File generation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention provides a file generation method and device, electronic equipment and a storage medium. The file processing method comprises the following steps: in response to first natural language input which is received by a session panel and indicates to generate a target file, determining an nth first question according to at least part of first to nth natural language input under the condition of determining that generation parameters are missing according to the first natural language input; wherein n is a positive integer; the nth first problem is used for acquiring first parameters which are not included in the generation parameters; based on the session panel, acquiring an (n + 1) th natural language input based on the nth first question output for the first time; based on the (n + 1) th natural language input, determining whether the generation parameters are complete or not; in response to incomplete generation parameters, according to at least part of natural language inputs from the first to the (n + 1) th, initially outputting the (n + 1) th first question; and in response to the generation parameter completeness, generating a target file according to the generation parameter.

Description

Technical Field

[0001] The present disclosure relates to but is not limited to the field of file processing technology, and in particular to a file generation method and device, an electronic device and a storage medium. Background Art

[0002] With the development of technology, various artificial intelligence (AI) models can do a lot of work. For example, AI models can automatically write documents.

[0003] However, the documents written by the AI ​​model need to meet the needs of users. Users may not be able to fully describe their needs at one time, so how to quickly obtain the user's writing needs and ensure that the documents written by AI meet the user's needs is a problem that needs to be further solved. Summary of the invention

[0004] Embodiments of the present disclosure provide a file generation method and device, an electronic device, and a storage medium.

[0005] A first aspect of an embodiment of the present disclosure provides a file generation method, the method comprising:

[0006] In response to the conversation panel receiving a first natural language input indicating generation of a target file, when it is determined according to the first natural language input that generation parameters are missing, determining at least part of the nth first question according to the first to nth natural language inputs; wherein n is a positive integer; the nth first question is used to obtain a first parameter that is not yet included in the generation parameters;

[0007] Based on the conversation panel, obtaining an n+1th natural language input based on the initial output of the nth first question;

[0008] Based on the (n+1)th natural language input, determining whether the generation parameters are complete;

[0009] In response to the generation parameter being incomplete, initially outputting the n+1th first question based on at least part of the first to n+1th natural language inputs;

[0010] In response to the generation parameters being complete, the target file is generated according to the generation parameters.

[0011] A second aspect of an embodiment of the present disclosure provides a file generation device, the device comprising:

[0012] A first determination module is configured to, in response to the conversation panel receiving a first natural language input indicating generation of a target file, determine at least part of an nth first question according to the first to nth natural language inputs when it is determined that generation parameters are missing according to the first natural language input; wherein n is a positive integer; and the nth first question is used to obtain a first parameter that is not yet included in the generation parameters;

[0013] An acquisition module, configured to acquire, based on the conversation panel, an (n+1)th natural language input based on the initial output of the (n)th first question;

[0014] A second determination module, configured to determine whether the generation parameters are complete based on the (n+1)th natural language input;

[0015] an output module, configured to, in response to the generation parameter being incomplete, initially output an n+1th first question based on at least a portion of the first to n+1th natural language inputs;

[0016] A generation module is used to generate the target file according to the generation parameters in response to the generation parameters being complete.

[0017] A third aspect of the present disclosure provides an electronic device, including:

[0018] processor;

[0019] a memory for storing processor-executable instructions;

[0020] Among them, the processor is configured to: implement the file processing method provided by any technical solution of the first aspect when running executable instructions.

[0021] A fourth aspect of an embodiment of the present disclosure provides a computer-readable storage medium, wherein the readable storage medium stores an executable program, wherein when the executable program is executed by a processor, the file processing method provided by any one of the solutions of the first aspect described above is implemented.

[0022] According to the technical solution provided by the embodiments of the present disclosure, the conversation panel guides the user to provide the generation parameters required to generate the target file in a natural language input manner through multiple rounds of questions, thereby achieving the smooth generation of the target file and / or generating a high-quality target file.

[0023] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0025] Figure 1 The figure is a flowchart of a file processing method according to an exemplary embodiment.

[0026] Figure 2A The present invention is a schematic diagram showing a display of a file interface and a session panel according to an exemplary embodiment.

[0027] Figure 2B The present invention is a schematic diagram showing a display of a file interface and a session panel according to an exemplary embodiment.

