AI auxiliary literature creation system
By designing an AI-assisted literary creation system, using dimension management and multi-person collaboration modules, the problems of disconnection from the previous article and difficulty in multi-person collaboration in existing AI writing tools are solved, and logical consistency and version synchronization are achieved.
Patent Information
- Application Number
- CN202510271662.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-09
- Publication Date
- 2025-04-18
AI Technical Summary
Due to the context length limitations of the large language model and the lack of a multi-person collaboration mechanism, existing AI writing tools have caused the generated content to be disconnected from the previous article, logically inconsistent, and the collaboration of multiple authors is difficult to synchronize, resulting in confusion in versions and conflicts in content.
An AI-assisted literary creation system was designed, and through dimension management modules, creation modules, collaborative management modules and output modules, conflict detection and collaborative creation of each part of the article is realized. The system ensures the logical consistency of the generated content through multiple dimension management (digital, weighted graphical, direct correlation) and machine learning training weights, and supports multi-person collaboration through permission level management and conflict detection mechanisms.
It effectively solves the problem of disconnection from the previous article and difficulty in collaborating with multiple people, realizes the logical consistency of text blocks and the synchronization of version management, and avoids content conflicts and version confusion.
Smart Images
Figure HDA0005302902550000011 
Figure HDA0005302902550000012
Abstract
Description
Technical Field
[0001] The present invention relates to the field of artificial intelligence, and in particular to an AI-assisted literary creation system, which is suitable for AI-assisted or multi-person collaborative creation of long literary works. Background Art
[0002] In existing AI writing tools that use large language models to assist in literary creation, only a portion can be generated each time due to the context length limit of the large language model. In order to ensure the relevance of the generated content to the previous text, retrieval enhancement technology is often used to retrieve fragments related to the content to be generated in the previous text, sort them by relevance, and take a portion of them as prompt words to let the large language model create new content. This method has some problems:
[0003] 1. Missing context information:
[0004] Retrieval enhancement usually uses multi-dimensional vector extraction based on embedding models and multi-dimensional space vector retrieval algorithms to retrieve and extract previous text information, but this method can only retrieve and sort according to semantic similarity, and cannot accurately extract the key plot nodes of the previous text. This causes the generated subsequent content to be out of touch with the previous text, resulting in logical inconsistencies, plot contradictions and other problems. For example, a character has died in the previous text, but appears again in the subsequent generated content, or key plot clues are omitted.
[0005] 2. Difficulty in multi-person collaboration:
[0006] Existing writing assistance software lacks an effective multi-person collaboration mechanism and is unable to support multiple authors to participate in the development of long texts at the same time. Modifications made by different authors to the plot and character settings cannot be synchronized in real time, leading to version confusion, content conflicts and other problems. Summary of the invention
[0007] In view of the defects existing in the prior art, the purpose of the present invention is to provide an AI-assisted literary creation system, which can not only use limited context to perform conflict detection on each part of the article, but also support collaborative creation by multiple people.
[0008] In order to achieve the above objectives, the technical solution adopted by the present invention is:
[0009] An AI-assisted literary creation system solves the above problems through multiple modules:
[0010] 1. Dimension management module, which allows users to manage multiple dimensions based on three types: digital, weighted graph, and direct association.
[0011] For numeric types, dimension distances, such as time, are calculated by subtraction.
[0012] For weighted graphs, a weighted graph structure is used to represent them, and the node distance is used as the dimension distance, such as a person relationship network.
[0013] The above four types of features are sorted out and extracted based on a large number of novels, and their weights are trained through machine learning after the values are normalized.
[0014] For the direct association type, it is composed of several keywords and phrases describing the relationship between the keywords. This type does not participate in the dimensional distance calculation, but participates in the creation by prompting the user or adding AI prompt words when any keyword is hit.
[0015] 2. The creation module guides and helps users to complete the creation of text blocks according to the following steps:
[0016] Step 1: Pre-creation: The user manually writes or manually writes prompt words to guide AI to generate pre-creation text block content.
[0017] Step 2: Dimension extraction: the system guides the large language model to analyze the pre-written text block in various dimensions according to the pre-set dimensional information, and guides the user to check and complete the dimensional information.
[0018] Step 3: Context extraction and conflict detection: the system searches for relevant text blocks in the database based on the extracted dimension information, and detects in turn whether the searched blocks conflict with the current pre-authored text blocks. If there is a conflict, the system interrupts and prompts.
[0019] Steps 1 to 3 are repeated until no conflict exists, and finally the detected text block is stored in the database together with the dimension information.
[0020] 3. Collaboration management module provides user management functions and defines the permission level of each user. When a user creates a text block, the text block he creates also has the same permission level as the user; when a conflict occurs, the high-authority user can discard the low-authority text block to complete the conflict detection.
[0021] 4. Output module: The user selects a series of text blocks and puts forward output requirements. The system guides the large language model to perform continuous output based on the selected text blocks and the relationship between the dimensional information text blocks until the output content covers all the selected text blocks and meets the output requirements put forward by the user. BRIEF DESCRIPTION OF THE DRAWINGS
[0022] The present invention has the following accompanying drawings:
[0023] Figure 1 Module diagram of the present invention.
