Information generation method and apparatus, electronic device, computer-readable storage medium

Abstract

The present disclosure provides an information generation method and device, relates to the technical field of artificial intelligence, in particular to the technical field of natural language processing, voice processing, deep learning, large model and the like, and can be applied to the content generation of artificial intelligence and the like. The specific implementation scheme is as follows: generating an initial text in an initial stage according to preset story basic information; in response to judging that the initial text does not include a key word of a conclusion, generating and sending interactive content information based on the initial text; in response to receiving user feedback information, determining key information in the initial stage based on the user feedback information and the initial text; and generating a work text of a work based on the initial text, the key information and the story basic information, thereby improving user experience and enhancing the generation quality of the work.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, specifically to technical fields such as natural language processing, speech processing, deep learning, and large models, and can be applied to scenarios such as artificial intelligence content generation, and in particular to an information generation method and device, electronic equipment, computer-readable storage medium, and computer program product. Background Art

[0002] In the field of literary creation, interactive novels became popular before large-scale models entered the public eye. However, the multiple story lines, interactive nodes and options of these early interactive novels were all set in advance by the producers, and users could not truly participate in the creation of the novels. Summary of the Invention

[0003] The present disclosure provides an information generation method and apparatus, an electronic device, a computer-readable storage medium, and a computer program product.

[0004] According to a first aspect, an information generation method is provided, the method comprising: generating an initial body text of an initial stage based on preset basic story information; in response to determining that the initial body text does not include ending keywords, generating and sending interactive content information based on the initial body text; in response to receiving user feedback information, determining key information of the initial stage based on the user feedback information and the initial body text; and generating a work text of the work based on the initial body text, the key information and the basic story information.

[0005] According to a second aspect, an information generating device is provided, which includes: a main text generating unit, configured to generate an initial main text of an initial stage based on preset basic story information; an interactive generating unit, configured to generate and send interactive content information based on the initial main text in response to a judgment that the initial main text does not include ending keywords; a determining unit, configured to determine key information of the initial stage based on the user feedback information and the initial main text in response to receiving user feedback information; and a work generating unit, configured to generate a work text of a work based on the initial main text, the key information and the basic story information.

[0006] According to a third aspect, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method described in any implementation manner of the first aspect.

[0007] According to a fourth aspect, a non-transitory computer-readable storage medium storing computer instructions is provided, where the computer instructions are used to cause a computer to execute the method as described in any implementation of the first aspect.

[0008] According to a fifth aspect, a computer program product is provided, comprising a computer program, which implements the method described in any implementation manner of the first aspect when executed by a processor.

[0009] The information generation method and device provided by the embodiments of the present disclosure first generate an initial body text of the initial stage based on preset basic story information; secondly, in response to determining that the initial body text does not include ending keywords, interactive content information is generated and sent based on the initial body text; thirdly, in response to receiving user feedback information, key information of the initial stage is determined based on the user feedback information and the initial body text; finally, the work text of the work is generated based on the initial body text, the key information, and the basic story information. Thus, by generating interactive content information based on the initial body text, users can participate in the creation of the work and continue to create the work based on the user feedback information of the users, which not only improves the user's participation in the creation of the work, but also increases the diversity of the plot development of the work and improves the generation quality of the work.

[0010] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0012] Figure 1 is a flow chart of an embodiment of the information generation method according to the present disclosure;

[0013] Figure 2 is a flow chart of another embodiment of the information generating method according to the present disclosure;

[0014] Figure 3 is a structural diagram of an embodiment of an information generating device according to the present disclosure;

[0015] Figure 4 It is a block diagram of an electronic device used to implement the information generating method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0016] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0017] The works disclosed herein include works of literature, art, natural science, social science, engineering technology, etc. created in the following forms: 1. Written works; 2. Oral works; 3. Music, drama, folk art, and dance works; 4. Fine arts and photography works; 5. Film, television, and video works; In traditional technologies, large language models and their related applications have received widespread attention. Among them, using large models to create works (such as novel writing) is an emerging field that has attracted many researchers. In the industrial world, interactive works became popular before large models entered the public eye, but the multiple story lines, interactive nodes, and options of such early interactive works were all set in advance by the producers, and users did not actually participate in the creation of the works.

[0018] To allow users to truly participate in the creation of interactive works, current character dialogue technology solutions include: 1) Creators set up a multi-branch, multi-ending narrative story, similar to a tree structure, with readers' choices at each interactive node determining unique branches and endings. 2) Utilize a large model to generate the main text directly without generating an outline, and then interact with the user.

[0019] Regarding the first technical solution, currently available interactive works on the market are all stories with multiple branches and endings pre-set by the creators. Users can only choose one of the pre-set options at a pre-set interactive node to determine the direction of the story. In essence, users are still passively reading the plot created by the creator. Users do not participate in the creative process, and the number of storylines is limited, with limited content available for users to choose from. In many cases, users cannot determine the direction of the story according to their own preferences.

[0020] Regarding the second technical solution, since the current ability of large models to create long works is not particularly satisfactory, due to the limitations of the maximum input allowed by the model and the model's long-term memory issues, directly generating the main text in one step often results in common problems such as plot discontinuities and repetitions. As a result, the works generated by this solution are not very readable and their creative quality is far from that of human writers.

[0021] In order to improve the participation of users in interactive works and improve the quality of work creation, let the plot of the work develop according to the personal preferences of the user, and create a personalized novel plot, the present disclosure provides an information generation method, Figure 1 The flow 100 of one embodiment of the information generation method according to the present disclosure is shown, and the information generation method comprises the following steps:

[0022] Step 101, generating an initial text of an initial stage according to preset story basic information.

[0023] In this embodiment, the story basic information is information such as the background and character setup of the work. The work can be an interactive novel, an interactive drama, an interactive poem, or other literary works. The work can also be a non-literary work. The story basic information can be used to determine the framework of each stage of the work. The execution subject on which the information generation method runs can obtain the preset story basic information in various ways, for example, the execution subject can obtain the story basic information stored in the database server through wired or wireless connection. For another example, the execution subject can communicate with the user terminal to obtain the story basic information sent by the user terminal in real time.