[0028] Figure 3 The figure is a flowchart of a file processing method according to an exemplary embodiment.

[0029] Figure 4 The figure is a schematic diagram showing the structure of a file processing device according to an exemplary embodiment.

[0030] Figure 5 The diagram is a schematic structural diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0031] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the embodiments of the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the embodiments of the present disclosure as detailed in the appended claims.

[0032] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0033] like Figure 1 As shown, the present disclosure provides a file processing method, the method comprising:

[0034] S1110: In response to the conversation panel receiving a first natural language input indicating generation of a target file, in the case where it is determined according to the first natural language input that generation parameters are missing, determining an nth first question according to at least part of the first to nth natural language inputs; wherein n is a positive integer; the nth first question is used to obtain a first parameter that is not yet included in the generation parameters;

[0035] S1120: Based on the conversation panel, obtaining an (n+1)th natural language input based on the initial output of the nth first question;

[0036] S1130: Determine whether the generated parameters are complete based on the (n+1)th natural language input;

[0037] S1140: In response to incomplete generation parameters, initially outputting the n+1th first question according to at least part of the first to n+1th natural language inputs;

[0038] S1150: In response to the generation parameters being complete, generate a target file according to the generation parameters.

[0039] The file processing method may be executed by an electronic device, which may include but is not limited to a terminal or a server, and the terminal may correspond to any form of terminal device. For example, the terminal device may include but is not limited to: a tablet computer, a desktop computer, a mobile phone, a vehicle-mounted device and / or a smart office device. The server may be an application server of a target application that generates a target file, etc.

[0040] If the target file is generated and executed locally on the terminal, it can also be executed by the server.

[0041] In some embodiments, the session panel may be a built-in session panel of the target application that generates the target file.

[0042] In other embodiments, the conversation interface can be a conversation panel of a file interface of an independent target file. The conversation interface can be used for a conversation between a user and a robot associated with an AI model.

[0043] In some embodiments, the target application may have an application window, in which an editing panel and a conversation panel may be provided. The target file may be manually edited in the editing panel. An AI dialogue based on a conversational language may be conducted in the conversation panel. The AI ​​dialogue may be conducted by an AI model in a natural language.

[0044] The editing panel and the session panel are arranged side by side in the application window, and usually, the area of ​​the editing panel is slightly larger than that of the session panel. FIG. 2A to FIG. 2B As shown, in FIG. 2A to FIG. 2B The left side of the figure shows the document interface of the presentation file, and the right side shows the conversation panel. In the conversation panel, there is a conversation in natural language.

[0045] In some other embodiments, the target application may have at least two small windows, which are respectively an editing window for manually editing the target file and a session window where the session panel is located.

[0046] When editing a target file based on a session panel, two small windows or two panels are displayed side by side. The two small windows can be displayed side by side horizontally or vertically. The two panels here are the aforementioned editing panel and session panel.

[0047] Specifically, it can be determined according to the aspect ratio of the display screen of the electronic device executing the target file. For example, if the aspect ratio is greater than 1, two small windows or two panels are displayed side by side horizontally. For another example, if the aspect ratio is less than 1, two small windows or two panels are displayed side by side vertically.

[0048] One or more controls related to target file editing may be provided in the conversation panel. In some embodiments, the conversation panel at least includes a conversation input box, which can be used for natural language input by the user.

[0049] The aforementioned AI model can run in the terminal or on the server. The AI ​​model can be trained using sample data. For example, a specified model is trained using sample data, and the loss value of the specified model is determined during the training process. When the loss value of the specified model is small enough, it is considered that the AI ​​model can be put online for the realization of specific functions. The AI ​​model here can be various neural network models and / or machine learning models.

[0050] Exemplarily, the AI ​​model involved in the file generation method in the embodiment of the present disclosure may include at least: a natural language processing model 1. The model 1 can be used to perform natural language-based dialogue processing with the user based on the conversation panel.

[0051] The AI ​​model can recognize the natural language input entered by the user, extract the semantics of each natural language input, and determine the generation parameters according to the semantics. The generation parameters can be the parameters required to generate the target file. The first parameter can be a parameter that has not yet been obtained in the generation parameters. That is, the first parameter is a parameter to be obtained in the generation parameters.

