Multi-modal large model for generating split script

Abstract

The invention discloses a multi-modal large model for generating a split script, and relates to the technical field of split scripts, the multi-modal large model comprises a data collection module, a data processing module and a script generation module, the data collection module is used for allowing a small part of readers to read for online testing before novels are put into the market, and the script generation module is used for generating a script; the data collection module is used for collecting data of reading habits of all readers after the novel is put into the market, the data processing module is used for analyzing data collected by online testing and reading preferences of the readers, and the script generation module is used for generating a split script according to the data collected by online testing; the data collection module comprises a reader account module, a test permission module, an eyeball tracking module, a sight judgment module, a paragraph position recording module and a page stay recording module, and the method and the system have the characteristic of conforming to public preferences.

Description

Technical Field

[0001] The present invention relates to the technical field of multimodal large models, and in particular to a multimodal large model for generating storyboard scripts. Background Art

[0002] A storyboard presents the content of a script in the form of a combination of visual images and short text descriptions, thereby helping creators visualize the storyline, shot transitions, character actions, and so on. A multimodal large model refers to a deep learning model that can process data in different modalities (such as text, images, sounds, etc.), and must take into account information in multiple input forms to achieve more complex tasks. Among them, illustrations are used in the field of online novels, allowing novels to generate images through text to convey some information and enrich the reader's reading experience. Studies have shown that when reading novels, readers generally spend more time reading passages that interest them, and spend less time skimming passages that they are not interested in.

[0003] When generating storyboards, existing big models usually preload the entire novel and then automatically determine which paragraphs need to generate images. However, in actual applications, it is found that big models are inaccurate when selecting paragraphs, resulting in images being generated for some unimportant paragraphs, while important paragraphs have no corresponding images. If the author manually sets this, it will take a lot of work. At the same time, different readers have different preferences for important paragraphs and different requirements for the frequency of image appearance. Therefore, it is necessary to design a multimodal big model for generating storyboards that meets the preferences of the public. Summary of the invention

[0004] The object of the present invention is to provide a multimodal large model for generating storyboard scripts to solve the problems raised in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions: a multimodal large model for generating storyboard scripts, comprising a data collection module, a data processing module, and a script generation module. The data collection module is used to allow a small number of readers to read the novel for online testing before the novel is released on the market, and to collect data on the reading habits of all readers after the novel is released on the market. The data processing module is used to parse the data collected from the online test and to parse the readers' reading preferences. The script generation module is used to generate storyboard scripts based on the data collected from the online test, and to optimize the paragraph positions of the storyboard scripts in a targeted manner based on the readers' reading preferences.

[0006] According to the above technical solution, the data collection module includes a reader account module, a test permission module, an eye tracking module, a gaze judgment module, a paragraph position recording module, and a page stop recording module. The reader account module is electrically connected to the test permission module and the eye tracking module, the eye tracking module is electrically connected to the gaze judgment module, and the gaze judgment module is electrically connected to the paragraph position recording module; the reader account module is used to manage the reader account information and the reader's login to the website to read the novel, so as to track individual preferences and data, the test permission module is used to control the reader's permission to participate in the online test, and apply for the permission to open the camera for the reader participating in the online test, the eye tracking module is used to identify and track the reader's pupil position, the gaze judgment module is used to determine the reader's gaze stop position in the text, the paragraph position recording module is used to record the current display position of each paragraph on the screen, and the page stop recording module is used to record the reader's current reading stop page;

[0007] The data processing module includes a word count module, a time count module, an importance calculation module, and a paragraph preference analysis module. The word count module, the paragraph position recording module, and the time count module are electrically connected to the importance calculation module, and the paragraph preference analysis module is electrically connected to the page stay recording module; the word count module is used to count the number of words in each paragraph, the time count module is used to count the time the reader stays on a certain page of the novel, and the time the reader's eyes stay on each paragraph, the importance calculation module is used to calculate the importance of the paragraph according to the time the reader's eyes stay on the paragraph, and the paragraph preference analysis module is used to analyze the reader's reading interest preference for each paragraph according to the time the reader stays on a certain page of the novel;