[0024] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of the story basic information are performed after authorization and comply with relevant laws and regulations.

[0025] In this embodiment, the step 101 comprises generating an initial plot summary based on the preset story basic information, and generating an initial text of an initial stage based on the initial plot summary.

[0026] The initial plot summary is a summary of the plot of the initial stage of the work. The initial plot summary can provide a story framework for the initial text. The execution subject on which the information generation method of the present disclosure runs can divide the generated text into multiple stages when creating the work, and the initial stage is the beginning of the work.

[0027] In this embodiment, the step 101 comprises inputting the story basic information into a plot summary generation large model to obtain an initial plot summary output by the plot summary generation large model. The plot summary generation large model is a model for representing the correspondence between the key information, the story basic information and the plot summary. When generating the initial plot summary, the key information input into the plot summary generation large model can be set to empty.

[0028] In this embodiment, the initial text is a story text description of the work. The initial text can depict the story development content of the initial stage.

[0029] In this embodiment, based on the plot development trend of the work to be created, the story can be divided into multiple stages, such as the initial stage of story development, subsequent stages, and multiple consecutive stages. In each stage before the work is completed, a body text for the corresponding stage is generated, such as the initial body text and the body text for each stage. In addition to generating key information for the initial stage based on the initial body text, the body text for each subsequent stage can be generated based on the key information of each previous stage.

[0030] In this embodiment, the above step 102 includes: inputting the initial plot summary into the main text generation model to obtain the initial main text output by the main text generation model.

[0031] Step 102 : In response to determining that the initial body text does not include the ending keyword, interactive content information is generated and sent based on the initial body text.

[0032] In this embodiment, the interactive content information is a specific content display of interactive information for interaction with the user. The interactive content information can be expressed in the form of images, text, and voice. Through the interactive content information, various plot development directions of the initial stage of the story can be displayed to the user. The user determines the development of the story in the subsequent stage by selecting one of the development directions.

[0033] The above step 102 includes: extracting key information from the initial body text, inputting the extracted key information into the interactive large model, and obtaining interactive content information.

[0034] Step 103 : In response to receiving the user feedback information, determine the key information of the initial stage based on the user feedback information and the initial body text.

[0035] In this embodiment, user feedback information refers to the information obtained after the user operates or changes the interactive content information after obtaining it. Through the user feedback information, the development trend of the main text expected by the user can be determined, providing a reliable implementation method for users to participate in the generation of the work text.

[0036] In this embodiment, key information is the main information that affects the development of the plot, and the key information can be keywords, key scenes, key images, etc.

[0037] In this embodiment, the above step 103 includes: converting the user feedback information into feedback text, splicing the feedback text and the initial main text to obtain a spliced ​​text; extracting keywords from the spliced ​​text, and determining the extracted keywords as key information. After determining the key information in the initial stage, the key information can be stored in the knowledge base.

[0038] Step 104 : Generate the work text of the work based on the initial body text, key information and basic story information.

[0039] In this embodiment, the work text refers to text data that describes the entire content of the work. The work text may include: an initial body text or a body text of at least one stage.

[0040] In this embodiment, the main text of the work can be generated directly based on the key information and the basic information of the story. When the main text includes the ending keyword, the ending text of the work is obtained, and the initial main text is merged with the main text to obtain the work text of the work; optionally, the above step 104 can also include: based on the key information and the basic information of the story, generating the main text of multiple stages, while detecting whether each main text includes the ending keyword, until the main text of the last stage includes the ending keyword, for this reason, the main text of the last stage is the ending text of the work, and the work text of the work is generated at this time, wherein, after generating the main text of multiple stages, new interactive content information and new key information can be generated, and the text of the work is generated based on the feedback information of each stage and the key information of all stages.

[0041] In this embodiment, the ending text is a text indicating that the work is completed. Through the ending text, it can be determined that the work to be created has been completed, and the work text of the entire work is a finished product.

[0042] This disclosure proposes pre-setting the overall story background and framework, but the specific plot development and interaction timing are all created and selected by the large model. During interaction, users can not only choose from the options provided by the large model, but also propose and input their own ideas, and the large model will then continue to create based on the user's choices or ideas. This not only increases user participation in the interactive work, but also greatly enhances the diversity of the plot within the work, allowing for endless possibilities for story development.

[0043] The information generation method provided by the embodiment of the present disclosure first generates an initial body text of the initial stage based on the preset basic information of the story; secondly, in response to determining that the initial body text does not include the ending keyword, interactive content information is generated and sent based on the initial body text; thirdly, in response to receiving user feedback information, the key information of the initial stage is determined based on the user feedback information and the initial body text; finally, the work text of the work is generated based on the initial body text, the key information and the basic information of the story. Thus, generating interactive content information based on the initial body text can enable users to participate in the creation of the work and continue to create the work based on the user feedback information of the user, which not only improves the user's participation in the creation of the work, but also increases the diversity of the plot development of the work and improves the generation quality of the work.

[0044] In some optional implementations of the present disclosure, the above-mentioned initial main text and the generation of the work text of the work based on the key information and the basic information of the story include: generating a plot summary of the current stage based on the key information and the basic information of the story; generating the main text of the current stage based on the plot summary and key information of the current stage; judging whether the main text of the current stage includes the ending keyword; in response to judging that the main text of the current stage includes the ending keyword, generating the work text of the work based on the initial main text, the main text of the current stage and the main text of the previous stage.

[0045] In this optional implementation, the above-mentioned key information is the key information of the initial stage. Through the key information of the initial stage and the basic information of the story, a plot summary of the next stage adjacent to the initial stage can be generated, wherein the plot summary of the next stage adjacent to the initial stage is the current stage.