[0052] Exemplarily, the generation parameters of the target file may include but are not limited to one of the following:

[0053] The file type of the target file;

[0054] The purpose of the target document;

[0055] The length of the target file;

[0056] The style of the target file;

[0057] The target audience of the target document;

[0058] Parameter information such as the purpose of the target file.

[0059] The above are only examples of generating parameters, and the specific implementation is not limited to the above examples.

[0060] In some embodiments, the first natural language input may include a phrase, one or more natural sentences, and / or a natural language paragraph input from the conversation panel.

[0061] The target file may be a file of various types, such as a presentation file, a text file, and / or a table file, etc. The application program that displays the file interface and / or the session panel may support automatic generation and editing of multiple files.

[0062] Exemplarily, the AI ​​model may include an LLM (Large Language Model, or “large-scale language model”).

[0063] For example, AI models may include GPT (Generative Pre-Trained Transformer Model). The training process of the GPT model is: unsupervised pre-training on a large-scale corpus, and then fine-tuning on specific tasks to achieve better results. The GPT model can be used for various natural language processing tasks, such as text generation, machine translation, sentiment analysis, etc.

[0064] In some embodiments, based on the natural language in the 1st to nth natural language inputs that can be used to determine the generation parameters, it is determined which other parameters are still needed, and the nth first question is determined based on the still needed parameters.

[0065] For example, there may be sentences composed of meaningless function words in the first to nth natural language inputs, in which case such sentences can be eliminated, and the nth first question can be determined based on the natural language inputs containing content words with clear directional meanings. In some embodiments, after obtaining the n+1th natural language input based on the nth first question, the semantics of the n+1th natural language input is analyzed to determine whether a newly determined generation parameter has been obtained.

[0066] In other embodiments, after obtaining the n+1th natural language input based on the nth first question, the generation parameters are obtained by combining the 1st to n+1th natural language inputs. When this method is used, the correlation between the 1st to n+1th natural language inputs can be extracted, and the generation parameters for generating the target file can be determined as much as possible, thereby reducing the output of questions and improving the efficiency of obtaining generation parameters.

[0067] In some embodiments, after obtaining the n+1th natural language input, the semantics of the xth to n+1th natural language inputs are comprehensively analyzed to determine whether new generation parameters are obtained, where x is greater than 2 and less than n.

[0068] Using this method, considering that the correlation between continuously acquired natural language inputs is closer than the correlation between discontinuous natural language inputs, on the one hand, the generation parameters of possible extraction points can be obtained based on the correlation between natural language inputs as much as possible, and on the other hand, the length of the natural language input processed by the AI ​​model this time is reduced, thereby improving processing efficiency.

[0069] In response to the first natural language input indicating the generation of the target file, the AI ​​model continues to determine whether the generation parameters are complete based on the first natural language input. If not, the AI ​​model enters a multi-round dialogue stage with the user based on the conversation panel to guide the user to give the complete generation parameters in the form of natural language. That is, the generation parameters are improved by entering into a multi-round dialogue in natural language.

[0070] In some embodiments, the method of determining whether the generation parameter is missing may be as follows:

[0071] Compare the determined generation parameters with the parameters that must be obtained, and determine whether the generation parameters are complete through comparison;

[0072] or,

[0073] The determined generation parameters are written into a prompt word template, and the prompt word template is input into the AI ​​model, and the AI ​​model determines whether the prompt word template output is complete based on the result.

[0074] In some embodiments, the determined generation parameters are compared with the parameters that must be obtained, and the comparison is used to determine whether the generation parameters are complete, which may include but is not limited to:

[0075] Determining a topic of a target document based on one or more natural language inputs received by the conversation panel;

[0076] Determine a target parameter list according to the association relationship between the subject and at least one candidate parameter list; wherein the target parameter list includes generation parameters required to generate a corresponding subject target file;

[0077] According to the target parameter list and the already acquired generation parameters, determining whether there are any generation parameters in the target parameter list but not acquired;

[0078] If there are generation parameters that have not been obtained in the target parameter list, it means that the current generation parameters are incomplete, otherwise the current production parameters are complete.

[0079] In some embodiments, determining the target parameter list according to the association relationship between the subject and at least one candidate parameter list may further include:

[0080] The target parameter list is selected from at least one candidate parameter list according to the subject, the acquired generation parameters and the association relationship.

