] The calculation formula of model training accuracy L is:

[0028]

[0029] Where P is the number of sentences predicted to be the label, and T is the number of sentences predicted to be the label and correct.

[0030] The calculation formula for the model training recall rate H is:

[0031]

[0032] Where T is the number of sentences that actually belong to the label, and A is the number of sentences that are predicted to be the correct label.

[0033] Furthermore, in step S33, the case analysis report generated by the platform system includes: the story structure complexity value, the proportion of novels of the same type, the compactness of the plot, the richness of the character's personality and the closeness of the relationship between the two characters.

[0034] The numerical analysis reported defines the story structure complexity value as C;

[0035]

[0036] The story structure i is divided into 1, 2, and 3 levels according to the complexity, 1 is a linear structure, 2 is a two-line parallel structure, and 3 is a multi-line cross-and-nested structure; n is the number of different structural parts in the story structure, L i is the complexity level corresponding to the i-th structural part, w iis the weight of the i-th structural part in the whole story;

[0037] For the statistics of different story structure patterns under a certain novel type, define the proportion of a story structure pattern i in the same type of novels as P i ;

[0038]

[0039] Where X i is the number of novels with story structure mode i, and Z is the total number of novels counted under this novel type;

[0040] The formula for plot compactness is J:

[0041]

[0042] Where T is the total length of the novel text, M is the number of plot development clues identified, and d is the average number of words between clues;

[0043] Character richness F R The calculation formula is as follows:

[0044]

[0045] In the formula, for the character R in the novel, t R is the number of different typical personality traits identified, f Rx To show the frequency of personality traits through character description sentences, F R Define the character richness value for ;

[0046] The closeness value Z of the relationship between the two characters AB The formula is:

[0047]

[0048] In the formula, for two characters A and B in the novel, x1 is the number of times they appear together in the same plot scene, S A and S B are the total number of plot scenes in which A and B participate respectively

[0049] Furthermore, the step S1 includes the following steps:

[0050] Step S11, registration and login module: supports creators to register and log in using usernames and passwords, collects creators' basic information, including age, gender, and occupation, and is compatible with multiple social account quick login methods;

[0051] Step S12, template providing module: classify and list the novel types, and prepare pre-set creation templates for different novel types;

[0052] Step S13, text editing module: when using the templates provided by the platform, it supports text input, typesetting and format adjustment functions, can set the font, font size and color format of chapter titles and body paragraphs, and adjust paragraph indents and line spacing. The text editing module integrates intelligent spelling check and grammar correction functions, and recognizes and prompts vocabulary and expression habits in different types of novels;

[0053] Step S14, creative material module: the platform creates a material library module to collect and organize material resources according to the type of novels, and the creator retrieves and calls the material content to add it into the creative content.

[0054] Furthermore, in step S2, the platform collects the creator's reading history data, records the specific information of the novels that the creator has browsed, and can judge the creator's viewing extent of a certain novel by the reading time. At the same time, it checks the collection preference data and collects readers' ratings for the novels they have read. After collecting the data, the platform uses a collaborative filtering algorithm to mine and analyze the data, find out other reader groups with similar interests to the target readers, and recommend works of interest to the target creator with reference to the novels that similar readers like.

[0055] Furthermore, the step S4 includes the following steps:

[0056] Step S41, review and release module: when the creator uploads the novel, the platform formulates a clear review rule system for multiple types of novels, and reviews the works from the aspects of text standardization, content logic and value orientation;

[0057] Step S42, AI review and manual re-examination: The system AI identifies various words in the uploaded novel text, and the vocabulary usage meets the set requirements and there are no spelling errors. The syntactic analysis determines the sentence structure and the logical connection of the sentences. During the manual review, the reviewer makes the final review decision and decides whether the work passes the review. When the work needs to be modified, the specific problems and modification suggestions will be pointed out and fed back to the creator, so that the creator can adjust and modify the work so that the novel meets the platform review standards and reaches the publishing quality requirements.