[0008] The script generation module includes a text analysis module, a storyboard judgment module, an image generation module, a frequency adjustment module, and a personalized adjustment module. The importance calculation module and the text analysis module are electrically connected to the storyboard judgment module, the storyboard judgment module is electrically connected to the image generation module, the frequency adjustment module is electrically connected to the storyboard judgment module, and the personalized adjustment module is electrically connected to the paragraph preference analysis module. The text analysis module uses an artificial intelligence algorithm to understand the meaning of text. The storyboard judgment module is used to draw a conclusion on whether the text in the current paragraph contains a word type that can generate an image based on the analysis result. The image generation module is used to generate a storyboard script image based on the meaning of the text using an algorithm for generating images from text. The frequency adjustment module is used to set the frequency of occurrence of the storyboard script according to the reader's personal preference. The personalized adjustment module is used to adjust the paragraphs of the generated storyboard script according to the reader's historical reading preference.

[0009] According to the above technical solution, the working method of the large model is:

[0010] S1. Before the novel is released to the market, the system first conducts a small-scale online test. Readers participating in the test register and log in. The account obtains its permissions to start reading the novel. At the same time, the display position of each paragraph in the reading page is recorded. After obtaining the monitoring permission, the camera is turned on to determine the gaze position by real-time monitoring the reader's pupil focus direction, and record the time the gaze stays on each paragraph of the novel;

[0011] S2. Analyze the collected data, calculate the number of words in each paragraph and the unit dwell time of the readers' eyes in each paragraph, and feed the results back to the importance calculation module to judge the importance of the paragraph according to the unit dwell time;

[0012] S3. After the data processing is completed, the artificial intelligence algorithm is used to understand the meaning of each paragraph, determine which paragraphs contain keywords that can generate images based on the text content, and decide whether to generate a storyboard script. Based on the analysis results, the corresponding image is generated as a storyboard script and loaded next to the paragraph;

[0013] S4. After the novel is officially released on the market, readers log in to read the novel. The system finds the test reader whose historical reading data is closest to the reader's historical reading data, adjusts the paragraphs of the generated storyboard script according to this test reader, and sets the frequency of appearance of the storyboard script. The system continuously collects data and analyzes readers' reading interests and preferences.

[0014] According to the above technical solution, in S1, the specific method of recording the time the gaze stays on each paragraph of the novel is:

[0015] S1-1. Mark paragraphs in the novel that have a specified number of words or more than X. When a marked paragraph appears on a reading page, record the starting line position and the last line position of the paragraph. The space between the starting line and the last line on the reading page is the display position of the paragraph. If the starting line and the last line are not on the same page, the space between the top of the page and the last line and the space between the starting line and the bottom of the page are both the display positions of the paragraph.

[0016] S1-2. Monitor the reader's pupil position and pupil focus direction in real time, determine the reader's gaze point by analyzing the pupil's movement trajectory, set a dwell time threshold t0, and when the reader's gaze point exceeds t0 on a certain paragraph, record the dwell time of the gaze on this paragraph. If the gaze moves quickly, it is considered that the reader is browsing.

[0017] According to the above technical solution, in S2, the specific calculation method of the unit dwelling time of the reader's gaze in each paragraph is:

[0018] S2-1. When a reader is reading, the time {t1, t2, ..., ti}, where i is the paragraph number, divided by the number of words in each marked paragraph {n1, n2, ..., n i}, and the unit residence time is obtained If the unit stay time of the current paragraph is T i The larger it is, the more interested the reader is in the paragraph;

[0019] S2-2. Count the average unit dwell time of all readers participating in the test on each marked paragraph Where m is the number of readers participating in the test, ∑T i It is the sum of the unit dwell time of all readers participating in the test on the current paragraph.

[0020] According to the above technical solution, in S3, the specific method of generating the corresponding image as a storyboard script and loading it next to the paragraph is: i Sort by, ΔT i The paragraphs that are judged to be able to generate images are given priority in generating storyboards. The number of storyboards is determined based on the total number of words in the novel, and then the paragraphs that need to generate storyboards are determined. Images that match the paragraph text are generated next to the paragraphs.