[0081] In other embodiments, prompt words are generated based on natural language input, and the prompt words are input into an AI model, and the AI ​​model outputs generation parameters required to generate a target file related to the subject. For example, the AI ​​model outputs a target parameter list; the parameters included in the target parameter list are complete generation parameters for generating the target file of the subject.

[0082] In some embodiments, a first template is constructed according to a topic involved in the natural language input. The first template is a type of prompt word template. The first template contains complete generation parameters, and the complete generation parameters include: parameter 1, parameter 2... parameter N.

[0083] A second template is constructed according to the received natural language input, and the second template is a template of the natural language input. The second template may include part or all of the received natural language input. The received natural language input includes: natural language input 1, natural language input 2, ..., natural language input m, where m is an integer.

[0084] The first template and the second template are input into the AI ​​model to obtain the output result of the AI ​​model's judgment on whether complete generation parameters have been obtained and / or the production parameters that have not yet been obtained.

[0085] In some embodiments, the nth first question is a question that has never been output by the conversation panel, so as to avoid the conversation panel from constantly repeating the same question.

[0086] After receiving a natural language input, semantic extraction will be performed on the natural language input to determine whether the natural language received this time has additional generation parameters. If so, it will be determined whether the generation parameters are complete.

[0087] If the supplement is not complete, the next first question that has not been output and is used to obtain the first parameter that has not been obtained will be continued. All first questions are output in natural language.

[0088] The nth first question gives a prompt for the user to supplement the first parameter. The first question can be in various forms, and can be specifically at least one of the following:

[0089] Open-ended questions such as, What type of document are you writing?

[0090] Closed-ended questions, such as, is the file type you want to edit a text file, a presentation file, or an excel file?

[0091] In some embodiments, the form of the first question may be determined according to the first parameter to be acquired.

[0092] For example, an open-ended question may be asked about the subject of the document, or a closed-ended question may be asked with several length options for the length of the document.

[0093] In summary, in the disclosed embodiment, considering that the user may not be able to generate the required parameters by outputting the target file in the conversation panel at one time, multiple rounds of natural language are used to guide the user to answer, thereby obtaining the generated parameters. However, when multiple rounds of conversations are conducted, each round of conversation will determine the next natural language question to be output based on the existing natural language input, thereby reducing repeated questions and unnecessary question output.

[0094] In the disclosed embodiment, after the complete generation parameters are acquired, the target file is automatically generated without the user providing specific content.

[0095] Exemplarily, the AI ​​model queries the knowledge base based on the file subject in the generation parameters to automatically compose a target file that meets the writing requirements reflected by the generation parameters, thereby achieving intelligent generation of the target file.

[0096] In some embodiments, determining the nth first question based on at least part of the first to nth natural language inputs includes:

[0097] Querying a question bank, wherein the question bank contains one or more second questions constructed in a natural language form, wherein the second questions are candidate questions for obtaining generation parameters required to generate a target file;

[0098] According to at least part of the first to nth natural language inputs and the association relationship between different second questions in the question bank, the nth first question output in the conversation panel is selected from the second questions that have not yet been output.

[0099] In some embodiments, a question bank is pre-set, and a plurality of second questions are pre-set in the question bank, and the plurality of second questions are all candidate questions for obtaining generation parameters in the target file.

[0100] By adopting the form of a question bank, there is no need to generate the first question on site (temporarily), thereby reducing delays.

[0101] Any question in the question bank (i.e., the second question) is constructed in natural language. The second question may include: a closed question and / or an open question.

[0102] Since there are many questions in the question bank, in the present embodiment, the first question to be outputted next and not outputted before is determined based on at least part of the 1st to nth natural language inputs obtained.

[0103] In some embodiments, question priorities are set between the second questions in the question bank, and / or jump relationships are set between questions, so that the conversation panel extracts question outputs for further responses from questions that have not been answered based on the natural language input already given by the user to obtain complete generation parameters.

[0104] On the one hand, by asking questions, we can obtain the results by adopting this method, and on the other hand, we can reduce the output of repeated questions.

[0105] Similarly, determining the next first question to be output and not outputted based on at least part of the 1st to nth natural language inputs may include: determining the next first question to be output and not outputted based on the 1st to nth natural language input query question bank, or determining the next first question to be output and not outputted based on the Yth to nth natural language input query question bank, or determining the next first question to be output and not outputted based on the parameters determined by the 1st to nth natural language input query question bank. Y may be a positive integer greater than or equal to 2 and less than n.