[0046] In this optional implementation, the above-mentioned generation of the plot summary of the current stage based on the key information and the basic story information includes: determining the story development direction related to the key information from the basic story information; and performing plot expansion on the story development direction to obtain the plot summary of the current stage.

[0047] In this optional implementation, the above-mentioned generation of the main text of the current stage based on the plot summary and key information of the current stage includes: sentence-patterning the plot summary of the current stage to obtain the processed text of the current stage; selecting text related to the key information of the initial stage from the current processed text to obtain the main text of the current stage.

[0048] In this optional implementation method, the main text of the current stage can be segmented to obtain the segmentation set of the current stage, and the similarity between the segmentations in the segmentation set of the current stage and the ending keywords is compared. If the similarity between the segmentations and the ending keywords is greater than the similarity threshold, it is determined that the main text of the current stage includes the ending keywords. When the main text of the current stage includes the ending keywords, the initial main text, the main text of the current stage, and the main text of the stage before the current stage are used as the work text of the work.

[0049] It should be noted that the body text of the previous stage is all stages between the initial stage and the current stage. When there is no stage between the initial stage and the current stage, the body text of the previous stage is empty.

[0050] The method for producing the ending text of a work provided by this optional implementation method first generates a plot summary of the current stage based on keyword information and basic story information; generates the main text of the current stage based on the plot summary of the current stage and key information of the initial stage; when it is determined that the main text of the current stage includes the ending keyword, generates the work text based on the initial main text, the main text of the current stage and the main text of the previous stage. By resolving the keywords, a reliable implementation method is provided for the termination of the main text of each stage.

[0051] Optionally, the above-mentioned generation of the work text of the work based on the initial main text, key information and basic story information includes: generating the main text of the current stage based on the key information and basic story information; judging whether the main text of the current stage includes the ending keyword; in response to judging that the main text of the current stage includes the ending keyword, generating the work text of the work based on the main text of the current stage, the initial main text and the main text of the previous stage.

[0052] In some optional implementations of the present disclosure, generating a plot summary for the current stage based on key information and basic story information includes: inputting the key information and basic story information into a plot summary generation model to obtain a plot summary output by the plot summary generation model; and preprocessing the plot summary output by the plot summary generation model to obtain a plot summary for the current stage.

[0053] In this optional implementation, the outline generation guide words, the key information in the initial stage, and the basic information of the story are input into the plot summary generation model, and a plot summary output by the plot summary generation model can be obtained, wherein the outline generation guide words are used to guide the plot summary generation model to generate a plot summary related to the key information in the initial stage and the basic information of the story from the perspective of outline generation.

[0054] In this optional implementation, preprocessing the plot summary output by the plot summary generation model includes: detecting whether the plot summary output by the plot summary generation model is consistent with the development of the story; if the plot summary output by the plot summary generation model is consistent with the development of the story, the plot summary output by the plot summary generation model is used as the plot summary of the current stage.

[0055] Optionally, in response to the large plot summary generation model outputting multiple plot summaries, a plot summary from the multiple plot summaries is selected as the plot summary of the current stage based on key information.

[0056] The method for generating a plot summary for the current stage provided by this optional implementation adopts a large plot summary generation model to generate a plot summary, and then preprocesses the plot summary output by the large plot summary generation model to obtain a plot summary for the current stage, thereby improving the accuracy of plot summary generation.

[0057] Optionally, generating the plot summary of the current stage based on the key information and the basic story information includes: inputting the key information and the basic story information into a plot summary generation model, and obtaining the plot summary of the current stage output by the plot summary generation model.

[0058] In some optional implementations of the present disclosure, the above-mentioned plot summary generation model is trained using the following steps: obtaining an initial basic corpus text including story background and character information; generating a basic outline of the initial basic corpus text through a pre-trained outline model; generating a training plot text based on the basic outline; combining the initial main text corpus text and the training plot text to obtain plot positive training data; and using the plot positive training data to train the initial model to obtain a trained plot summary generation model.

[0059] In this optional implementation, the initial basic corpus text is the key corpus text that can generate a complete storyline. The key corpus text includes the story background and character information, where the character information is the roles and names of the people in the story, and the story background is the time and scene where the story takes place.

[0060] In this optional implementation, the pre-trained outline model is an excellent model in the industry, which can generate an outline based on a text with a storyline.

[0061] In this optional implementation, the plot-forward training data refers to the initial corpus consisting of the novel's story background and character information. An outline is then generated using an industry-leading large model, and then the training plot text is generated based on the outline. The advantage of this plot-forward training data is that the generated plot summary is complete and logically coherent.

[0062] In this optional implementation, the training plot text is the text corresponding to the plot summary in the plot positive training data. The training plot text serves as the true value of the plot summary and provides a reliable training basis for the plot summary generation model.

[0063] In this optional implementation, the large plot summary generation model is invisible to the user, supports context windows up to 8KB, and considers approximately 8,000 words of textual information when generating summaries. This feature is crucial for maintaining plot coherence and contextual relevance, especially when working with long articles or serialized stories, ensuring that the summary does not lose key information while maintaining the integrity of the text. Furthermore, the model's small size allows for rapid response at a relatively low cost.

[0064] When applying the disclosed method to interactive novels, a small model is first used to generate a novel outline (content connected by plot summaries of multiple stages), and then a slightly larger model is used to generate the main text based on the novel outline. Finally, the RAG method for extracting key information can effectively improve the plot coherence of the interactive novel. And unlike traditional interactive novel products that have preset fixed plot branches, the plot branches of interactive novels are infinite, because every time at an interactive point, the user can choose or input their own ideas, and then the model continues to create according to the user's choice or input, making the plot development of the novel have infinite possibilities, and the content created is in line with the user's personal preferences. This makes the applied product more interactive, can enhance the user's interest in the product and stickiness to the product, increase the length of time the user uses the product, and thus effectively improve the product's market competitiveness.