[0058] Furthermore, in step S5, the indicators of the submitted novels are displayed in charts and graphs on a visual interface, and the displayed content is the reading volume change trend. A line chart or a bar chart of the reading volume of the work in different time periods is drawn with time as the horizontal axis and the reading volume as the vertical axis. The creator can observe the trend of the chart to understand whether the reading volume of the work increases, fluctuates or decreases after the release of the work. When the reading volume decreases after the release of a chapter, it indicates that this chapter has factors that affect the reading experience. The novel platform analyzes readers' comments, extracts high-frequency keywords, and displays them in a keyword cloud map. A high frequency of keyword occurrence indicates a high degree of reader attention. By collecting readers' current comment feedback, the current chapter can be modified and improved and the creation of subsequent chapters can be optimized accordingly.

[0059] The present invention has the following beneficial effects:

[0060] 1. The present invention collects and processes novel text data, uses deep learning technology to construct story lines, and can accurately analyze the story structure, plot development clues, character creation key points and other contents of the novels written by the creators, and generate case analysis reports. Creators can understand the characteristics and advantages of their own creations and discover their shortcomings based on quantitative analysis of the complexity of the story structure, the compactness of the plot, the richness of the characters' personalities, and the closeness of the character relationships presented in these reports, so as to optimize the story structure, adjust the plot rhythm, enrich the character image, and continuously improve the quality of creation.

[0061] 2. The present invention has a clear review rule system by improving the review and release module, and controls multiple aspects such as text standards, content logic and value orientation. First, the system AI conducts preliminary vocabulary, syntax and other aspects of review, and then manual review makes the final judgment. For works that do not meet the requirements, it can also point out specific problems and modification suggestions and feedback to the creator. This dual review mechanism can effectively avoid the release of low-quality, non-standard works, and ensure that the novels presented on the platform have high text quality and positive values, which helps to improve the overall quality and image of the entire platform's novels.

[0062] 3. The present invention uses a visual interface to display relevant indicators of submitted novels and reader review feedback in the form of charts, keyword clouds, etc. Creators can intuitively see the changing trend of the reading volume of the works and understand the attractiveness of different chapters to readers. At the same time, they can know the focus of readers through high-frequency keyword clouds. Based on these feedbacks, they can accurately locate the problems that affect the reading experience in the current chapter, so as to modify and improve the current chapter in time, and optimize the content in subsequent creations to better meet readers' reading expectations, enhance interaction with readers, and improve reader stickiness.

[0063] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for describing the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0065] Figure 1 The present invention is a flowchart of a multi-type novel creation platform. DETAILED DESCRIPTION

[0066] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0067] See also Figure 1 As shown, the present invention is a multi-type novel creation platform, comprising the following steps:

[0068] Step S1: Building the basic functional architecture of the multi-type novel creation platform;

[0069] Step S2: making personalized recommendations for novels of the same style according to the author's novel style;

[0070] Step S3: When the creator creates on the multi-genre novel platform, the multi-genre novel platform AI analyzes the author’s style architecture;

[0071] Step S4: After the novel created by the author is uploaded, the multi-type novel platform will review it;

[0072] Step S5: Collecting reader feedback on the creator's novel on multiple types of novel platforms;

[0073] Step S3 includes the following sub-steps:

[0074] Step S31, data collection and processing: using a web crawler tool to collect different types of novel text data, classifying and sorting the collected novel texts according to different types, establishing a labeled novel data set, performing text cleaning and formatting operations on the collected text format, using text processing technology in natural language processing to remove special symbols, redundant spaces, and HTML tags in the text, and converting the text into a plain text format;

[0075] Step S32: The storyline construction process is as follows:

[0076] Step S321, deep learning is used to construct the story line: the platform inputs the novel text data into the platform AI, allowing the platform AI to learn and mine the distribution of plot themes in the text, extract the frequency and distribution patterns of different themes in each novel text, and sort out the story structure pattern; manually annotate the sentences in the novel text, set the "starting plot" label to mark the sentence at the beginning of the plot, the "development plot" label to correspond to the sentence that drives the plot forward, the "climax plot" label is used for the sentence where the plot reaches the climax, and the ending plot label marks the sentence that ends the plot;