[0106] In some embodiments, according to the association relationship between the first to nth natural language inputs and different second questions in the question bank, selecting the nth first question output in the conversation panel from the second questions that have not yet been output includes:

[0107] Determine, based on at least a portion of the first to nth natural language inputs, a second parameter obtained in the generation parameters;

[0108] According to the second parameter, select part or all of the second questions that have not been output from the question bank to construct an alternative question set;

[0109] The nth first question is selected based on the relevance of each second question in the set of alternative questions.

[0110] The second parameter is a parameter in the generated parameters that has been obtained according to one or more natural language inputs. Thus, the parameters to be obtained other than the second parameter in the generated parameters can be known according to the second parameter.

[0111] In this way, the query questions of the parameters to be obtained are used as candidate questions to be output, and then questions are selected from the candidate questions to be output. The nth first question to be output is selected, thereby ensuring that the output questions are output from high to low importance.

[0112] In some embodiments, the association relationship includes at least one of the following:

[0113] Jump relationship when multiple second questions are output;

[0114] Output priority of multiple second questions.

[0115] For example, for question 3, if a negative answer is received for question 3, the process jumps to question 4, and if a positive answer is received for question 3, the process jumps to question 10. For another example, if question 3 is answered, the process jumps to question 5, and / or, if it is not answered, the process jumps to question 4. This is an example of a jump relationship.

[0116] The plurality of second questions have priorities, and questions of the same priority level can be randomly selected for output, and questions of different priority levels are outputted with priority to the second questions of higher priority. In short, the association relationship between the plurality of second questions mentioned above can be determined in this way.

[0117] In some embodiments, the nth first question is determined according to the 1st to nth natural language inputs; wherein n is a positive integer; the nth first question includes:

[0118] Based on the 1st to nth natural languages ​​and the generation parameters to be completed, an nth first question is generated.

[0119] The nth first question can be automatically generated by the AI / ML model.

[0120] If there are multiple generation parameters to be completed, the most important parameter to be completed can be selected to construct the nth first problem according to the importance of the generation parameters to be completed.

[0121] In some embodiments, based on the first to nth natural languages ​​and the generation parameters to be completed, the nth first question is generated, which may include but is not limited to the following examples:

[0122] According to the first to nth natural language inputs, a determined generation parameter is obtained;

[0123] Determine generation parameters to be completed according to the determined generation parameters;

[0124] Generate the parameters to be generated in the session panel according to the generation parameters to be completed.

[0125] For example, part or all of the first to nth natural language inputs are identified to obtain determined generation parameters, and the generation parameters to be completed are determined according to the aforementioned target parameter list.

[0126] In some embodiments, a prompt word template is written according to the generation parameters to be completed;

[0127] The prompt word template is input into the AI ​​model to obtain the first question to be output generated by the AI ​​model.

[0128] In some other embodiments, part or all of the first to nth natural language inputs are written into a natural language template;

[0129] Input the natural language template into the AI ​​model, obtain the AI ​​model output generation parameters to be completed, and obtain the first problem of the generation parameters to be completed.

[0130] In some embodiments, part or all of the first to nth natural language inputs and the generated parameters to be completed or the completed generated parameters are written into different parts of the template, and then the template is input into the AI ​​model. At this time, the AI ​​model will output the generated parameters to be completed and the first question to obtain the generated parameters to be completed.

[0131] In some embodiments, a template may include natural language input and generation parameters to be completed.

[0132] The template may include: the natural language input is: natural language input 1, natural language input 2 ... natural language input M. The template may also include: the generation parameters to be completed may include: parameter 1, parameter 2 ... parameter x.

[0133] or,

[0134] The template may include: the natural language input is: natural language input 1, natural language input 2 ... natural language input M. The template may also include: the complete generation parameters may include: parameter 1, parameter 2 ... parameter Y.

[0135] Of course, the above are just examples, and the specific implementation is not limited to the above examples.

[0136] In some embodiments, the method further comprises:

[0137] Before determining the nth first question according to the first to nth natural language inputs, determining whether a preset event occurs;

[0138] In response to detecting a preset event, the selection and output of the nth first question is stopped to change the way of obtaining the missing generation parameters.

