Novel generation method and device based on reverse cue word, electronic equipment and medium

Abstract

The invention provides a novel generation method and device based on a reverse cue word, electronic equipment and a medium, and the method comprises the steps: constructing the reverse cue word and inputting a large language model after a first target text is modified, and summarizing the updated content of the modified first target text, thereby improving the generation efficiency of the first target text. The general content is used for updating the Nth-level outline generating the first target body, the updated content of the Nth-level outline can reflect the overall modification condition of the first target body, associated objects of the Nth-level outline can be determined step by step by generating a relation chain, conflict judgment and updating are carried out on each associated object, and the overall modification condition of the first target body can be reflected by the updated content of the Nth-level outline. According to the method, the first target text is traced to the second target text until the first target text is traced to the text level, so that the overall modification condition of the first target text is synchronized to the second target text, all the texts do not conflict with the updated content of the first target text, the logic consistency and plot continuity of the whole novel text are ensured, and dynamic memory of modification is realized.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and particularly to a novel generation method, device, electronic device and medium based on reverse prompts. Background Art

[0002] With the continuous development of artificial intelligence technology, the novel generation technology based on artificial intelligence has gradually emerged. It can help people improve writing efficiency, so the importance of this technology has become increasingly prominent. The current method of generating novels using artificial intelligence is mainly to gradually generate multi-level outlines first, and then generate corresponding multi-chapter texts according to the last-level outline. After the text is generated, users mostly need to modify the text content so that the novel can meet their personalized creation needs. The full text of the novel needs to maintain logical consistency and plot coherence. If a user modifies the text content of one or several chapters, the plot and logic of this part may conflict with the text content of other chapters.

[0003] Currently, in this situation, users often rely on memory to view other texts that may have conflicts and then manually modify them. However, for novels with a long length and many chapters in the text, it is often difficult to timely and comprehensively discover which text content conflicts with the modified content. In addition, there are currently algorithms for consistency. For example, if the attributes of a certain character are modified, subsequent content that involves this character will inherit the modified attributes. However, this algorithm only targets the local area and cannot maintain consistency and coherence as a whole. Moreover, such algorithms mostly directly modify the basic settings or outlines, and then generate new text content according to the modified outlines. However, for the modifications made to the generated text, it is very difficult to directly synchronize them to other texts.

[0004] Therefore, the current method of generating novels through artificial intelligence has the technical problem that it is difficult to ensure the logical consistency and plot coherence of the full text after the text is modified, and improvement is needed. Summary of the Invention

[0005] Embodiments of this application provide a novel generation method, device, electronic device and medium based on reverse prompts to alleviate the technical problem that it is difficult to ensure the logical consistency and plot coherence of the full text of the current novel after the text is modified.

[0006] To solve the above technical problems, the embodiments of this application provide the following technical solutions:

[0007] This application provides a novel generation method based on reverse prompts. The novel includes N-level outlines, each upper-level outline generates at least two lower-level outlines, and each Nth-level outline generates corresponding text. N is an integer not less than 1. The method includes:

[0008] In response to a modification operation on the first target text, update the initial content of the first target text to obtain the updated content of the first target text;

[0009] Construct a reverse prompt based on the updated content of the first target text, and process the reverse prompt through a large language model to obtain a summary content of the updated content of the first target text;

[0010] Obtain a generation relationship chain formed by outlines at all levels and texts. Two objects with a generation relationship in the generation relationship chain are associated with each other. Determine the target object that generates the first target text from the generation relationship chain, and update the initial content of the target object based on the summary content to obtain the updated content of the target object;

[0011] Determine the associated object of the target object from the generation relationship chain, and use the large language model to determine whether there is a conflict between the initial content of the associated object and the updated content of the target object;

[0012] If so, update the initial content of the associated object through the large language model to obtain the updated content of the associated object, and the updated content of the associated object does not conflict with the updated content of the target object;

[0013] Update the associated object to the target object, and loop through the operations of determining the associated object of the target object from the generation relationship chain, using the large language model to determine whether there is a conflict between the initial content of the associated object and the updated content of the target object, when the judgment result is yes, updating the initial content of the associated object through the large language model to obtain the updated content of the associated object, and updating the associated object to the target object, until the target object is the second target text.

[0014] Meanwhile, an embodiment of the present application also provides a novel generation device based on reverse prompts. The novel includes N-level outlines. Each upper-level outline generates at least two lower-level outlines. Each N-level outline generates a corresponding text. N is an integer not less than 1. The device includes:

[0015] A response module, configured to update the initial content of the first target text in response to a modification operation on the first target text to obtain the updated content of the first target text;

[0016] A first obtaining module, configured to construct a reverse prompt based on the updated content of the first target text, and process the reverse prompt through a large language model to obtain a summary content of the updated content of the first target text;

[0017] A second obtaining module, configured to obtain a generation relationship chain formed by outlines at all levels and each main text, two objects with a generation relationship in the generation relationship chain are associated with each other, determine a target object that generates the first target main text from the generation relationship chain, and update the initial content of the target object based on the summary content to obtain the updated content of the target object;

[0018] A conflict judgment module, configured to determine an associated object of the target object from the generation relationship chain, and judge whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model;

[0019] A third obtaining module, configured to, if so, update the initial content of the associated object through the large language model to obtain the updated content of the associated object, and the updated content of the associated object does not conflict with the updated content of the target object;

[0020] A loop module, configured to update the associated object to the target object, and loop to execute the operations of determining the associated object of the target object from the generation relationship chain, judging whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model, when the judgment result is yes, updating the initial content of the associated object through the large language model to obtain the updated content of the associated object, and updating the associated object to the target object, until the target object is the second target main text.

[0021] The present application further provides an electronic device, including a memory and a processor; the memory stores an application program, and the processor is configured to run the application program in the memory to execute the steps in the novel generation method described in any one of the above.

[0022] An embodiment of the present application provides a computer-readable storage medium, which stores multiple instructions, and the instructions are suitable for being loaded by a processor to execute the steps in the above novel generation method.