[0077] Step S322, annotation training: the annotated text data is divided into a training set, a validation set, and a test set according to a proportion, and the sentence vectors and the corresponding manually annotated labels in the training set are input into the constructed long short-term memory network model in batches, and iterative training is performed according to the set training rounds. In each round of training, the model calculates the predicted label sequence according to the input data, and then calculates the loss value through the loss function, and uses the optimizer to update the model parameters according to the loss value. When the model training is completed, the sentence vectors of the test set are input into the model to obtain the model's annotation results for the test set sentences, and then compared with the manually annotated real labels in the test set, and the accuracy, precision, and recall evaluation indicators are calculated;

[0078] The calculation formula for the model training accuracy Z is:

[0079]

[0080] Where, the total number of sentences in the test set is N, and the number of sentences correctly labeled by the model is C;

[0081] The calculation formula of model training accuracy L is:

[0082]

[0083] Where, for one of the plot labels, the number of sentences predicted as that label is P, and the number of sentences predicted as that label and correct is T;

[0084] The calculation formula for the model training recall rate H is:

[0085]

[0086] In the formula, for one of the plot labels, the number of sentences that actually belong to this label is T, and the number of sentences predicted to be this label and correct is also A;

[0087] Step S323, deep learning annotation, through long short-term memory network combined with CRF sequence annotation model, the sentences in the novel text are sequenced and annotated, and the sentences with plot significance, the start, development, climax and ending of the event are marked. By connecting these annotated sentences in chronological order, a plot chain is constructed, and the plot development sequence and structure combination in different novels are analyzed; for the part of the novel text involving time description, the time expression recognition technology in natural language processing is used to extract and standardize it, and then the development of the story in the time dimension is sorted out according to the time nodes, and the changes in the plot at different time stages are analyzed;

[0088] Step S33: AI architecture analysis: The platform system generates a case analysis report based on the story architecture, plot development clues and character creation key points analyzed by the platform AI, for various types of novels. The report content is presented in the form of text and formula numerical structure, where the report numerical analysis defines the story architecture complexity value as C;

[0089]

[0090] The story structure i is divided into 1, 2, and 3 levels according to the complexity, 1 is a linear structure, 2 is a two-line parallel structure, and 3 is a multi-line cross-and-nested structure; where n is the number of different structural parts in the story structure, L i is the complexity level corresponding to the i-th structural part, w i is the weight of the i-th structural part in the whole story;

[0091] For the statistics of different story structure patterns under a certain novel type, define the proportion of a story structure pattern i in the same type of novels as P i ;

[0092]

[0093] Where X i is the number of novels belonging to story structure mode i, and Z is the total number of novels counted under this novel type;

[0094] The total length of a novel is T, the number of identified plot development clues is M, and the average number of words between clues is d. The formula for plot compactness is J:

[0095]

[0096] For the character R in the novel, the number of different typical personality traits identified is t R , the frequencies of personality traits reflected through character description sentences are f Rx , then define the character richness value as F R, character richness F R The calculation formula is as follows:

[0097]

[0098] The closeness value Z of the relationship between the two characters AB The formula is:

[0099]

[0100] In the formula, for two characters A and B in the novel, the number of times they appear together in the same plot scene is x1, and the total number of plot scenes in which A and B participate is S respectively. A and S B .