[0139] In some embodiments, the preset event may be any event indicating that the user does not want to be asked questions again.

[0140] After the preset event is detected, the selection and output of the nth first question is stopped to change the way of obtaining the missing generation parameters, thereby improving the user experience when the target file is generated.

[0141] In some embodiments, after stopping the selection and output of the nth first question, the method further comprises:

[0142] Generate a prompt word (prompt) by inputting all or part of the first to nth natural language;

[0143] The first model generates parameters based on the prompt word completion.

[0144] In some embodiments, the first model is used to automatically complete the generation parameters based on a prompt.

[0145] For example, after the first model obtains the prompt word, it also obtains the user information of the editing user, and combines the user information and the prompt word to complete the generation parameters.

[0146] Assuming that the user is a third-grade primary school student, if the natural language input instructs him to write an essay about autumn, the third-grade primary school student's essay is about 800 words, which is obviously not 10,000 words. At this time, some generation parameters can be derived based on the user information. For example, the generation parameters for this scenario may include a length of about 800.

[0147] The prompt word may include at least information for determining the subject of the target document.

[0148] In some embodiments, the first model can also complete the generation parameters based on the prompt word alone. The theme is to write a PPT about dry ice production. The first model uses "dry ice" as the keyword prompt word, and knows whether it is necessary to write a popular science presentation file or a practical presentation file for probability judgment. Then based on the probability judgment, it is highly likely that a popular science presentation file is to be written, and the generation parameters are determined according to the length of the popular science presentation file.

[0149] In summary, in some cases, the first model can be used to deduce generation parameters that have not yet been acquired in combination with one or more natural languages ​​that have already been acquired.

[0150] Exemplarily, the step of generating prompt words from the first to the nth natural language inputs may specifically include:

[0151] Input at least part of the first to nth natural language inputs into an AI model for natural language processing. Exemplarily, the AI ​​model for generating the prompt word may be a second model. The second model here may be another AI model, which may correspond to the same AI model as the aforementioned first model, or may be a different AI model.

[0152] The AI ​​model processes the natural language by performing keyword extraction, semantic extraction, and association between natural language inputs, and outputs one or more prompt words. Exemplarily, the AI ​​model is used to extract keywords from the natural language, construct prompt words, and form a prompt word template.

[0153] Exemplarily, these prompt words are written into an empty template of prompt words. Optionally, these prompt words are filled into the empty template according to a preset order to obtain a prompt word template.

[0154] The prompt word template is input into another AI model (ie, the first model), for example, the LLM model to obtain the complete generation parameters output by the LLM model.

[0155] In some embodiments, after the prompt word template and the determined generation parameters are spliced, they are input into the first model, and the first model will complete the missing parameters in the determined generation parameters to obtain complete generation parameters.

[0156] In some embodiments, the first model also has the function of correcting generation parameter errors. For example, when the generation parameters determined by the first model based on the prompt words conflict with the acquired generation parameters, the error correction and / or error correction function will be automatically started so that the obtained complete generation parameters do not conflict with each other. For example, according to the acquisition method of the acquired generation parameters, the generation parameters determined by the first model are used for replacement, or the generation parameters input by the user are used as the basis and the generation parameters completed by the first model are discarded, or the first model will combine two generation parameters that have a certain conflict to obtain complete generation parameters. In some embodiments, the setting method of the prompt words can be generated according to at least one of the following:

[0157] Natural language input entered by the user:

[0158] Keywords extracted based on natural language input;

[0159] Additional content.

[0160] In some embodiments, before extracting the prompt word, the natural language input and the additional content are combined to obtain the target content.

[0161] For example, the additional content may include but is not limited to at least one of the following:

[0162] Account information of the account associated with the target file;

[0163] The intent of the target file.

[0164] The target content is processed by a designated model capable of natural language processing or text processing to obtain output prompt words.

[0165] Exemplarily, the account information may include a user portrait and / or document editing habit information, etc.