[0065] This optional implementation provides a training method for a plot summary generation model. The method generates a basic outline for an initial basic corpus text using a pre-trained outline model. Based on the basic outline, training plot text is generated. The initial body corpus text and the training plot text are combined to obtain positive plot training data. The positive plot training data is used to train the initial model to obtain a trained plot summary generation model, thereby improving the reliability of the generation of the plot summary generation model.

[0066] In some optional implementations of the present disclosure, the above-mentioned generation of the main text of the current stage based on the plot summary and key information of the current stage includes: inputting the key information and the plot summary of the current stage into a main text generation model to obtain the main text output by the main text generation model; and preprocessing the main text output by the main text generation model to obtain the main text of the current stage.

[0067] In this optional implementation method, the main text generation guide words, the key information of the initial stage, and the plot summary of the current stage are input into the main text generation model, and the main text input by the main text generation model can be obtained. Among them, the main text generation guide words are used to guide the main text generation model to generate main text related to the key information and plot summary of the initial stage from the perspective of story main text generation.

[0068] In the optional implementation, the preprocessing of the text text output by the text generation large model includes: detecting whether the text text output by the text generation large model conforms to the story development logic, and if the text text output by the text generation large model conforms to the story development logic, taking the text text output by the text generation large model as the text text of the current stage.

[0069] The method for generating the text text of the current stage provided by the optional implementation adopts the text generation large model to generate the text text of the current stage, and then pre-processes the text text output by the text generation large model to obtain the text text of the current stage, thereby improving the accuracy of text text generation.

[0070] In some optional implementations of the present disclosure, the text generation large model is trained by the following steps: obtaining initial basic corpus text including story background and character information; generating a basic outline of the initial basic corpus text by a pre-trained outline large model; generating training text based on the basic outline; combining the initial text corpus text and the training text to obtain text forward training data; training the initial large model using the text forward training data to obtain the trained text generation large model.

[0071] In the optional implementation, the initial basic corpus text is a text with a complete story plot, and the story background and character information are included in the initial basic corpus text. The pre-trained outline large model is an excellent large model in the industry, which can generate an outline based on a text with a story plot.

[0072] In the optional implementation, the text forward training data refers to the story background and character information of the initial corpus as a novel, then an outline is generated by an excellent large model in the industry, and then a training text is generated according to the outline. The advantage of the final training data, the text forward training data, is that the generated story plot summary is complete, logical, and smooth.

[0073] In the optional implementation, the training text is the text corresponding to the text text of the current stage in the text forward training data, and the training text as the true value of the text text of the current stage provides a reliable training basis for the text generation large model.

[0074] In this optional implementation, the large text generation model excels at understanding and generating complex texts, and with its support for a 128KB context window, it is crucial for maintaining narrative coherence and contextual consistency in long texts, particularly when creating texts that require long-term memory and multi-layered narrative structures. The long-term memory capability of the large text generation model is one of its core strengths, enabling it to remember and reference information from earlier in the text, thereby generating content that maintains consistency in plot, style, and tone. This long-term memory enables the model to deliver high-quality output when creating long texts.

[0075] In addition, large text generation models can also support streaming output, which can process input data in parallel while generating text, rather than waiting for the entire document to be processed. This approach allows the model to maintain smooth output even when processing large amounts of data.

[0076] This optional implementation provides a training method for a large text generation model. A basic outline of an initial basic corpus text is generated through a pre-trained outline large model. Based on the basic outline, a training text is generated. The initial text corpus text and the training text are combined to obtain text positive training data. The text positive training data is used to train the initial large model to obtain a trained text generation model, thereby improving the reliability of the generation of the large text generation model.

[0077] In view of the above situation, the present disclosure provides another embodiment of the information generation method, see Figure 2 , shows a process 200 according to another embodiment of the information generation method of the present disclosure, the information generation method comprising the following steps:

[0078] Step 201 : Generate the initial body text of the initial stage according to the preset basic information of the story, and then execute step 202 .

[0079] Step 202 , in response to determining that the initial body text does not include the ending keyword, generates and sends interactive content information based on the initial body text, and then executes step 203 .

[0080] Step 203 , in response to receiving the user feedback information, determining the key information of the initial stage based on the user feedback information and the initial body text, and then executing step 204 .

[0081] It should be understood that the operations and features in the above steps 201-203 correspond to the operations and features in steps 101-103 respectively. Therefore, the above descriptions of the operations and features in steps 101-103 are also applicable to steps 201-203 and will not be repeated here.

[0082] Step 204 , based on the key information and basic story information, generates a plot summary for the current stage, and then executes step 205 .

[0083] Step 205 , based on the plot summary and key information of the current stage, generate the main text of the current stage, and then execute step 206 .

[0084] Step 206, determine whether the main text of the current stage includes the ending keyword; if the main text of the current stage includes the ending keyword, execute step 207; if the main text of the current stage does not include the ending keyword, execute step 209.

[0085] Step 207 , based on the initial body text, the body text of the current stage and the body text of the previous stage, generates the work text of the work, and then executes step 208 .

[0086] Step 208, end.

[0087] Step 209: Generate and send interactive content information of the current stage based on the main text of the current stage.

[0088] Step 210 : In response to receiving feedback information on the interactive content information of the current stage, determine key information of the current stage based on the feedback information.

[0089] Step 211 , based on the key information of the current stage, the key information before the current stage and the basic story information, generate the main text of the next stage, use the main text of the next stage as the main text of the current stage, and continue to execute step 206 .

[0090] The above step 211 includes: generating a plot summary of the next stage based on the key information of the current stage, the key information before the current stage and the basic story information; generating the main text of the next stage based on the plot summary of the next stage and the above key information of the current stage and the key information before the current stage.

[0091] Optionally, the key information of the current stage and the key information before the current stage are stored in the knowledge base. The information in the knowledge base can be retrieved through RAG (Retrieval-Augmented Generation) to obtain retrieval information. Based on the retrieval information and the basic information of the story, the main text of the next stage is generated, and the main text of the next stage is used as the main text of the current stage, and step 206 is continued.