[0101] Step S1 includes the following steps:

[0102] Step S11, registration and login module: supports creators to register and log in using usernames and passwords, collects creators' basic information, including age, gender, and occupation, and is compatible with multiple social account quick login methods. The registration and login module provides creators with convenient and diverse ways to enter the platform, supports traditional username and password registration and login methods. During the registration process, the platform will collect some basic information of the creators, which will help the platform to better understand the composition characteristics of the creator group in the future, so as to provide them with more tailored services. Creators can directly use commonly used social accounts for one-click login, which greatly simplifies the login process, avoids cumbersome registration steps, and saves creators' time;

[0103] Step S12, template providing module: classify and list novel types, prepare pre-set creation templates for different novel types, and according to the creation needs and creation habits of different creators for different novel types, the platform pre-sets corresponding types of creation templates. These templates are designed according to the typical characteristics and writing structure of various novels;

[0104] Step S13, text editing module: when using the templates provided by the platform, it supports text input, typesetting and format adjustment functions, can set the font, font size and color format of chapter titles and body paragraphs, and adjust paragraph indents and line spacing. The text editing module integrates intelligent spelling check and grammar correction functions, and recognizes and prompts vocabulary and expression habits in different types of novels;

[0105] Step S14, creative material module: the platform creates a material library module to collect and organize material resources according to the type of novels, and the creator retrieves and calls the material content to add it into the creative content;

[0106] In step S2, the platform collects the creator's reading history data and records the specific information of the novels the creator has browsed. It can judge the creator's viewing extent of a certain novel by the reading time. At the same time, it checks the collection preference data and collects readers' ratings for the novels they have read. After collecting the data, the platform uses collaborative filtering algorithms to mine and analyze the data to find other reader groups with similar interests to the target readers, and recommends works of interest to the target creator with reference to the novels liked by similar readers.

[0107] Step S4 includes the following steps:

[0108] Step S41, review and release module: When the creator uploads the novel, the platform formulates a clear review rule system for multiple types of novels, and reviews the works from the aspects of text standardization, content logic and value orientation; from the perspective of content logic, review whether the plot development of the work is coherent and reasonable, whether the character image is consistent, and whether the story is natural and smooth, to prevent plot loopholes and logical confusion. In terms of value orientation, strictly review the ideas conveyed by the work to ensure that it is positive and healthy, in line with the mainstream values ​​of society, and does not spread bad, negative or even harmful ideological content

[0109] Step S42, AI review and manual re-examination: The system AI identifies various words in the uploaded novel text, and the vocabulary usage meets the set requirements and there are no spelling errors. The syntactic analysis determines the sentence structure and the logical connection of the sentences. When conducting manual review, the reviewer uses his professional review experience and understanding of the platform review standards to conduct a comprehensive review and judgment on the AI-marked parts and the entire work, and finally decides whether the work passes the review. When there are places in the work that need to be modified, the reviewer will carefully point out the specific problems and modification suggestions, and promptly feedback these feedback information to the creator, so that the creator can adjust and modify the work according to the feedback, so that the novel can meet the platform review standards and meet the publishing quality requirements, thereby ensuring that the novels published on the platform are of high quality and provide readers with high-quality reading content.

[0110] In step S5, the indicators of the submitted novels are displayed in charts and graphs on a visual interface. The displayed content is the reading volume change trend. With time as the horizontal axis and the reading volume as the vertical axis, a line chart or a bar chart of the reading volume of the work in different time periods is drawn. The creator can observe the trend of the chart to understand whether the reading volume rises, fluctuates or decreases after the work is released. When the reading volume declines after the release of a chapter, it indicates that this chapter has factors that affect the reading experience. The novel platform analyzes readers' comments, extracts high-frequency keywords, and displays them in the form of a keyword cloud map. A high frequency of keyword occurrence indicates high reader attention. The platform will continue to collect readers' current comment feedback. Creators can analyze the suggestions and opinions based on these direct feedback from readers, and make targeted modifications and improvements to the current chapter accordingly. At the same time, they can optimize the content arrangement of subsequent chapters so that their works can better meet readers' reading expectations.