[0166] In some embodiments, the preset event includes at least one of the following:

[0167] whether the number of first questions output after receiving the first natural language input is equal to or greater than a specified number;

[0168] Whether the duration of the multi-round conversation based on the conversation panel reaches the first duration;

[0169] A stop operation is detected, wherein the stop operation includes: an operation acting on a stop questioning control in a conversation panel;

[0170] When the duration of multiple rounds of conversations based on the conversation panel has reached a second duration and a complete third parameter has been obtained, wherein the second duration is less than the first duration; the generated parameters include a third parameter and a fourth parameter, wherein the importance of the third parameter is stronger than that of the fourth parameter.

[0171] If the number of questions output in the conversation panel has reached the specified number, it means that continuing to ask questions will make the user disgusted. In this case, in order to improve the user experience, the method of obtaining the generation parameters will be changed.

[0172] If a stop operation based on a physical button or virtual control input is detected by the customer, the questioning will also stop.

[0173] If the question time is long enough, the user experience will be poor, so another way to obtain the generation parameters can be used.

[0174] Of course, in this case, a prompt message may be output before changing the method for obtaining the generation parameters, and after receiving a confirmation reply from the user, the method for obtaining the generation parameters may be changed.

[0175] In some embodiments, the generated parameters may be divided into two parts, one of which is more important than the other. In this case, the questioning is stopped when the third parameter is obtained.

[0176] In short, by setting the above preset events, the generation intelligence of the target file and the user experience can be improved.

[0177] In some embodiments, the method further comprises:

[0178] When the m+1th natural language input for the mth first question does not answer the mth first question, the second question with the same meaning but different description as the mth second question is determined as the m+1th first question; wherein the mth first question is the second question for obtaining the third parameter; and m is a positive integer.

[0179] Exemplarily, if no option check or option input is received for a closed-ended question within a specific time period starting from the output of the mth first question, or if no natural language input related to an open-ended question is received within a specific time period starting from the output of the mth first question, it means that the user has not answered the mth first question.

[0180] In the disclosed embodiment, when the conversation panel outputs a question that is not answered by the user or the user replies to the question, the corresponding generation parameters cannot be determined. In this case, it is possible that the user does not correctly understand the question itself, so the question is output in a different way to obtain important generation parameters. In this way, the AI / ML model is used to simulate the user's intelligent question and answer, so as to obtain the required generation parameters.

[0181] In some embodiments, m can be a positive integer less than or equal to n. That is, when it is found in the process of outputting questions that no user response is obtained regarding particularly critical generation parameters, the first question for obtaining the same key parameter is output with a different description to ensure that the most critical parameter among the generation parameters is successfully obtained.

[0182] In some embodiments, the first to nth natural languages ​​are input to generate prompt words, including:

[0183] Get the account information of the account associated with the target file;

[0184] The first to nth natural language inputs and account information are used to generate prompt words.

[0185] The account information of the associated account may include but is not limited to at least one of the following:

[0186] Preference information for linked accounts;

[0187] User attribute information of the associated account.

[0188] The user attribute information may include but is not limited to at least one of the following:

[0189] User age;

[0190] User's gender;

[0191] User education level;

[0192] The industry in which the user works.

[0193] For example, based on the user's age and gender, it can be inferred that young men may like metallic styles, and young women may like pink styles.

[0194] The preference information of the associated account may include at least one of the following:

[0195] The editing account of the target file;

[0196] The account for creating the target file;

[0197] The login account of the client that opens the target file;

[0198] Specify an account. For example, if user A wants to edit his target file in the style that user B likes, he can specify account B.

[0199] The management account of the target file, the account level of the management account may include: the creation account of the presentation file and / or the upper level or multiple levels of accounts of the creation account.

[0200] Preference information may include but is not limited to at least one of the following:

[0201] The associated account is marked with the generated parameters it likes to use;

[0202] The generation template most frequently used by the associated account; the generation template includes one or more pre-set generation parameters.

[0203] Generate parameters that the associated account does not like;

[0204] The associated account does not use the generated parameters;

[0205] Generate templates that have not been used by the associated account.

[0206] In some embodiments, if the electronic device starts an editing application for a presentation file but does not log in using an account, the associated account may include the account that is most frequently used by the editing application.

[0207] In this way, based on the account information of the associated account, a presentation file with the personal characteristics or preferences of the associated account will be generated, thereby realizing personalized AI / ML generation of the presentation file.

[0208] In summary, in the disclosed embodiment, in addition to obtaining generation parameters by asking questions, the generation parameters can also be automatically completed based on the prompt words and account information through the first model.