[0023] Beneficial effects: The present application provides a novel generation method, device, electronic device and medium based on reverse prompts. The novel includes N-level outlines, each upper-level outline generates at least two lower-level outlines, and each N-level outline generates corresponding text. After modifying the first target text, the method constructs a reverse prompt using the updated content of the first target text and inputs it into a large language model to obtain a summary of the updated content of the first target text. Then, it determines the target object that generates the first target text, that is, a certain N-level outline, from the generation relationship chain formed by each level of outline and each text, and updates the initial content of the target object based on the summary to obtain the updated content of the target object. Next, it determines the associated object of the target object from the generation relationship chain, that is, a certain (N-1)-level outline. If there is a conflict between the initial content of the associated object and the updated content of the target object, it updates the initial content of the associated object to obtain the updated content of the associated object so that it does not conflict with the updated content of the target object, and then updates the associated object to the target object. Finally, it repeatedly executes the above operations of determining the associated object, judging the conflict between the associated object and the target object, updating the associated object when the judgment result is yes, and updating the associated object to the target object until the target object is the second target text, and there is a conflict between the initial content of the second target text and the updated content of the first target text. After the above steps, the updated content of the second target text does not conflict with the updated content of the first target text. After the first target text is modified, the present application can summarize the updated content of the modified first target text by constructing a reverse prompt and inputting it into a large language model, and update the N-level outline that generates the first target text with the summary. Then, the updated content of the N-level outline can reflect the overall modification of the first target text. Through the generation relationship chain, the associated objects of the N-level outline can be determined level by level, and conflict judgment and update are performed on each associated object until the text level is traced, so as to synchronize the overall modification of the first target text to the second target text, ensuring that all texts do not conflict with the updated content of the first target text. Therefore, the logical consistency and plot coherence of the entire novel text are guaranteed, and dynamic memory of the modification is achieved. Brief Description of the Drawings

[0024] The following will clearly show the technical solutions and other beneficial effects of the present application by describing the specific embodiments of the present application in detail with reference to the drawings.

[0025] Figure 1 It is a schematic diagram of the scenario of the novel generation method based on reverse prompts provided by an embodiment of the present application.

[0026] Figure 2 It is a schematic diagram of the generation relationship of the novel in an embodiment of the present application.

[0027] Figure 3It is a schematic flowchart of the novel generation method based on reverse prompts provided in the embodiments of the present application.

[0028] Figure 4 It is the first schematic diagram of the novel generation page in the embodiments of the present application.

[0029] Figure 5 It is the second schematic diagram of the novel generation page in the embodiments of the present application.

[0030] Figure 6 It is the third schematic diagram of the novel generation page in the embodiments of the present application.

[0031] Figure 7 It is the fourth schematic diagram of the novel generation page in the embodiments of the present application.

[0032] Figure 8 It is a schematic structural diagram of the novel generation setting based on reverse prompts provided in the embodiments of the present application.

[0033] Figure 9 It is a schematic structural diagram of the electronic device provided in the embodiments of the present application.

[0034] Explanation of reference numerals:

[0035] Generation device 11; large language model 12; response module 10; first obtaining module 20; second obtaining module 30; conflict judgment module 40; third obtaining module 50; loop module 60; radio frequency circuit 101; memory 102; input unit 103; display unit 104; sensor 105; audio circuit 106; WiFi module 107; processor 108. Detailed implementation manners

[0036] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of protection of the present application.

[0037] Please refer to Figure 1 , Figure 1Schematic diagram of the scenario to which the novel generation method based on reverse prompts provided by the embodiments of this application is applied. This scenario includes a generation device 11 and a large language model 12. A novel generation application / website runs in the generation device 11. A user sends an instruction to generate an N-level outline and the main text in the relevant interface of the novel generation application / website. The generation device 11 responds to each instruction, constructs various prompts, and calls the large language model 12 to process the prompts. After processing, it returns the generated N-level outline and the main text in sequence, and finally obtains the entire novel and displays it in the generation device 11.

[0038] After the novel is generated, the user performs a modification operation on the first target main text in the relevant interface of the novel generation application / website. The generation device 11 responds to this operation, updates the first target main text, obtains the updated content of the first target main text, constructs a reverse prompt based on this updated content, and sends it to the large language model 12, so that the large language model 12 generalizes this updated content to generate a generalized content, and then uses the generalized content to update the target object that generated the first target main text to obtain the updated content. A generation relationship chain is pre-set in the novel generation application / website, and the generation relationship chain records the generation relationships between each level of outline and each main text. After the above target object is updated, based on the generation relationship chain, the associated objects that have a generation relationship with the target object are determined level by level, and it is judged whether the initial content of the associated object conflicts with the updated content of the target object. If there is a conflict, the initial content of the associated object is updated to the updated content, and the associated object is determined as the target object, and the associated object of this target object is determined again based on the generation relationship chain. The above steps are repeatedly executed until the target object is the second target main text, that is, traced back to the main text level, so as to synchronize the overall modification situation of the first target main text to the second target main text.

[0039] By setting the reverse prompt and the generation relationship chain, it is ensured that all main texts do not conflict with the updated content, thus guaranteeing the logical consistency and plot coherence of the entire novel main text, and realizing the dynamic memory of the modification. In the following embodiments, the above novel generation process will be described in detail.

[0040] Please refer to Figure 2 , Figure 2This is a schematic diagram of the generation relationship of the novel in the embodiments of the present application. The novel includes N-level outlines. Each upper-level outline generates at least two lower-level outlines, and each Nth-level outline generates corresponding text. N is an integer not less than 1. After generating a certain upper-level outline, a prompt will be constructed based on the upper-level outline and sent to the large language model. After processing by the large language model, the specific content of the corresponding lower-level outline will be generated and returned. After generating a certain Nth-level outline, a prompt will also be constructed based on this outline and sent to the large language model. After processing by the large language model, the specific content of the corresponding text will be generated and returned. Since each upper-level outline generates at least two lower-level outlines, and each lowest-level outline generates corresponding text, the same first-level outline will extend to at least two chapters of text.