[0092] In this embodiment, key information from each stage can be stored in a knowledge base. The knowledge base acts as an information warehouse, not only storing the extracted knowledge points but also organizing them for future retrieval and application. This storage is structured, so each knowledge point is given a certain context and relevance, allowing for more accurate matching to relevant queries in subsequent use.

[0093] The next time a plot is generated or content is created, the RAG method can be used to retrieve all key historical plot points from the knowledge base based on the previous round of plot points. This is crucial for maintaining the coherence and richness of the story, ensuring that the newly generated content is consistent with the previous content and that previous plot elements and facts can be introduced and reused when necessary. Through this mechanism, the created works are not only more intelligent but also closer to the creative methods of human authors. The accumulation and application of knowledge can continuously improve the depth and quality of the content.

[0094] In this embodiment, after the work text of the work is generated, a completed work can be obtained. The work can be an interactive novel. Specifically, the implementation of the interactive novel is divided into three steps. The first step is to create / continue the plot summary of each stage based on the preset story background, character setting, user decision-making and text. The second step is to generate the text of each stage based on the plot summary of each stage. The third step is to extract key information from the previously generated text based on the knowledge base and RAG method after generating the text, and store the extracted information in the knowledge base. The key information will be retrieved from the knowledge base when the plot is generated next time. The implementation of these three steps will use different large models that have been fine-tuned and trained.

[0095] The above method generates interactive novels in segments. First, a plot summary is generated, then the main text is generated based on the plot summary. Finally, key information is extracted from the main text, stored in the knowledge base, and used in the next generation of the plot summary and main text. Because the plot summary is relatively short, the model does not forget the previous content due to excessive length, which makes it easy to ensure plot coherence. Therefore, the main text generated from the plot summary will not have plot contradictions or incoherence. Using the RAG method to extract key information can further ensure plot coherence.

[0096] The information generation method provided by this embodiment generates and sends the interactive content information of the current stage based on the main text and key information of the current stage when it is determined that the main text of the current stage does not include the ending keyword; in response to receiving feedback information on the interactive content information of the current stage, based on the feedback information, the key information of the current stage is determined, and the main text of the next stage is generated based on the key information of the current stage, the key information before the current stage and the basic information of the story, and the main text of the next stage is used as the main text of the current stage, and it is continued to be determined whether the main text of the current stage includes the ending keyword. For this reason, interactive content information can be generated at each stage of the main text generation, and user feedback information can be received, thereby improving the user's work creation experience.

[0097] In some embodiments of the present disclosure, the above-mentioned response to receiving user feedback information, based on the user feedback information and the initial main text, determining the key information of the initial stage includes: based on the initial main text, selecting multiple initial information from a preset knowledge base; based on the user feedback information, determining the plot development direction word; based on the direction indicated by the plot development direction word, selecting the key information of the initial stage from multiple initial information.

[0098] In this embodiment, a variety of keywords are stored in the knowledge base, and the initial main text contains multiple words, and the initial information is keywords related to the initial main text and different plot development directions. For example, the keyword for the tragic direction is: "leave"; the keyword for the perfect direction is: "together". Through user feedback information, the development direction of the user's story can be determined.

[0099] In this embodiment, based on the plot development direction word, keywords in the initial information that are in the same direction as indicated by the plot development direction word are determined to obtain key information.

[0100] The method for determining key information in the initial stage provided by this embodiment first selects initial information related to the initial main text from a knowledge base based on the initial main text, and then determines the plot development direction word based on user feedback information; based on the direction indicated by the plot development direction word, the key information of the initial stage is selected from the initial information, thereby improving the reliability of obtaining the key information in the initial stage.

[0101] In some optional implementations of the present disclosure, the above-mentioned selection of multiple initial information from a preset knowledge base based on the initial main text includes: inputting the initial main text into a key information extraction model to obtain the main text key information output by the key information extraction model; and selecting multiple initial information related to the main text key information from the preset knowledge base.

[0102] In this optional implementation method, key information extraction guide words and initial body text are input into the key information extraction big model, and the body key information output by the key information extraction big model can be obtained, wherein the key information extraction guide words are used to guide the key information extraction big model to generate body key information related to the initial body text from the perspective of key information extraction.

[0103] In this optional implementation, the key information extraction model will conduct in-depth analysis and information distillation on the generated initial body text. This process is not just a simple selection of keywords or phrases, but a more complex understanding of the semantic content of the text, identifying those information elements that are crucial to the overall narrative framework and theme, and obtaining the key information of the body text.

[0104] In this embodiment, after the key information of the text is obtained through the key information extraction model, the key information of the text can be stored in a preset knowledge base, so that the corresponding key information of the stage can be provided for each subsequent stage.

[0105] The method for selecting multiple initial information provided by this optional implementation method first obtains the key information of the initial body text through a key information extraction large model; and selects multiple initial information related to the key information of the body text from the knowledge base, thereby improving the accuracy of obtaining multiple initial information.

[0106] In some optional implementations of the present disclosure, the above-mentioned key information extraction big model is trained by the following steps: obtaining the initial main text corpus text including the main text of the story; generating the main text outline of the initial main text corpus text through the pre-trained outline big model; generating the training background and character text based on the main text outline; combining the initial main text corpus text with the training background and character text to obtain text reverse training data; using the text reverse training data to train the initial big model to obtain a trained key information extraction big model.

[0107] In this optional implementation, the initial body text corpus is a body text with a complete storyline, which may include multiple stages of the story. The pre-trained outline model is an industry-leading model that can generate an outline based on the initial body text corpus with a complete storyline.

[0108] In this optional implementation, the text reverse training data refers to the initial corpus being the main text of the novel. Then, the industry's existing high-performance large model is used to summarize the outline, story background and character information based on the main text, and the key text of the main text is obtained based on the outline, story background and character information. The advantage of text reverse training data is that the main text content is complete and the plot is coherent.

[0109] In this optional implementation, the main text key text is the text corresponding to the key information in the text reverse training data. The main text key text serves as the true value of the key information and provides a reliable training basis for the key information extraction model.