[0021] According to the above technical solution, in S4, the specific method of continuously collecting data and analyzing readers' reading interests and preferences is:

[0022] S4-1, when the reader reads the novel online, the system requests to turn on the camera to identify the reader's pupils. If the reader agrees to the request, the system collects data according to the method of S1-S2;

[0023] S4-2. If the request is not approved, the page where the reader is currently reading is recorded. When a marked paragraph appears on the reading page, the time t that the reader stayed on this page of the novel is recorded. k , where k is the total number of pages in the novel, and the total number of words in this page is recorded H k ,calculate The proportion of the marked paragraphs to the total number of words on the page is obtained, Equivalent to the time t that the reader's eyes stay on the marked paragraph i If there are multiple marked paragraphs on the same page, then the multiple paragraphs on this page are marked according to Calculate t separately i If a paragraph appears on two pages, then the t calculated twice i This method can estimate the paragraphs that readers are interested in without using a camera or interfering with privacy.

[0024] According to the above technical solution, in S4, the specific method for fine-tuning the paragraphs of the generated storyboard script is as follows: compare the historical reading records of each reader participating in the test with the historical reading records of the currently logged-in reader. If there are reading records of the same novel, calculate the unit stay time T of both readers for all marked paragraphs of the same novel. i The sum of the differences is selected, and the participating reader with the smallest sum of differences is selected as the preference reference object of the currently logged-in reader. The T of each marked paragraph is i Sort by T i The paragraphs that are judged to be able to generate images are given priority in generating storyboards. If there is no reading record of the same novel, the storyboard is selected according to S3. This method can form a more personalized storyboard generation solution.

[0025] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: when the online novel is put on the market, the present invention pre-determines the important paragraphs that attract the reader's attention by capturing the time the reader's page stays on each paragraph of text, and generates storyboards for the paragraphs that can generate images without the author's manual setting, and records the reader's personal preferences for paragraphs that he likes to read, generates targeted storyboards, and sets the frequency of appearance of the storyboards according to the preferences, so as to achieve a personalized reading experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0027] Figure 1 It is a schematic diagram of the overall module structure of the present invention. DETAILED DESCRIPTION

[0028] 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.

[0029] See also Figure 1The present invention provides a technical solution: a multimodal large model for generating storyboard scripts, including a data collection module, a data processing module, and a script generation module. The data collection module is used to allow a small number of readers to read the novel for online testing before the novel is released on the market, and to collect data on the reading habits of all readers after the novel is released on the market. The data processing module is used to parse the data collected from the online test and the reading preferences of the readers. The script generation module is used to generate the storyboard script according to the data collected from the online test, and to optimize the paragraph positions of the storyboard script according to the reading preferences of the readers.

[0030] The data collection module includes a reader account module, a test authority module, an eye tracking module, a gaze judgment module, a paragraph position recording module, and a page stop recording module. The reader account module is electrically connected to the test authority module and the eye tracking module, the eye tracking module is electrically connected to the gaze judgment module, and the gaze judgment module is electrically connected to the paragraph position recording module; the reader account module is used to manage the reader account information and the reader's login to the website to read the novel, so as to track individual preferences and data, the test authority module is used to control the authority of readers participating in the online test, and apply for the permission to open the camera for the readers participating in the online test, the eye tracking module is used to identify and track the reader's pupil position, the gaze judgment module is used to determine the reader's gaze stop position on the text, the paragraph position recording module is used to record the current display position of each paragraph on the screen, and the page stop recording module is used to record the reader's current reading stop page;

[0031] The data processing module includes a word count module, a time count module, an importance calculation module, and a paragraph preference analysis module. The word count module, the paragraph position recording module, and the time count module are electrically connected to the importance calculation module, and the paragraph preference analysis module is electrically connected to the page stay recording module; the word count module is used to count the number of words in each paragraph, the time count module is used to count the time the reader stays on a certain page of the novel, and the time the reader's eyes stay on each paragraph, the importance calculation module is used to calculate the importance of the paragraph according to the time the reader's eyes stay on the paragraph, and the paragraph preference analysis module is used to analyze the reader's reading interest preference for each paragraph according to the time the reader stays on a certain page of the novel;