[0024] Figure 2 The invention is a flow chart of the invention. DETAILED DESCRIPTION
[0025] The present invention is further described in detail below in conjunction with the accompanying drawings.
[0026] like Figure 1 As shown, the present invention solves the above problems through multiple modules:
[0027] The dimension management module allows users to manage multiple dimensions based on three types: digital, weighted graph, and direct association.
[0028] For numeric types, dimension distances, such as time, are calculated by subtraction.
[0029] For weighted graphs, a weighted graph structure is used to represent them, and the node distance is used as the dimension distance, such as a person relationship network.
[0030] The above two types of features are sorted and extracted based on a large number of novel samples. After the values are normalized, their weights are trained through supervised machine learning methods.
[0031] like Figure 2 As shown, the creation module, the collaborative management module and the output module work together to guide the user to complete the text block creation and output the finished product according to the following steps.
[0032] Step 1 Pre-creation: Users manually write or manually write prompt words to guide AI to generate pre-creation text block content.
[0033] In step 2, dimension extraction, the system guides the large language model to analyze the characteristics of the pre-written text block in various dimensions based on the pre-set dimension information, and guides the user to check and improve the dimension information.
[0034] Step 3: Context extraction and conflict detection: the system searches for relevant text blocks in the database based on the extracted dimension information, and detects in turn whether the searched blocks conflict with the current pre-authored text blocks. If there is a conflict, the system interrupts and prompts.
[0035] Steps 1 to 3 are repeated until no conflict exists, and the detected text block is stored in the database together with the dimension information.
[0036] The collaboration management module manages the permission levels of multiple users. When a user creates a text block, the text block he creates also has the same permission level as the user; when a conflict occurs, the high-authority user can discard the low-authority text block to complete the conflict detection.
[0037] The output module allows users to select a series of text blocks and make output requirements. The system combines the output prompt words provided by the user, the time sequence relationship of the selected text blocks, and the dimension information to guide the AI to organize the selected text blocks into a continuous long text in sequence. At the same time, the system sets the permission level of all specified text blocks to the higher of the text block permission and the output user permission to avoid low-authority users from creating new text blocks that conflict with the published content after the package is released.
[0038] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.
Claims
1. An AI-assisted literary creation system, whose features include: Dimension management module: used to allow users to manage multiple dimensions based on three types: digital, weighted graph, and direct association; Creation module: guides and helps users to complete the creation of text blocks according to the steps of pre-creation, dimension extraction, context extraction, conflict detection, and archiving; Collaboration management module: provides user authority management functions; Output module: The user selects a series of text blocks and puts forward output requirements. The system guides the large language model to perform continuous output based on the selected text blocks combined with dimensional information and text block relationships until the output content covers all selected text blocks and meets the output requirements put forward by the user.
2. The dimension management module according to claim 1, characterized in that: Dimensions are based on three types: digital, weighted graph, and direct association: For numeric types, dimension distances, such as time, are calculated by subtraction; For weighted graphs, a weighted graph structure is used, and the node distance is used as the dimension distance, such as the person relationship network; The above two types of features are sorted out and extracted based on a large number of novel samples, and their values are normalized and their weights are trained through supervised machine learning methods; For the direct association type, it is composed of several keywords and phrases describing the relationship between the keywords. This type does not participate in the dimensional distance calculation, but participates in the creation by prompting the user or adding AI prompt words when any keyword is hit.
3. The creation module according to claim 1, characterized in that: Guide users to complete text block creation in the following steps: Step 1: Pre-creation: Users manually write or manually write prompt words to guide AI to generate pre-creation text block content; Step 2: Dimension extraction: the system guides the large language model to analyze the characteristics of the pre-written text block in various dimensions based on the pre-set dimension information, and guides the user to check and improve the dimension information; Step 3: Context extraction and conflict detection: the system searches for relevant text blocks in the database based on the extracted dimension information, and sequentially detects whether the searched blocks conflict with the current pre-authored text blocks. If there is a conflict, the system interrupts and prompts; Steps 1 to 3 are repeated until no conflict exists, and the detected text block is stored in the database together with the dimension information.
4. The collaborative management module as claimed in claim 1 is characterized in that Permission levels are managed for multiple users. When a user creates text, the text block he creates also has the same permission level as the user. When a conflict occurs, the high-permission user can discard the low-permission text block to complete the conflict detection.
5. The output module according to claim 1, characterized in that Allow users to select a series of text blocks for output. The system will guide AI to organize the selected text blocks into continuous long texts in sequence based on the output prompt words given by the user, the time sequence relationship of the selected text blocks, and the dimensional information of the selected text. At the same time, the permission level of all specified text blocks is set to the higher of the text permission and the output user permission, to prevent low-privilege users from creating new text blocks that conflict with the packaged content after packaging and publishing.