[0110] This optional implementation provides a training method for a key information extraction large model. The method generates a text outline of the initial text corpus text through a pre-trained outline large model, and generates a training background and character text based on the text outline; based on the text outline, the training background and character text, the text key text is obtained, and the initial text corpus text and the text key text are combined to obtain text reverse training data; the text reverse training data is used to train the initial large model to obtain a trained key information extraction model, thereby improving the reliability of the generation of the key information extraction model.

[0111] In some optional implementations of the present disclosure, in response to determining that the initial body text does not include the ending keyword, generating and sending interactive content information based on the initial body text includes: matching the initial body text with a preset ending keyword; in response to the initial body text and the ending keyword not being successfully matched, determining that the initial body text does not include the ending keyword, inputting the initial body text into a pre-trained interactive large model, and obtaining interactive content information output by the interactive large model.

[0112] In this optional implementation, the ending keyword is a keyword in the story that is related to the ending.

[0113] In this optional implementation, the interactive content information is information for interacting with the user, and the information is used to display information of different plot development directions to the user. The interactive big model is a big model for generating interactive content information. The interactive big model obtains interactive content information including multiple plot development directions by performing multi-development plot analysis on the initial main text. After displaying the interactive content information to the user, the user can determine one of the plot development directions from the interactive content information.

[0114] The method for generating interactive content information provided by this optional implementation method inputs the initial body text into a pre-trained interactive large model when the initial body text and the ending keyword are not successfully matched, and obtains the interactive content information output by the interactive large model, thereby improving the reliability of obtaining the interactive content information.

[0115] Further references Figure 3 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of an information generating device, which is similar to Figure 1 Corresponding to the method embodiment shown, the device can be specifically applied to various electronic devices.

[0116] like Figure 3As shown, the information generation device 300 provided in this embodiment includes: a main text generation unit 301, an interactive generation unit 302, a determination unit 303, and a work generation unit 304. The main text generation unit 301 can be configured to generate an initial main text of the initial stage based on preset basic story information. The interactive generation unit 302 can be configured to generate and send interactive content information based on the initial main text in response to determining that the initial main text does not include ending keywords. The determination unit 303 can be configured to determine the key information of the initial stage based on the user feedback information and the initial main text in response to receiving user feedback information. The work generation unit 304 can be configured to generate the work text of the work based on the initial main text, key information and basic story information.

[0117] In this embodiment, the specific processing of the text generation unit 301, the interaction generation unit 302, the determination unit 303, and the work generation unit 304 and the technical effects thereof can be referred to in the respective Figure 1 The relevant descriptions of step 101, step 102, step 103, and step 104 in the corresponding embodiment are not repeated here.

[0118] In some optional implementations of this embodiment, the above-mentioned work generation unit 305 is further configured to: generate a plot summary of the current stage based on key information and basic story information; generate the main text of the current stage based on the plot summary and key information of the current stage; determine whether the main text of the current stage includes ending keywords; in response to determining that the main text of the current stage includes ending keywords, generate the work text of the work based on the initial main text, the main text of the current stage and the main text of the previous stage.

[0119] In some optional implementations of this embodiment, the above-mentioned work generation unit 305 is further configured to: input key information and basic story information into the plot summary generation model to obtain the plot summary output by the plot summary generation model; pre-process the plot summary output by the plot summary generation model to obtain the plot summary of the current stage.

[0120] In some optional implementations of this embodiment, the above-mentioned plot summary generation big model is trained by a plot training unit (not shown in the figure), and the plot training unit is configured to: use plot positive training data to train the initial big model to obtain a trained plot summary generation big model, wherein the plot positive training data is obtained by the plot training unit using the following steps: obtaining an initial basic corpus text including story background and character information; generating a basic outline of the initial basic corpus text through a pre-trained outline big model; generating a training plot text based on the basic outline; and combining the initial main text corpus text and the training plot text to obtain plot positive training data.

[0121] In some optional implementations of this embodiment, the above-mentioned work generation unit 305 is further configured to: input key information and the plot summary of the current stage into the main text generation model to obtain the main text output by the main text generation model; pre-process the main text output by the main text generation model to obtain the main text of the current stage.

[0122] In some optional implementations of this embodiment, the above-mentioned main text generation big model is trained by the main text training unit, and the main text training unit (not shown in the figure) is configured to: use text positive training data to train the initial big model to obtain a trained main text generation big model, wherein the text positive training data is obtained by the main text training unit using the following steps: obtaining the initial basic corpus text including story background and character information; generating a basic outline of the initial basic corpus text through a pre-trained outline big model; generating a training main text based on the basic outline; combining the initial main text corpus text and the training main text to obtain text positive training data.

[0123] In some optional implementations of this embodiment, the above-mentioned work generation unit 305 is further configured to: in response to determining that the main text of the current stage does not include the ending keyword, generate and send the interactive content information of the current stage based on the main text of the current stage; in response to receiving feedback information on the interactive content information of the current stage, determine the key information of the current stage based on the feedback information; generate the main text of the next stage based on the key information of the current stage, the key information before the current stage and the basic information of the story, use the main text of the next stage as the main text of the current stage, and continue to determine whether the main text of the current stage includes the ending keyword until the work text of the work is generated.

[0124] In some optional implementations of the embodiment, the determination unit 304 is configured to: select a plurality of initial information from a preset knowledge base based on the initial text; determine the plot development direction word based on the user feedback information; and select the key information of the initial stage from the plurality of initial information based on the direction indicated by the plot development direction word.

[0125] In some optional implementations of the embodiment, the determination unit 304 is further configured to: input the initial text into a key information extraction large model to obtain text key information output by the key information extraction large model; and select a plurality of initial information related to the text key information from a preset knowledge base.