[0041] Figure 2 Taking N equal to 2 as an example, the novel includes two-level outlines. The first-level outline is the part outline, and the second-level outline is the chapter outline. There are 10 part outlines, Figure 2 which are represented by the 1st part to the 10th part. The part outlines are generated according to the basic settings of the novel. Each part outline generates 10 chapter outlines. In the order of generation, the 1st part outline generates the 1st to 10th chapter outlines, the 2nd part outline generates the 11th to 20th chapter outlines, and so on. Each chapter outline generates a corresponding chapter of text. In the order of generation, the 1st chapter outline generates the 1st chapter of text, the 2nd chapter outline generates the 2nd chapter of text, and so on.

[0042] According to the above generation relationship, a generation relationship chain of each level of outline and each text can be established. The highest level to the lowest level in the generation relationship chain are the first-level outline, the second-level outline,..., the Nth-level outline, and each chapter of text in turn. The levels of the first-level outline to the Nth-level outline decrease in turn. The objects in the generation relationship chain include outlines and texts. Among them, two objects with a generation relationship are mutually associated objects. For example, the 1st part outline can generate the 1st to 10th chapter outlines, so the 1st part outline and any one of the 1st to 10th chapter outlines are mutually associated objects. The 1st chapter outline can generate the 1st chapter of text, so the 1st chapter outline and the 1st chapter of text are mutually associated objects.

[0043] When establishing the generation relationship chain, a specific name can be set for each object. In the embodiments of the present application, the names of the 1st to 10th part outlines are set as lp1 to lp10 respectively, the names of the 1st to 100th chapter outlines are set as sp1 to sp100 respectively, and the names of the 1st to 100th chapter texts are set as zn1 to zn100 respectively. When the content of a certain object is newly generated, the new content will be stored at the specific name of this object for easy retrieval.

[0044] Please refer to Figure 3 , Figure 3The flowchart shows the method for generating a novel based on reverse prompts provided by an embodiment of this application. The method specifically includes:

[0045] S1: In response to a modification operation on the first target text, update the initial content of the first target text to obtain the updated content of the first target text.

[0046] The first target text can be any chapter text among all the generated text. Define the generated content of the first target text as the initial content. The user performs a modification operation on the display page of the first target text. The generation device responds to this modification operation, updates the initial content, obtains the updated content, and displays it to the user.

[0047] Exemplarily, as Figure 4 shown, the novel generation page is used to display the currently generated outlines at all levels and the corresponding text. Figure 4 Specifically shows the outline of Part 1, the outlines of Chapters 1 to 6 associated with the outline of Part 1, and the text of Chapter 1 generated from the outline of Chapter 1. It should be noted that the text of Chapter 1 may include 5 large paragraphs, and each large paragraph contains several small paragraphs. When generating the text of Chapter 1 according to the outline of Chapter 1, 5 paragraph outlines will be generated first, and then 5 large paragraphs corresponding to each paragraph outline will be generated. That is, there are also paragraph outlines between the chapter outline and the chapter text. However, since paragraph outlines can only be newly generated in the actual product and do not participate in subsequent updates after generation, their main role is to be displayed to the user for viewing. Therefore, in the following embodiments of this application, the chapter outline is still used as the lowest-level outline, and paragraph outlines are not included in the generation relationship chain.

[0048] As Figure 5 shown, the user modifies the initial content "text plot" of the first target text in the text of Chapter 1 on the novel generation page. Figure 5 Specifically, in [reference figure], the "text plot" at the end is modified to "new plot". After the modification, the generation device responds to this operation, can update the text of Chapter 1, and display the other "text plot" and the "new plot" at the end together as the updated content of the text of Chapter 1 to the user.

[0049] S2: Construct a reverse prompt based on the updated content of the first target text, and process the reverse prompt through a large language model to obtain a summary content of the updated content of the first target text.

[0050] A large language model is a language model with a relatively large number of parameters, aiming to understand and generate human language. It is trained with a large amount of text data and can perform a wide range of tasks including text summarization, translation, analysis, generation, etc. In the embodiments of this application, a prompt word that requires further refinement of a certain content is defined as a positive prompt word, and a prompt word that requires summarization of a certain content is defined as a negative prompt word. After obtaining the updated content of the first target text, a negative prompt word is constructed, which requires summarization of the updated content of the first target text. After inputting this negative prompt word into the large language model and processing it, the summary content of the updated content is returned.

[0051] S3: Obtain the generation relationship chain formed by each level of outline and each text. The two objects with a generation relationship in the generation relationship chain are mutually associated. Determine the target object that generates the first target text from the generation relationship chain, and update the initial content of the target object based on the summary content to obtain the updated content of the target object.

[0052] The generation relationship chains of each outline have been set in advance. The two outlines with a generation relationship in the generation relationship chain are mutually associated objects, and a certain outline and a certain text with a generation relationship are also mutually associated objects. It should be noted that the generation relationship in the embodiments of this application only includes direct generation and does not include indirect generation. For example, if A generates B and B generates C, then A and B have a generation relationship, B and C have a generation relationship, but A and C do not have a generation relationship.

[0053] After determining the first target text, the generation relationship chain can be searched to determine a certain Nth-level outline that generates the first target text, and use it as the target object. Since the initial content of the first target text has been updated, the initial content of the target object will not exactly match the initial content of the first target text, and there may be inconsistencies in the plot or logic. At this time, it is necessary to update the initial content of the target object based on the above summary content to obtain the updated content of the target object.

[0054] When updating the initial content of the target object, there are various methods.

[0055] In one embodiment, S3 specifically includes: covering the initial content of the target object with the summary content to obtain the updated content of the target object.

[0056] Generally, the summary content can meet the user's creation needs. Therefore, after obtaining the summary content, it can be directly returned to a specific naming place of the target object to cover its initial content. After covering, the updated content of the target object is formed.

[0057] For this direct covering method, steps S2 and S3 can be completed as a whole. An example of the content of the negative prompt word in this process is as follows:

[0058] Please concisely summarize the following {page} into a summary introduction {newsp}, presenting the plot clearly and methodically, removing rhetorical expressions, and noting the elements of the beginning, development, climax, and ending of the novel plot, as well as clarifying the character relationships and scene locations.