[0111] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0112] The preferred embodiments of the present invention disclosed above are only used to help illustrate the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and use the present invention well. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A multi-type novel creation platform, characterized by: The steps include: Step S1: Building the basic functional architecture of the multi-type novel creation platform; Step S2: Personalized recommendation of novels of the same style based on the author's reading style; Step S3: When the creator creates on the multi-genre novel platform, the multi-genre novel platform AI analyzes the author’s style architecture; Step S4: After the novel created by the author is uploaded, the multi-type novel platform will review it; Step S5: Collecting reader feedback on the creator's novel on multiple types of novel platforms; The step S3 includes the following sub-steps: Step S31, data collection preprocessing; Step S32: The storyline construction process is as follows: Step S321, story line construction adopts deep learning: the platform inputs the novel text data into the platform AI, allowing the platform AI to learn and mine the distribution of plot themes in the text, extract the frequency and distribution of different themes in each novel text, and sort out the story structure pattern; Step S322, annotation training: train the annotated text to obtain the annotation results of the model for the test set sentences, and then compare them with the manually annotated real labels in the test set to calculate the accuracy, precision, and recall evaluation indicators; Step S323, deep learning annotation, through long short-term memory network combined with CRF sequence annotation model, the sentences in the novel text are sequenced and annotated, and the sentences with plot significance, the start, development, climax and ending of the event are marked. The plot chain is constructed by connecting the annotated sentences in chronological order, and the plot development sequence and structure combination in different novels are analyzed; for the part of the novel text involving time description, the time expression recognition technology in natural language processing is used to extract and standardize it, and then the development of the story in the time dimension is sorted out according to the time nodes, and the changes in the plot at different time stages are analyzed; The step S33: AI architecture analysis: The platform AI analyzes the story structure, plot development clues and character creation key points. For various types of novels, the platform system generates a case analysis report, and the report content is presented in text and formula numerical structure.

2. A multi-type novel creation platform according to claim 1, characterized in that: The specific contents of step S31, data collection preprocessing, are as follows: Different types of novel text data are collected through web crawler tools, and the collected novel texts are classified and organized according to different types to establish a labeled novel data set. The collected text formats are cleaned and formatted, and the text processing technology in natural language processing is used to remove special symbols, redundant spaces, and HTML tags in the text, and convert the text into plain text format.

3. A multi-type novel creation platform according to claim 1, characterized in that: The step S321, deep learning step is as follows: By manually annotating the sentences in the novel text, we set the "starting plot" label to mark the sentences where the plot begins, the "development plot" label to correspond to the sentences that drive the plot forward, the "climax plot" label to be used for the sentences where the plot reaches the climax, and the ending plot label to mark the sentences where the plot ends.

4. A multi-type novel creation platform according to claim 1, characterized in that: The specific steps of the step S322, labeling training are as follows: The annotated text data is divided into training set, validation set and test set in proportion, and the sentence vectors and corresponding manually annotated labels in the training set are input into the constructed long short-term memory network model in batches. The training is iterated according to the set training rounds. In each round of training, the model calculates the predicted label sequence based on the input data, and then calculates the loss value through the loss function. The optimizer is used to update the model parameters according to the loss value. After the model training is completed, the sentence vectors of the test set are input into the model.

5. A multi-type novel creation platform according to claim 1, characterized in that: The evaluation index formulas of the accuracy, precision and recall rate of the model training in step S322 are as follows: The calculation formula for the model training accuracy Z is: Where N is the total number of sentences in the test set, and C is the number of sentences correctly labeled by the model; The calculation formula of model training accuracy L is: Where P is the number of sentences predicted to be the label, and T is the number of sentences predicted to be the label and correct. The calculation formula for the model training recall rate H is: Where T is the number of sentences that actually belong to the label, and A is the number of sentences that are predicted to be the correct label.