[0032] The script generation module includes a text analysis module, a storyboard judgment module, an image generation module, a frequency adjustment module, and a personalized adjustment module. The importance calculation module and the text analysis module are electrically connected to the storyboard judgment module, the storyboard judgment module is electrically connected to the image generation module, the frequency adjustment module is electrically connected to the storyboard judgment module, and the personalized adjustment module is electrically connected to the paragraph preference analysis module. The text analysis module uses an artificial intelligence algorithm to understand the meaning of the text. The storyboard judgment module is used to give a conclusion on whether the text of the current paragraph contains a word type that can generate an image based on the analysis result. The image generation module is used to generate a storyboard script image based on the meaning of the text using an algorithm for generating images from text. The frequency adjustment module is used to set the frequency of occurrence of the storyboard script according to the reader's personal preference. The personalized adjustment module is used to adjust the paragraphs of the generated storyboard script according to the reader's historical reading preference.

[0033] The working method of this large model is:

[0034] S1. Before the novel is released to the market, the system first conducts a small-scale online test. Readers participating in the test register and log in. The account obtains its permissions to start reading the novel. At the same time, the display position of each paragraph in the reading page is recorded. After obtaining the monitoring permission, the camera is turned on to determine the gaze position by real-time monitoring the reader's pupil focus direction, and record the time the gaze stays on each paragraph of the novel;

[0035] S2. Analyze the collected data, calculate the number of words in each paragraph and the unit dwell time of the readers' eyes in each paragraph, and feed the results back to the importance calculation module to judge the importance of the paragraph according to the unit dwell time;

[0036] S3. After the data processing is completed, the artificial intelligence algorithm is used to understand the meaning of each paragraph, determine which paragraphs contain keywords that can generate images based on the text content, and decide whether to generate a storyboard script. Based on the analysis results, the corresponding image is generated as a storyboard script and loaded next to the paragraph;

[0037] S4. After the novel is officially released to the market, readers log in to read the novel. The system finds the test reader with the closest historical reading data based on the reader's historical reading data, adjusts the paragraphs of the generated storyboard script based on this test reader, and sets the frequency of the storyboard script, continuously collects data, and analyzes the reader's reading interests and preferences;

[0038] In S1, the specific method for recording the time the gaze stays on each paragraph of the novel is as follows:

[0039] S1-1. Mark paragraphs in the novel that have a specified number of words or more than X. When a marked paragraph appears on a reading page, record the starting line position and the last line position of the paragraph. The space between the starting line and the last line on the reading page is the display position of the paragraph. If the starting line and the last line are not on the same page, the space between the top of the page and the last line and the space between the starting line and the bottom of the page are both the display positions of the paragraph.

[0040] S1-2, monitor the pupil position and pupil focus direction of the reader in real time, determine the reader's gaze point by analyzing the pupil's movement trajectory, set a dwell time threshold t0, and when the reader's gaze point exceeds t0 on a certain paragraph, record the dwell time of the gaze on this paragraph. If the gaze moves quickly, it is considered that the reader is browsing;

[0041] In S2, the specific calculation method of the unit dwell time of the reader's gaze in each paragraph is:

[0042] S2-1. When a reader is reading, the time {t1, t2, ..., t i}, where i is the paragraph number, divided by the number of words in each marked paragraph {n1, n2, ..., n i}, and the unit residence time is obtained If the unit stay time of the current paragraph is T i The larger it is, the more interested the reader is in the paragraph;

[0043] S2-2. Count the average unit dwell time of all readers participating in the test on each marked paragraph Where m is the number of readers participating in the test, ∑T i It is the sum of the unit dwell time of all readers participating in the test on the current paragraph;

[0044] In S3, the specific method of generating the corresponding image as a storyboard script and loading it next to the paragraph is as follows: for each marked paragraph, ΔT i Sort by, ΔT i The paragraphs that are judged to be able to generate images are given priority in generating storyboards. The number of storyboards is determined based on the total number of words in the novel, and then the paragraphs that need to generate storyboards are determined. Images that match the paragraph text are generated next to the paragraphs.