[0059] {page1}:

[0060] Note when modifying:

[0061] 1. Extract the core plot and key events from the text, ensuring the summary covers the main story line. Avoid omitting important plots or character actions.

[0062] 2. Delete all decorative language, figures of speech, exaggerations, etc., and retain factual descriptions.

[0063] 3. Organize the content in chronological or logical order to ensure the plot is coherent and easy to understand.

[0064] 4. Avoid subjective evaluations or emotional tones, and only state facts. The language style should be neutral, objective, and concise.

[0065] 5. Ensure the summary can independently convey the core content of the story without referring to the original text.

[0066] 6. Control the word count between 50 - 100 words. Output in JSON format.

[0067] Among them, {page} is the updated content of the first target text, {newsp} is the updated content of the target object. The process of generating the summary content and the process of the summary content covering the initial content are background processes and invisible to the user. Only the final updated content will be shown on the front - end page.

[0068] Of course, the above is only one description method of the prompt words. Those skilled in the art can make adaptive settings and modifications to the description content of the prompt words according to actual needs, as long as it enables the large - language model to complete summarization and updating.

[0069] Exemplarily, as Figure 5 shown, after modifying the text of Chapter 1, if you want to leave the current page, a pop - up window will be displayed, which prompts the user "You have adjusted the text content. Do you want to synchronize the modification to the corresponding chapter outline?" If the user selects to confirm synchronization, the background will automatically execute the steps of constructing the reverse prompt words, inputting the reverse prompt words into the large - language model, obtaining the summary content returned by the large - language model, and covering the Chapter 1 outline with the summary content. During the update process, the page is as Figure 6 shown, and after the update is completed, the page is as Figure 7As shown, the initial content "AAA..." of the outline of Chapter 1 is updated to the updated content "GGG...".

[0070] In one embodiment, S3 specifically further includes: in response to a modification operation on the summary content, updating the summary content to obtain the updated summary content; and overwriting the initial content of the target object with the updated summary content to obtain the updated content of the target object.

[0071] The summary content is completely generated by a large language model. In some cases, users may think that the summary content has defects or does not meet expectations, and directly using it as the updated content does not meet the creation requirements. To solve this technical problem, after obtaining the summary content, it can be first displayed on a relevant page so that users can see the summary content. If the user thinks that the summary content meets the creation requirements, a confirmation operation can be performed on it, and the generation device responds to this operation and overwrites the initial content of the target object with the summary content. After overwriting, the updated content of the target object is formed. If the user thinks that the summary content does not meet the creation requirements, a modification operation can be performed on it, and the generation device responds to this operation and can update the summary content to obtain the updated summary content. If the user thinks that the updated summary content meets the creation requirements, a confirmation operation can be performed on it, and the generation device responds to this operation and overwrites the initial content of the target object with the updated summary content. After overwriting, the updated content of the target object is formed.

[0072] By combining the automatic generation of the large language model and user modification, the updated content can be made more accurate and play a better reference effect in subsequent steps.

[0073] In one embodiment, before the step of updating the summary content in response to a modification operation on the summary content, it further includes: comparing and displaying the initial content and the summary content of the target object on the same page.

[0074] In the above embodiment, after generating the summary content, the initial content of the target object will not be displayed continuously, and only the summary content will be displayed on the relevant page. After the user modifies the summary content, it will be overwritten. However, in some cases, when the user modifies the summary content, they may need to refer to the initial content. For example, the user wants to know which parts of the summary content have been changed compared to the initial content. However, since the initial content is no longer displayed, the user may not remember what information the initial content recorded, which will increase the difficulty of the modification work.

[0075] To solve this technical problem, in this embodiment, after generating the summary content, the initial content and the summary content of the target object can be compared and displayed on the same page. For example, the summary content can be displayed below the initial content, which is convenient for users to compare the content of the two, know the differences between the summary content and the initial content according to the comparison, and then modify the summary content on this basis, so that the modification is smoother and more accurate, and the work efficiency is improved.

[0076] Those skilled in the art can, according to needs, choose not to display the summary content, only display the summary content, or compare and display the initial content and the summary content to meet the user needs in different scenarios.

[0077] S4: Determine the associated object of the target object from the generated relationship chain, and judge whether there is a conflict between the initial content of the associated object and the updated content of the target object through a large language model.

[0078] After determining the target object, the generated relationship chain can be searched to determine the associated object that has a generation relationship with the target object. The associated object also has initial content. Since the target object has been updated and the associated object has a generation relationship with the target object, the updated content of the target object may have a logical or plot conflict with the initial content of the associated object. To solve this problem, a corresponding prompt word is first constructed and input into the large language model, and the large language model detects and judges whether there is a conflict between the two.

[0079] S5: If so, update the initial content of the associated object through the large language model to obtain the updated content of the associated object, and the updated content of the associated object does not conflict with the updated content of the target object.

[0080] If the judgment result is yes, it means that there is a conflict between the initial content of the associated object and the updated content of the target object. To maintain the consistency and coherence between the two outlines, the initial content of the associated object needs to be updated to the content that does not conflict with the updated content of the target object, and this content is used as the updated content of the associated object.

[0081] If the judgment result is no, it means that there is no conflict between the initial content of the associated object and the updated content of the target object. Then the associated object and the lower-level outline or text of the associated object will not conflict with the updated content, so no update is required.

[0082] In an actual scenario, steps S4 and S5 can be completed as a whole. An example of the content of the prompt word in this process is as follows:

[0083] As a master novelist who pays attention to the rationality and logic of the plot, please check the following two parts, sp1' and lp1, and match the plots. If a plot that appears in sp1' is found in lp1 but the ending does not match sp1', modify the content of lp1 and save it as lp1'. Always pay attention to the consistency of the plot before and after, and maintain the coherence and logic of the story description when modifying. Output in JSON format.

[0084] Among them, sp1' is the updated content of the outline of Chapter 1, lp1 is the initial content of the outline of Part 1, and lp1' is the updated content of the outline of Part 1.