6. A multi-type novel creation platform according to claim 1, characterized in that: In step S33, the case analysis report generated by the platform system includes: the story structure complexity value, the proportion of novels of the same type, the compactness of the plot, the richness of the character's personality and the closeness of the relationship between the two characters. The numerical analysis reported defines the story structure complexity value as C; The story structure i is divided into 1, 2, and 3 levels according to the complexity, 1 is a linear structure, 2 is a two-line parallel structure, and 3 is a multi-line cross-and-nested structure; where n is the number of different structural parts in the story structure, Li is the complexity level corresponding to the i-th structural part, and wi is the weight of the i-th structural part in the entire story; For statistics of different story structure patterns under a certain novel type, define the proportion of a story structure pattern i in novels of the same type as Pi; Where Xi is the number of novels with story structure mode i, and Z is the total number of novels counted under this novel type; The formula for plot compactness is J: Where T is the total length of the novel text, M is the number of plot development clues identified, and d is the average number of words between clues; The character richness FR calculation formula is as follows: In the formula, for the character R in the novel, tR is the number of different typical personality traits identified, fRx is the frequency of reflecting the personality traits through the character description sentences, and FR is the definition of the character's personality richness value; The formula for the closeness of the relationship between two characters, ZAB, is: In the formula, for two characters A and B in the novel, x1 is the number of times they appear together in the same plot scene, SA and SB are the total number of plot scenes in which A and B participate respectively.

7. A multi-type novel creation platform according to claim 1, characterized in that: The step S1 includes the following steps: Step S11, registration and login module: supports creators to register and log in using usernames and passwords, collects creators' basic information, including age, gender, and occupation, and is compatible with multiple social account quick login methods; Step S12, template providing module: classify and list the novel types, and prepare pre-set creation templates for different novel types; Step S13, text editing module: when using the templates provided by the platform, it supports text input, typesetting and format adjustment functions, can set the font, font size and color format of chapter titles and body paragraphs, and adjust paragraph indents and line spacing. The text editing module integrates intelligent spelling check and grammar correction functions, and recognizes and prompts vocabulary and expression habits in different types of novels; Step S14, creative material module: the platform creates a material library module to collect and organize material resources according to the type of novels, and the creator retrieves and calls the material content to add it into the creative content.

8. A multi-type novel creation platform according to claim 1, characterized in that: In step S2, the platform collects the creator's reading history data, records the specific information of the novels that the creator has browsed, and can judge the creator's viewing degree of a certain novel by the reading time. At the same time, it checks the collection preference data and collects readers' ratings for the novels they have read. After collecting the data, the platform uses a collaborative filtering algorithm to mine and analyze the data, find out other reader groups with similar interests to the target readers, refer to the novels that similar readers like, and recommend works of interest to the target creator.

9. A multi-type novel creation platform according to claim 1, characterized in that: The step S4 includes the following steps: Step S41, review and release module: when the creator uploads the novel, the platform formulates a clear review rule system for multiple types of novels, and reviews the works from the aspects of text standardization, content logic and value orientation; Step S42, AI review and manual re-examination: The system AI identifies various words in the uploaded novel text, and the vocabulary usage meets the set requirements and there are no spelling errors. The syntactic analysis determines the sentence structure and the logical connection of the sentences. During the manual review, the reviewer makes the final review decision and decides whether the work passes the review. When the work needs to be modified, the specific problems and modification suggestions will be pointed out and fed back to the creator, so that the creator can adjust and modify the work so that the novel meets the platform review standards and reaches the publishing quality requirements.

10. A multi-type novel creation platform according to claim 1, characterized in that: In step S5, the indicators of the submitted novel are displayed in charts and graphs on a visual interface. The displayed content is the reading volume change trend. A line chart or a bar chart of the reading volume of the work in different time periods is drawn with time as the horizontal axis and the reading volume as the vertical axis. The creator can understand whether the reading volume of the work increases, fluctuates or decreases after the release of the work by observing the trend of the chart. When the reading volume decreases after the release of a chapter, it indicates that this chapter has factors that affect the reading experience. The novel platform analyzes readers' comments, extracts high-frequency keywords, and displays them in a keyword cloud map. A high frequency of keyword occurrence indicates a high degree of reader attention. By collecting readers' current comment feedback, the current chapter can be modified and improved and the creation of subsequent chapters can be optimized accordingly.

Images