[0126] In some optional implementations of the embodiment, the key information extraction large model is trained by a key training unit (not shown in the figure), and the key training unit is configured to: train an initial large model by using text reverse training data to obtain a trained key information extraction large model, wherein the text reverse training data is obtained by the key training unit by the following steps: obtaining an initial text corpus including a story text; generating a text outline of the initial text corpus by a pre-trained outline large model; generating a training background and a character text based on the text outline; obtaining a text key text based on the text outline, the training background and the character text; and combining the initial text corpus and the text key text to obtain the text reverse training data.

[0127] In some optional implementations of the embodiment, the interactive generation unit 303 is configured to: match the initial text with a preset ending keyword; in response to that the initial text does not match the ending keyword successfully, determine that the initial text does not include the ending keyword; and input the initial text into a pre-trained interactive large model to obtain interactive content information output by the interactive large model.

[0128] The information generation apparatus provided by the embodiment of the present disclosure first generates an initial text of an initial stage according to preset story basic information by the text generation unit 301; secondly, the interactive generation unit 302 generates and sends interactive content information based on the initial text in response to determining that the initial text does not include an ending keyword; thirdly, the determination unit 303 determines key information of the initial stage based on user feedback information and the initial text in response to receiving the user feedback information; and finally, the work generation unit 304 generates a work text of a work based on the initial text, the key information and the story basic information. Thus, the interactive content information is generated based on the initial text, which can enable a user to participate in the creation of the work, and the work is continuously created according to the user feedback information of the user, thereby not only improving the participation of the user in the creation of the work, but also improving the diversity of the plot development of the work and the generation quality of the work.

[0129] In the technical solutions disclosed herein, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0130] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0131] Figure 4 A schematic block diagram of an example electronic device 400 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are provided as examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0132] like Figure 4 As shown, the device 400 includes a computing unit 401, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 402 or a computer program loaded from a storage unit 408 into a random access memory (RAM) 403. Various programs and data required for the operation of the device 400 can also be stored in the RAM 403. The computing unit 401, the ROM 402, and the RAM 403 are connected to each other via a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0133] Various components in device 400 are connected to I / O interface 405, including an input unit 406, such as a keyboard, mouse, etc.; an output unit 407, such as various types of displays, speakers, etc.; a storage unit 408, such as a magnetic disk, optical disk, etc.; and a communication unit 409, such as a network card, modem, wireless communication transceiver, etc. Communication unit 409 allows device 400 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0134] The computing unit 401 can be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 401 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units that run machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 401 performs the various methods and processes described above, such as the information generation method. For example, in some embodiments, the information generation method can be implemented as a computer software program that is tangibly contained in a machine-readable medium, such as a storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 400 via the ROM 402 and / or the communication unit 409. When the computer program is loaded into the RAM 403 and executed by the computing unit 401, one or more steps of the information generation method described above can be performed. Alternatively, in other embodiments, the computing unit 401 can be configured to perform the information generation method by any other appropriate means (e.g., by means of firmware).

[0135] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0136] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable information generating device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0137] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0138] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0139] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0140] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0141] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not limited herein.

[0142] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A method for generating information, the method comprising: Generate the initial body text of the initial stage based on the preset basic information of the story; In response to determining that the initial body text does not include an ending keyword, generating and sending interactive content information based on the initial body text; In response to receiving user feedback information, determining key information of the initial stage based on the user feedback information and the initial body text; wherein the key information is main information that affects the development of the plot; Generating a work text based on the initial body text, the key information, and the basic story information; The step of generating the work text based on the initial main text, the key information, and the basic story information includes: Based on the key information and the basic story information, a plot summary for the current stage is generated; wherein a story development direction related to the key information is determined from the basic story information; and the story development direction is plot-expanded to obtain a plot summary for the current stage; Generate the main text of the current stage based on the plot summary of the current stage and the key information; Determine whether the main text of the current stage includes the ending keyword; In response to determining that the main text of the current stage includes the ending keyword, a work text of the work is generated based on the initial main text, the main text of the current stage, and the main text of the previous stage.

2. The method according to claim 1, wherein Generating a plot summary of the current stage based on the key information and the basic story information includes: Inputting the key information and the basic story information into a plot summary generation model to obtain a plot summary output by the plot summary generation model; The plot summary output by the plot summary generation model is preprocessed to obtain a plot summary of the current stage.

3. The method according to claim 2, wherein: The plot summary generation large model is trained by the following steps: the initial large model is trained with plot positive training data to obtain the trained plot summary generation large model, wherein the plot positive training data is obtained by the following steps: an initial basic corpus text including story background and character information is obtained; a basic outline of the initial basic corpus text is generated by a pre-trained outline large model; a training plot text is generated based on the basic outline; and the initial main text corpus text and the training plot text are combined to obtain plot positive training data.

4. The method according to claim 1, wherein Generating the main text of the current stage based on the plot summary of the current stage and the key information includes: Inputting the key information and the plot summary of the current stage into the main text generation model to obtain the main text output by the main text generation model; The main text output by the main text generation model is preprocessed to obtain the main text of the current stage.

5. The method according to claim 4, wherein The text generation model is trained using the following steps: The initial large model is trained using text forward training data to obtain a trained text generation large model; wherein, the text forward training data is obtained through the following steps: obtaining an initial basic corpus text including story background and character information; generating a basic outline of the initial basic corpus text through a pre-trained outline large model; generating a training text based on the basic outline; and combining the initial text corpus text and the training text to obtain text forward training data.

6. The method according to claim 1, wherein Generating the work text based on the initial body text, the key information and the basic story information further includes: In response to determining that the main text of the current stage does not include the ending keyword, generating and sending interactive content information of the current stage based on the main text of the current stage; In response to receiving feedback information on the interactive content information of the current stage, determining key information of the current stage based on the feedback information; Based on the key information of the current stage, the key information before the current stage and the basic information of the story, the main text of the next stage is generated, the main text of the next stage is used as the main text of the current stage, and it is continued to be determined whether the main text of the current stage includes the ending keywords until the work text is generated.