[0085] Of course, the above is only one way of describing the prompt words. Those skilled in the art can make adaptive settings and modifications to the description content of the prompt words according to actual needs, as long as it enables the large language model to complete conflict detection and update.

[0086] Exemplarily, assume that the content of sp1 includes "Xiaoming writes a letter to Li Hua and Li Hua receives the letter". After modifying the text of Chapter 1, update the content of sp1 to sp1', and the content of sp1' includes "Xiaoming writes a letter to Li Hua and Li Hua does not receive the letter". Assume that the content of lp1 in the outline of Part 1 includes "Li Hua departs for Xiaoming's home after receiving Xiaoming's letter". After conflict detection, it is found that it conflicts with the content of sp1', so lp1 will be updated to lp1', and the content of lp1' includes "Li Hua does not receive Xiaoming's message and departs after calling Xiaoming". After the update, the content of lp1' does not conflict with the content of sp1', and thus can be consistent with the modified text of Chapter 1.

[0087] S6: Update the associated object to the target object, and loop through the operations of determining the associated object of the target object from the generated relationship chain, judging whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model, updating the initial content of the associated object through the large language model when the judgment result is yes to obtain the updated content of the associated object, and updating the associated object to the target object until the target object is the second target text.

[0088] After the associated object is updated, use it as the new target object, then determine its associated object from the generated relationship chain again, and perform conflict judgment again. If a conflict is judged to exist, update the initial content of the associated object again to obtain the updated content of the associated object, and finally use this associated object as the new target object again. The specific implementation of this process is the same as that in the above embodiment and will not be elaborated here. Loop through the above steps until the target object obtained from the last update is a certain text, and define this text as the second target text, and the loop ends.

[0089] After completion, the initial content of the second target text is updated to the updated content, which does not conflict with the updated content of the previous target object (a certain N-level outline), and based on the previous conflict detection and update results, it also does not conflict with the updated content of the first target text. The novel finally generated through the above steps can ensure consistency and coherence.

[0090] When the target object is a certain text, it is the lowest-level object, and its associated object can only be a certain N-level outline; when the target object is the first-level outline, since the first-level outline is the highest-level object, its associated object can only be a certain or some second-level outlines; when the target object is other-level objects except the first level and a certain text, its associated objects will include a certain upper-level object and a certain or some lower-level objects. Since a target object may have multiple associated objects, for those associated objects that do not conflict, no update is performed, and for one or several conflicting associated objects, each needs to be updated. After the update, each associated object is used as a new target object.

[0091] In one embodiment, each step of using the large language model to determine whether there is a conflict between the initial content of the associated object and the updated content of the target object includes:

[0092] S61: Determine whether a viewing request for the associated object is currently received.

[0093] S62: If so, use the large language model to determine whether there is a conflict between the initial content of the associated object and the updated content of the target object.

[0094] If the initial content of the associated object has been generated, even if there is a conflict between the initial content of the associated object and the updated content of the target object, it will not be automatically updated without the user's knowledge, otherwise it will cause chaos. Therefore, after determining the associated object, it will first be judged whether a viewing request for the associated object is currently received. When the user wants to view the associated object, a viewing operation will be performed on the relevant page, and this operation generates a corresponding viewing request. After receiving the viewing request, the conflict judgment is then carried out through the large language model.

[0095] In one embodiment, each step of using the large language model to update the initial content of the associated object to obtain the updated content of the associated object includes:

[0096] S63: Determine whether an update request for the associated object is currently received.

[0097] S64: If so, use the large language model to update the initial content of the associated object to obtain the updated content of the associated object.

[0098] If it is determined that there is a conflict, it is also necessary to determine whether an update request for the associated object has been received. When the user issues an update request for the associated object, specifically, a pop-up window showing whether an update is required can be displayed to the user on the current page. If the user confirms the need for an update, a confirmation operation will be performed on the options to be updated in the pop-up window. Performing this operation is regarded as issuing an update request for the associated object. The generating device responds to this request and then performs an update operation, and finally obtains the updated content of the associated object.

[0099] In the above manner, the update operation is based on the user's permission, selection, and confirmation. Then, the user can know which outlines or texts are in conflict during the viewing process and make corresponding updates as needed.

[0100] In one embodiment, after the step of determining the associated object of the target object from the generated relationship chain, it further includes:

[0101] S6a: Determine whether the initial content of the associated object has been generated.

[0102] S6b: If not, generate the initial content of the associated object based on the updated content of the target object through a large language model.

[0103] The novel generation method provided by the embodiments of the present application is applicable to the scenario where the initial content of each object in the generated relationship chain is completely generated, and is also applicable to the scenario where the initial content of some objects has not been generated. For the former, since the initial content of the associated object has been generated, direct conflict detection and judgment can be performed each time. For the latter, for example, only the chapter outlines of Chapters 1 to 3 have been generated, and the chapter outlines of subsequent chapters have not been generated. When the associated object is the chapter outlines of Chapters 4 to 10, its initial content has not been generated, and conflict detection and judgment are not required. Another example is that only the texts of Chapters 1 to 3 have been generated, and the texts of subsequent chapters have not been generated. When the associated object is the texts of Chapters 4 to 10, its initial content has not been generated, and conflict detection and judgment are not required either. Therefore, after determining the associated object, it is also necessary to first determine whether the initial content of the associated object has been generated. If the initial content of the associated object has not been generated, directly generate the initial content of the associated object based on the updated content of the target object, and the subsequent lower-level outlines or texts of the associated object will be based on this initial content.

[0104] For the above two scenarios, when the novel generation method provided by the embodiments of the present application is not adopted, no matter what modifications are made to the previous text, the subsequent outlines and texts will only respond to the initially set content, and the above modifications by the user actually have no impact on the generation of the subsequent chapter texts, resulting in a break in the plot. After adopting the above method, whether the newly generated chapter texts or the updated chapter texts subsequently will be consistent with the modified previous text content, so the coherence of the plot can be ensured.

[0105] The following will specifically describe the above embodiments in conjunction with Figure 2 the above.