[0045] In S4, the specific methods for continuously collecting data and analyzing readers’ reading interests and preferences are as follows:

[0046] S4-1, when the reader reads the novel online, the system requests to turn on the camera to identify the reader's pupils. If the reader agrees to the request, the system collects data according to the method of S1-S2;

[0047] S4-2. If the request is not approved, the page where the reader is currently reading is recorded. When a marked paragraph appears on the reading page, the time t that the reader stayed on this page of the novel is recorded. k , where k is the total number of pages in the novel, and the total number of words in this page is recorded H k ,calculate The proportion of the marked paragraphs to the total number of words on the page is obtained, Equivalent to the time t that the reader's eyes stay on the marked paragraph i If there are multiple marked paragraphs on the same page, then the multiple paragraphs on this page are marked according to Calculate t separately i If a paragraph appears on two pages, then the t calculated twice i Average; this method can estimate the paragraphs that readers are interested in without using the camera and without interfering with privacy.

[0048] In S4, the specific method for fine-tuning the paragraphs of the generated storyboard script is as follows: the historical reading records of each reader participating in the test are compared with the historical reading records of the currently logged-in reader. If there are reading records of the same novel, the unit stay time T of both readers for all marked paragraphs of the same novel is calculated. i The sum of the differences is selected, and the participating reader with the smallest sum of differences is selected as the preference reference object of the currently logged-in reader. The T of each marked paragraph is i Sort by T i The paragraphs that are high and judged to be able to generate images are given priority in generating storyboards. If there is no reading record of the same novel, the storyboard is selected according to S3. This method can form a more personalized storyboard generation solution.

[0049] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0050] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A multimodal large model for generating storyboards, characterized in that: It includes a data collection module, a data processing module, and a script generation module. The data collection module is used to allow a small number of readers to read the novel for online testing before the novel is released on the market, and to collect data on the reading habits of all readers after the novel is released on the market. The data processing module is used to parse the data collected from the online test and the readers' reading preferences. The script generation module is used to generate a storyboard script based on the data collected from the online test, and to optimize the paragraph positions of the storyboard script in a targeted manner based on the readers' reading preferences.

2. A multimodal macromodel for generating storyboards according to claim 1, characterized in that: The data collection module includes a reader account module, a test permission module, an eye tracking module, a gaze judgment module, a paragraph position recording module, and a page stop recording module; the reader account module is used to manage the reader account information and the reader's login to the website to read the novel, so as to track individual preferences and data, the test permission module is used to control the reader's permission to participate in the online test, and apply for the permission to open the camera for the reader participating in the online test, the eye tracking module is used to identify and track the reader's pupil position, the gaze judgment module is used to determine the reader's gaze stop position in the text, the paragraph position recording module is used to record the current display position of each paragraph on the screen, and the page stop recording module is used to record the reader's current reading stop page; The data processing module includes a word count module, a time count module, an importance calculation module, and a paragraph preference analysis module; the word count module is used to count the number of words in each paragraph, the time count module is used to count the time the reader stays on a certain page of the novel, and the time the reader's eyes stay on each paragraph, the importance calculation module is used to calculate the importance of the paragraph according to the time the reader's eyes stay on the paragraph, and the paragraph preference analysis module is used to analyze the reader's reading interest preference for each paragraph according to the time the reader stays on a certain page of the novel; The script generation module includes a text analysis module, a storyboard judgment module, an image generation module, a frequency adjustment module, and a personalized adjustment module. The text analysis module uses an artificial intelligence algorithm to understand the meaning of text. The storyboard judgment module is used to draw a conclusion on whether the text in the current paragraph contains a word type that can generate an image based on the analysis result. The image generation module is used to generate a storyboard script image based on the meaning of the text using an algorithm for generating images from text. The frequency adjustment module is used to set the frequency of occurrence of the storyboard script based on the reader's personal preference. The personalized adjustment module is used to adjust the paragraphs of the generated storyboard script based on the reader's historical reading preference.

3. A multimodal macromodel for generating storyboards according to claim 2, characterized in that: The working method of this large model is: S1. Before the novel is released to the market, the system first conducts a small-scale online test. Readers participating in the test register and log in. The account obtains its permissions to start reading the novel. At the same time, the display position of each paragraph in the reading page is recorded. After obtaining the monitoring permission, the camera is turned on to determine the gaze position by real-time monitoring the reader's pupil focus direction, and record the time the gaze stays on each paragraph of the novel; S2. Analyze the collected data, calculate the number of words in each paragraph and the unit dwell time of the readers' eyes in each paragraph, and feed the results back to the importance calculation module to judge the importance of the paragraph according to the unit dwell time; S3. After the data processing is completed, the artificial intelligence algorithm is used to understand the meaning of each paragraph, determine which paragraphs contain keywords that can generate images based on the text content, and decide whether to generate a storyboard script. Based on the analysis results, the corresponding image is generated as a storyboard script and loaded next to the paragraph; S4. After the novel is officially released on the market, readers log in to read the novel. The system finds the test reader whose historical reading data is closest to the reader's historical reading data, adjusts the paragraphs of the generated storyboard script according to this test reader, and sets the frequency of appearance of the storyboard script. The system continuously collects data and analyzes readers' reading interests and preferences.