7. The method according to claim 1, wherein In response to receiving user feedback information, determining the key information of the initial stage based on the user feedback information and the initial body text includes: Based on the initial body text, selecting a plurality of initial information from a preset knowledge base; Determining a plot development direction word based on the user feedback information; Based on the direction indicated by the plot development direction word, key information of the initial stage is selected from the multiple initial information.

8. The method according to claim 7, wherein: The selecting of a plurality of initial information from a preset knowledge base based on the initial body text includes: Inputting the initial body text into a key information extraction model to obtain the body key information output by the key information extraction model; A plurality of initial information related to the key information of the text is selected from a preset knowledge base.

9. The method according to claim 8, wherein The key information extraction model is trained by the following steps: The initial large model is trained using text reverse training data to obtain a trained key information extraction large model, wherein the text reverse training data is obtained using the following steps: obtaining an initial main text corpus text including the main text of the story; generating a main text outline of the initial main text corpus text through a pre-trained outline large model; generating a training background and character text based on the main text outline; obtaining a main text key text based on the main text outline, the training background and the character text; and combining the initial main text corpus text and the main text key text to obtain text reverse training data.

10. The method according to claim 1, wherein In response to determining that the initial body text does not include an ending keyword, generating and sending interactive content information based on the initial body text includes: Matching the initial body text with preset ending keywords; In response to the initial body text not being successfully matched with the ending keyword, it is determined that the initial body text does not include the ending keyword, and the initial body text is input into a pre-trained interactive model to obtain interactive content information output by the interactive model.

11. An information generating device, comprising: The body text generation unit is configured to generate an initial body text of the initial stage according to the preset basic information of the story; an interaction generating unit configured to generate and send interactive content information based on the initial body text in response to determining that the initial body text does not include an ending keyword; a determining unit configured to, in response to receiving user feedback information, determine key information of the initial stage based on the user feedback information and the initial body text; wherein the key information is main information that affects the development of the plot; a work generating unit configured to generate a work text of the work based on the initial body text, the key information and the basic story information; Wherein, the work generation unit is further configured to: Based on the key information and the basic story information, a plot summary for the current stage is generated; wherein a story development direction related to the key information is determined from the basic story information; and the story development direction is plot-expanded to obtain a plot summary for the current stage; Generate the main text of the current stage based on the plot summary of the current stage and the key information; Determine whether the main text of the current stage includes the ending keyword; In response to determining that the main text of the current stage includes the ending keyword, a work text of the work is generated based on the initial main text, the main text of the current stage, and the main text of the previous stage.

12. The device according to claim 11, wherein The work generation unit is further configured to: input the key information and the basic story information into a plot summary generation model to obtain a plot summary output by the plot summary generation model; and pre-process the plot summary output by the plot summary generation model to obtain a plot summary of the current stage.

13. The device according to claim 12, wherein The plot summary generation large model is trained by a plot training unit, and the plot training unit is configured to: train the initial large model with plot positive training data to obtain a trained plot summary generation large model, wherein the plot positive training data is obtained by the plot training unit using the following steps: obtaining an initial basic corpus text including story background and character information; generating a basic outline of the initial basic corpus text through a pre-trained outline large model; generating a training plot text based on the basic outline; and combining the initial body corpus text and the training plot text to obtain plot positive training data.

14. The device according to claim 11, wherein The work generation unit is further configured to: input the key information and the plot summary of the current stage into the main text generation model to obtain the main text output by the main text generation model; pre-process the main text output by the main text generation model to obtain the main text of the current stage.

15. The device according to claim 14, wherein The main text generation large model is trained by a main text training unit, and the main text training unit is configured to: use text positive training data to train the initial large model to obtain a trained main text generation large model, wherein the text positive training data is obtained by the main text training unit using the following steps: obtaining an initial basic corpus text including story background and character information; generating a basic outline of the initial basic corpus text through a pre-trained outline large model; generating a training main text based on the basic outline; and combining the initial main text corpus text and the training main text to obtain text positive training data.

16. The device according to claim 11, wherein The work generation unit is further configured to: in response to determining that the body text of the current stage does not include the ending keyword, generate and send interactive content information of the current stage based on the body text of the current stage; In response to receiving feedback information on the interactive content information of the current stage, determining key information of the current stage based on the feedback information; Based on the key information of the current stage, the key information before the current stage and the basic information of the story, the main text of the next stage is generated, the main text of the next stage is used as the main text of the current stage, and it is continued to be determined whether the main text of the current stage includes the ending keywords until the work text is generated.

17. The device according to claim 11, wherein The determining unit is configured to: select a plurality of initial information from a preset knowledge base based on the initial body text; determine a plot development direction word based on the user feedback information; Based on the direction indicated by the plot development direction word, key information of the initial stage is selected from the multiple initial information.

18. The device according to claim 17, wherein The determination unit is further configured to: input the initial body text into the key information extraction model to obtain the body key information output by the key information extraction model; and select multiple initial information related to the body key information from a preset knowledge base.

19. The method according to claim 18, wherein The key information extraction big model is trained by a key training unit, and the key training unit is configured to: train the initial big model with text reverse training data to obtain a trained key information extraction big model, wherein the text reverse training data is obtained by the key training unit using the following steps: obtaining an initial main text corpus text including the main text of the story; generating a main text outline of the initial main text corpus text through a pre-trained outline big model; generating a training background and character text based on the main text outline; obtaining a main text key text based on the main text outline, the training background and the character text; and combining the initial main text corpus text and the main text key text to obtain text reverse training data.

20. The device according to claim 11, wherein The interaction generation unit is configured to: match the initial body text with a preset ending keyword; in response to the initial body text not being successfully matched with the ending keyword, determine that the initial body text does not include the ending keyword, input the initial body text into a pre-trained interaction model, and obtain the interaction content information output by the interaction model.

21. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 10.

22. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to enable the computer to execute the method according to any one of claims 1 to 10.

23. A computer program product comprising a computer program, which, when executed by a processor, implements the method of any one of claims 1 to 10.

Images