[0106] In Figure 2 , the initial contents of the 1st to 10th part outlines are respectively represented as lp1 to lp10, the updated contents are respectively represented as lp1' to lp10', the initial contents of the 1st to 100 chapter outlines are respectively represented as sp1 to sp100, the updated contents are respectively represented as sp1' to sp10', and the initial contents of the texts of the 1st to 100 chapters are respectively represented as zn1 to zn100, and the updated contents are respectively represented as zn1' to zn10'.

[0107] Assume that the text of the 2nd chapter is the first target text. After modifying it and generating a summary content, determine the 2nd chapter outline of the 2nd chapter text as the target object. Update its initial content sp2 based on the summary content to obtain the updated content sp2'.

[0108] Based on the generated relationship chain, determine that the associated object of the 2nd chapter outline is the 1st part outline, and use a large language model to determine whether there is a conflict between the initial content lp1 of the associated object and the updated content sp2' of the target object.

[0109] If it is determined that lp1 and sp2' conflict, then update lp1 to obtain the updated content lp1' of the 1st part outline, and lp1' does not conflict with sp1'.

[0110] After the update, use the 1st part outline as the new target object. Its associated objects include the 1st to 10th chapter outlines. If only sp1 of the 1st chapter outline and sp6 of the 6th chapter outline conflict with lp1', then only update sp1 to sp1' and sp6 to sp6', and do not update the other 8 chapter outlines. After the update, the 1st chapter outline and the 6th chapter outline are respectively used as new target objects, and their associated objects are respectively the text of the 1st chapter and the text of the 6th chapter. If zn1 of the text of the 1st chapter conflicts with sp1' and zn6 of the text of the 6th chapter also conflicts with sp6', then the text of the 1st chapter and the text of the 6th chapter are both second target texts, and it is necessary to update zn1 to zn1' and zn6 to zn6'. After the update, the text of the 1st chapter and the text of the 6th chapter do not conflict with the modified text of the 2nd chapter, thus achieving consistency and coherence.

[0111] It should be noted that although the modification of the main text of Chapter 1 in the above embodiments will only be synchronized to Chapters 2 to 10 based on the generation relationship chain, in actual scenarios, the degree of modification to the main text of a certain chapter usually only affects several chapters before and after it, and will not be so large as to affect all 100 chapters. If the degree of modification is large enough to affect 100 chapters, generally, the basic settings or the first-level outlines will be directly modified, rather than the main text of a certain chapter. Therefore, although the modification of the main text of Chapter 1 above will not be synchronized to all 100 chapters of the main text, the consistency and coherence of the full text can still be ensured.

[0112] As can be seen from the above embodiments, for the novel generation method provided in this application, after the first target main text is modified, by constructing a reverse prompt word and inputting it into the large language model, the updated content of the modified first target main text can be summarized, and the N-level outline for generating the first target main text can be updated with this summarized content. Then, the updated content of this N-level outline can reflect the overall modification situation of the first target main text. And through the generation relationship chain, the associated objects of this N-level outline can be determined level by level, and conflict judgment and update are performed on each associated object until the text level is traced, so as to synchronize the overall modification situation of the first target main text to the second target main text, making all main texts not conflict with the updated content of the first target main text. Therefore, the logical consistency and plot coherence of the entire novel main text are ensured, and dynamic memory of the modification is realized.

[0113] It should be noted that in the above embodiments, the situation of modifying the main text and finally synchronizing it to other main texts is described, but this application is not limited to this. If the modification is not for the main text of a certain chapter, but for a certain outline, the principle is similar, and the upper-level and / or lower-level associated objects can also be found level by level according to the generation relationship chain, and conflict judgment and update are performed until the text level is traced, and consistency and coherence can also be achieved.

[0114] Based on the method described in the above embodiments, this embodiment will further describe from the perspective of a novel generation device based on reverse prompt words. The novel includes N-level outlines, each upper-level outline generates at least two lower-level outlines, and each N-level outline generates the corresponding main text, where N is an integer not less than 1. Please refer to Figure 8 , the novel generation device may include:

[0115] A response module 10, configured to respond to a modification operation on the first target main text, update the initial content of the first target main text, and obtain the updated content of the first target main text;

[0116] A first obtaining module 20, configured to construct a reverse prompt word according to the updated content of the first target main text, process the reverse prompt word through a large language model, and obtain a summary content of the updated content of the first target main text;

[0117] A second obtaining module 30, configured to obtain a generation relationship chain formed by outlines at all levels and each main text, two objects with a generation relationship in the generation relationship chain are associated with each other, determine a target object that generates the first target main text from the generation relationship chain, and update the initial content of the target object based on the summary content to obtain the updated content of the target object;

[0118] A conflict judgment module 40, configured to determine an associated object of the target object from the generation relationship chain, and judge whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model;

[0119] A third obtaining module 50, configured to, if so, update the initial content of the associated object through the large language model to obtain the updated content of the associated object, and the updated content of the associated object does not conflict with the updated content of the target object;

[0120] A loop module 60, configured to update the associated object to the target object, and loop to execute the operation of determining the associated object of the target object from the generation relationship chain, the operation of judging whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model, the operation of updating the initial content of the associated object through the large language model to obtain the updated content of the associated object when the judgment result is yes, and the operation of updating the associated object to the target object until the target object is the second target main text.

[0121] In one embodiment, the second obtaining module 30 includes:

[0122] A first obtaining unit, configured to overwrite the initial content of the target object with the summary content to obtain the updated content of the target object.

[0123] In one embodiment, the second obtaining module 30 includes:

[0124] A second obtaining unit, configured to update the summary content in response to a modification operation on the summary content to obtain an updated summary content;

[0125] A third obtaining unit, configured to overwrite the initial content of the target object with the updated summary content to obtain the updated content of the target object.

[0126] In one embodiment, the second obtaining module 30 further includes a comparison display unit that works before the second obtaining unit, and the comparison display unit is configured to compare and display the initial content of the target object and the summary content on the same page.