4. A multimodal macromodel for generating storyboards according to claim 3, characterized in that: In S1, the specific method of recording the time the gaze stays on each paragraph of the novel is: S1-1. Mark paragraphs in the novel that have more than the specified number of words X. When a marked paragraph appears on the reading page, record the starting line position and the last line position of the paragraph. The space between the starting line and the last line on the reading page is the display position of the paragraph. If the starting line and the last line are not on the same page, the space between the top of the page and the last line and the space between the starting line and the bottom of the page are both the display position of the paragraph. S1-2. Monitor the reader's pupil position and pupil focus direction in real time, determine the reader's gaze point by analyzing the pupil's movement trajectory, set a dwell time threshold t0, and when the reader's gaze point exceeds t0 on a certain paragraph, record the dwell time of the gaze on this paragraph. If the gaze moves quickly, it is considered that the reader is browsing.

5. A multimodal macromodel for generating storyboards according to claim 4, characterized in that: In S2, the specific calculation method of the unit dwelling time of the reader's gaze in each paragraph is: S2-1. When a reader is reading, the time {t1, t2, ..., t i }, where i is the paragraph number, divided by the number of words in each marked paragraph {n1, n2, ..., n i }, and the unit residence time is obtained If the unit stay time of the current paragraph is T i The larger it is, the more interested the reader is in the paragraph; S2-2. Count the average unit dwell time of all readers participating in the test on each marked paragraph Where m is the number of readers participating in the test, ∑T i It is the sum of the unit dwell time of all readers participating in the test on the current paragraph.

6. A multimodal macromodel for generating storyboards according to claim 5, characterized in that: In S3, the specific method of generating the corresponding image as a storyboard script and loading it next to the paragraph is: for each marked paragraph, ΔT i Sort by, ΔT i The paragraphs that are judged to be able to generate images are given priority in generating storyboards. The number of storyboards is determined based on the total number of words in the novel, and then the paragraphs that need to generate storyboards are determined. Images that match the paragraph text are generated next to the paragraphs.

7. A multimodal macromodel for generating storyboards according to claim 6, characterized in that: In S4, the specific method of continuously collecting data and analyzing readers' reading interests and preferences is: S4-1, when the reader reads the novel online, the system requests to turn on the camera to identify the reader's pupils. If the reader agrees to the request, the system collects data according to the method of S1-S2; S4-2. If the request is not approved, the page where the reader is currently reading is recorded. When a marked paragraph appears on the reading page, the time t that the reader stayed on this page of the novel is recorded. k , where k is the total number of pages in the novel, and the total number of words in this page is recorded H k ,calculate The proportion of the marked paragraphs to the total number of words on the page is obtained, Equivalent to the time t that the reader's eyes stay on the marked paragraph i If there are multiple marked paragraphs on the same page, then the multiple paragraphs on this page are marked according to Calculate t separately i If a paragraph appears on two pages, then the t calculated twice i Perform an average.

8. A multimodal macromodel for generating storyboards according to claim 7, characterized in that: In S4, the specific method for fine-tuning the paragraphs of the generated storyboard script is as follows: compare the historical reading records of each reader participating in the test with the historical reading records of the currently logged-in reader. If there are reading records of the same novel, calculate the unit stay time T of both readers for all marked paragraphs of the same novel. i The sum of the differences is selected, and the participating reader with the smallest sum of differences is selected as the preference reference object of the currently logged-in reader. The T of each marked paragraph is i Sort by T i The paragraphs that are high and judged to be able to generate images are given priority in generating storyboards. If there is no reading record of the same novel, the storyboard is selected according to S3.

Images