[0127] In one embodiment, the loop module 60 includes a first judgment unit and a second judgment unit;

[0128] The first judgment unit is configured to judge whether a viewing request for the associated object is currently received;

[0129] The second judgment unit is configured to, if so, judge whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model;

[0130] In one embodiment, the loop module 60 includes a third judgment unit and a fourth obtaining unit;

[0131] The third judgment unit is configured to judge whether an update request for the associated object is currently received;

[0132] The fourth obtaining unit is configured to, if so, update the initial content of the associated object through the large language model to obtain the updated content of the associated object.

[0133] In one embodiment, the novel generation device further includes a generation judgment module and a generation module; after determining the associated object of the target object from the generation relationship chain, the generation judgment module is configured to judge whether the initial content of the associated object has been generated; the generation module is configured to, if not, generate the initial content of the associated object based on the updated content of the target object through the large language model.

[0134] Different from the prior art, the novel generation device based on reverse prompt words provided by the present application, after the first target text is modified, by constructing a reverse prompt word and inputting it into the large language model, can summarize the updated content of the modified first target text, and update the N-level outline for generating the first target text with the summarized content, then the updated content of the N-level outline can reflect the overall modification situation of the first target text, and through the generation relationship chain, the associated object of the N-level outline can be determined level by level, and conflict judgment and update are performed on each associated object until the text level is traced, so as to synchronize the overall modification situation of the first target text to the second target text, so that all texts do not conflict with the updated content of the first target text, thus ensuring the logical consistency and plot coherence of the entire novel text and realizing the dynamic memory of the modification.

[0135] Correspondingly, the embodiment of the present application further provides an electronic device, such as Figure 9As shown in the figure, the electronic device may include a radio frequency (RF) circuit 101, a memory 102 including one or more computer-readable storage media, an input unit 103, a display unit 104, a sensor 105, an audio circuit 106, a WiFi module 107, a processor 108 including one or more processing cores, and a power supply 109 and other components. Those skilled in the art can understand that Figure 9 the structure of the electronic device shown in

[0136] does not limit the electronic device, and may include more or fewer components than shown in the figure, or combine some components, or have different component arrangements. Among them:

[0137] The RF circuit 101 can be used for receiving and sending signals during information reception or call processes. Specifically, after receiving the downlink information from the base station, it is handed over to one or more processors 108 for processing; in addition, the data related to the uplink is sent to the base station. The memory 102 can be used to store software programs and modules, and the processor 108 executes various functional applications and novel generation by running the software programs and modules stored in the memory 102. The input unit 103 can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to customer settings and function controls.

[0138] The electronic device may further include at least one sensor 105, such as a light sensor, a motion sensor, and other sensors. The audio circuit 106 includes a speaker, and the speaker can provide an audio interface between the customer and the electronic device.

[0139] WiFi belongs to wireless transmission technology. The electronic device can help customers send and receive emails, browse the web, and follow up on streaming media through the WiFi module 107, which provides customers with wireless broadband Internet access. Although Figure 9 the WiFi module 107 is shown, it can be understood that it does not belong to an essential component of the electronic device and can be omitted completely according to needs without changing the essence of the application.

[0140] The processor 108 is the control center of the electronic device, connecting various parts of the entire mobile phone through various interfaces and lines, executing various functions of the electronic device and processing data by running or executing the software programs and / or modules stored in the memory 102, and calling the data stored in the memory 102, thereby monitoring the mobile phone as a whole.

[0141] The electronic device further includes a power source 109 (such as a battery) for supplying power to each component. Preferably, the power source can be logically connected to the processor 108 through a power management system, so as to realize functions such as charging management, discharging management, and power consumption management through the power management system.

[0142] Although not shown, the electronic device may further include a camera, a Bluetooth module, etc., which will not be elaborated here. Specifically, in this embodiment, the processor 108 in the server will load the executable files corresponding to the processes of one or more application programs into the memory 102 according to the following instructions, and the processor 108 will run the application programs stored in the memory 102 to realize the following functions:

[0143] In response to a modification operation on the first target text, update the initial content of the first target text to obtain the updated content of the first target text;

[0144] Construct a reverse prompt word according to the updated content of the first target text, and process the reverse prompt word through a large language model to obtain a summary content of the updated content of the first target text;

[0145] Obtain a generation relationship chain formed by various outlines and texts. Two objects with a generation relationship in the generation relationship chain are mutually associated. Determine the target object that generates the first target text from the generation relationship chain, and update the initial content of the target object based on the summary content to obtain the updated content of the target object;

[0146] Determine the associated object of the target object from the generation relationship chain, and judge whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model;

[0147] If so, update the initial content of the associated object through the large language model to obtain the updated content of the associated object, and the updated content of the associated object does not conflict with the updated content of the target object;

[0148] Update the associated object to the target object, and loop to execute the operation of determining the associated object of the target object from the generation relationship chain, the operation of judging whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model, the operation of updating the initial content of the associated object through the large language model when the judgment result is yes to obtain the updated content of the associated object, and the operation of updating the associated object to the target object until the target object is the second target text.

[0149] For the electronic device provided in this application, after the first target text is modified, by constructing a reverse prompt and inputting it into the large language model, the updated content of the modified first target text can be summarized. Using this summarized content to update the Nth-level outline that generates the first target text, the updated content of this Nth-level outline can reflect the overall modification of the first target text. By generating a relationship chain, the associated objects of this Nth-level outline can be determined level by level, and a conflict judgment and update are performed on each associated object until the text level is traced, so as to synchronize the overall modification of the first target text to the second target text, ensuring that all texts do not conflict with the updated content of the first target text, thus guaranteeing the logical consistency and plot coherence of the entire novel text and realizing dynamic memory of the modification.

[0150] In the above embodiments, the descriptions of the embodiments each have their own focuses. For parts not detailed in a certain embodiment, reference can be made to the detailed description above, and details will not be repeated here.

[0151] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by instructions, or by controlling relevant hardware through instructions. These instructions can be stored in a computer-readable storage medium and loaded and executed by a processor.

[0152] Therefore, an embodiment of this application provides a computer-readable storage medium, in which multiple instructions are stored, and these instructions can be loaded by a processor to implement the following functions:

[0153] In response to a modification operation on the first target text, update the initial content of the first target text to obtain the updated content of the first target text;

[0154] Construct a reverse prompt according to the updated content of the first target text, and process the reverse prompt through the large language model to obtain a summary content of the updated content of the first target text;

[0155] Obtain the generation relationship chain formed by each level of outline and each text. Two objects with a generation relationship in the generation relationship chain are associated with each other. Determine the target object that generates the first target text from the generation relationship chain, and update the initial content of the target object based on the summary content to obtain the updated content of the target object;

[0156] Determine the associated object of the target object from the generation relationship chain, and judge whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model;

[0157] If so, update the initial content of the associated object through the large language model to obtain the updated content of the associated object, and the updated content of the associated object does not conflict with the updated content of the target object;

[0158] Update the associated object to the target object, and loop through the operations of determining the associated object of the target object from the generated relationship chain, determining whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model, updating the initial content of the associated object through the large language model to obtain the updated content of the associated object when the judgment result is yes, and updating the associated object to the target object until the target object is the second target text.

[0159] The computer-readable storage medium provided by this application, after the first target text is modified, can summarize the updated content of the modified first target text by constructing a reverse prompt and inputting it into the large language model, and update the Nth-level outline for generating the first target text with the summarized content. Then, the updated content of the Nth-level outline can reflect the overall modification of the first target text. By generating a relationship chain, the associated object of the Nth-level outline can be determined level by level, and conflict judgment and update are performed on each associated object until the text level is traced, so as to synchronize the overall modification of the first target text to the second target text, making all texts not conflict with the updated content of the first target text. Therefore, the logical consistency and plot coherence of the entire novel text are ensured, and dynamic memory of the modification is realized.

[0160] The above has introduced in detail a novel generation method, device, electronic device, and computer-readable storage medium based on reverse prompt words provided by this application. Specific examples are used in this article to elaborate on the principle and implementation method of this application. The description of the above embodiments is only used to help understand the technical solution and its core idea of this application; those of ordinary skill in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A novel generation method based on reverse prompts, characterized in that, The novel includes N-level outlines, each upper-level outline generates at least two lower-level outlines, and each N-level outline generates corresponding text. N is an integer not less than 1. The method includes: In response to a modification operation on the first target text, update the initial content of the first target text to obtain the updated content of the first target text; Construct a reverse prompt based on the updated content of the first target text, and process the reverse prompt through a large language model to obtain a summary content of the updated content of the first target text; Obtain the generation relationship chain formed by each level of outline and each text. Two objects with a generation relationship in the generation relationship chain are associated with each other. Determine the target object that generates the first target text from the generation relationship chain, and update the initial content of the target object based on the summary content to obtain the updated content of the target object; Determine the associated object of the target object from the generation relationship chain, and judge whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model; If so, update the initial content of the associated object through the large language model to obtain the updated content of the associated object, and the updated content of the associated object does not conflict with the updated content of the target object; Update the associated object to the target object, and loop through the operations of determining the associated object of the target object from the generation relationship chain, judging whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model, updating the initial content of the associated object through the large language model when the judgment result is yes to obtain the updated content of the associated object, and updating the associated object to the target object until the target object is the second target text.

2. The novel generation method according to claim 1, wherein The step of updating the initial content of the target object based on the summary content to obtain the updated content of the target object includes: Cover the initial content of the target object with the summary content to obtain the updated content of the target object.

3. The novel generation method according to claim 1, characterized in that, The step of updating the initial content of the target object based on the summary content to obtain the updated content of the target object includes: In response to a modification operation on the summary content, update the summary content to obtain the updated summary content; Cover the initial content of the target object with the updated summary content to obtain the updated content of the target object.

4. The novel generation method according to claim 3, wherein Before the step of updating the summary content in response to a modification operation on the summary content, it further includes: Compare and display the initial content of the target object and the summary content on the same page.

5. The novel generation method according to claim 1, wherein Each step of judging whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model includes: Judge whether a viewing request for the associated object is currently received; If so, judge whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model.

6. The novel generation method according to claim 5, characterized in that, Each step of updating the initial content of the associated object through the large language model to obtain the updated content of the associated object includes: Determine whether an update request for the associated object is currently received; If so, update the initial content of the associated object through the large language model to obtain the updated content of the associated object.

7. The novel generation method according to claim 1, wherein After the step of determining the associated object of the target object from the generation relationship chain, it further includes: Determine whether the initial content of the associated object has been generated; If not, generate the initial content of the associated object through the large language model based on the updated content of the target object.

8. A novel generation device based on reverse prompts, characterized in that, The novel includes N-level outlines, each upper-level outline generates at least two lower-level outlines, and each N-level outline generates corresponding text. N is an integer not less than 1. The device includes: A response module for updating the initial content of the first target text in response to a modification operation on the first target text to obtain the updated content of the first target text; A first obtaining module for constructing a reverse prompt word according to the updated content of the first target text, and processing the reverse prompt word through the large language model to obtain a summary content of the updated content of the first target text; A second obtaining module for obtaining a generation relationship chain formed by each level of outline and each text, two objects with a generation relationship in the generation relationship chain are associated with each other, determining a target object that generates the first target text from the generation relationship chain, and updating the initial content of the target object based on the summary content to obtain the updated content of the target object; A conflict judgment module for determining an associated object of the target object from the generation relationship chain, and judging whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model; A third obtaining module for, if so, updating the initial content of the associated object through the large language model to obtain the updated content of the associated object, and the updated content of the associated object does not conflict with the updated content of the target object; A loop module for updating the associated object to the target object, and looping to execute the operation of determining the associated object of the target object from the generation relationship chain, the operation of judging whether there is a conflict between the initial content of the associated object and the updated content of the target object through the large language model, the operation of updating the initial content of the associated object through the large language model to obtain the updated content of the associated object when the judgment result is yes, and the operation of updating the associated object to the target object until the target object is the second target text.

9. An electronic device, characterized in that, It includes a memory and a processor; the memory stores an application program, and the processor is used to run the application program in the memory to execute the steps in the novel generation method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, A computer program is stored on the computer-readable storage medium, and the computer program is executed by the processor to implement the steps in the novel generation method according to any one of claims 1 to 7.

Images