[0209] In some other embodiments, after stopping the selection and output of the nth first question, the method further includes:

[0210] Output the recommended parameters for missing generation parameters;

[0211] Completes generated parameters based on feedback operations acting on the generated parameters.

[0212] When the necessary parameters for generating the target file have not been obtained when a preset event is detected, the AI ​​and / or ML model can propose one or more sets of recommended parameters and output these recommended parameters in the session panel for the user to select, thereby completing all remaining generation parameter completions at one time.

[0213] The feedback operation may include but is not limited to at least one of the following:

[0214] Selection operations for multiple sets of feedback parameters;

[0215] A selection operation for a set of feedback parameters;

[0216] A selection operation or a cancellation operation on a set of different options with missing parameters, etc.

[0217] This is a method of semi-automatically obtaining generation parameters by AI and / or ML models to improve the generation effect of the target file.

[0218] like Figure 3 As shown, the present disclosure provides a file processing method, the method comprising:

[0219] S1210: In response to the conversation panel receiving a first natural language input indicating generation of a target file, in the case where it is determined according to the first natural language input that generation parameters are missing, determining an nth first question according to at least part of the first to nth natural language inputs; wherein n is a positive integer; the nth first question is used to obtain a first parameter that is not yet included in the generation parameters;

[0220] S1220: Based on the conversation panel, obtain an (n+1)th natural language input based on the initial output of the nth first question;

[0221] S1230: Determine whether the generated parameters are complete based on the (n+1)th natural language input;

[0222] S1240: In response to incomplete generation parameters, initially outputting the n+1th first question based on at least part of the first to n+1th natural language inputs;

[0223] S1250: Write a document outline according to the generated parameters;

[0224] S1260: Generate a target file according to the file outline.

[0225] In the disclosed embodiment, after the electronic device conducts multiple rounds of language dialogue based on the conversation panel, it first writes a file outline based on the obtained generation parameters, and then generates a target file based on the file outline.

[0226] In some embodiments, while the file outline is being generated, generation of the target file is started according to the generated file outline, for example, receiving input information from the UI to generate the target file is simulated, thereby improving the generation efficiency of the target file.

[0227] In other embodiments, after the file outline is completely generated, the target file may be generated according to the complete file outline.

[0228] It is worth noting that: in this embodiment, a file outline is first generated according to the generation parameters, and then a target file is generated according to the file outline. In some cases, the AI ​​model can directly generate the target file according to the generation parameters and skip the step of generating the outline, thereby improving the generation efficiency of the target file.

[0229] In some cases, after the file outline is generated, it may not be displayed, and the target file can be written directly according to the outline device.

[0230] In some implementations, the method further comprises:

[0231] In the Session panel, display the file outline;

[0232] Updates the document outline based on partial or complete changes to the document outline.

[0233] In some embodiments, the third model generates a prompt word (prompt) according to the semantic information of the generation parameter, and generates a document outline according to the prompt. The generated document outline may be displayed in the session panel or may not be displayed in the session panel.

[0234] The file outline can be a file outline in any file format, such as a presentation file format, or a text format such as a word text format or a TXT text format.

[0235] According to the position sequence of each outline content in the document outline, each outline content is arranged sequentially in the first interface.

[0236] Exemplarily, each outline content may be written word for word, so that the outline content may be displayed one by one in a streaming manner in the session panel, thereby improving the display effect of the session panel and improving the friendliness of the interaction with the user.

[0237] In the disclosed embodiment, the generated file outline will be displayed in the session panel, and the user is allowed to modify part or all of the file outline manually or by voice. If the file outline is updated, the updated file outline will generate the target file, or the target file will be updated according to the updated file outline.

[0238] In some examples, the session panel may be displayed in the application window in response to a user triggering a shortcut key, or the session panel may be displayed in the application window in response to a user clicking a predefined function menu.

[0239] In some examples, one or more controls related to the processing of the document outline may be provided in the conversation panel. For example, the conversation panel at least includes a conversation input box, which can be used to receive a natural language input of the user.

[0240] In some examples, the document outline includes multiple outline contents, such as a topic and multiple headings with hierarchical information. Headings of different levels include first-level headings, second-level headings, third-level headings, and so on. For example, Heading 1 is a first-level heading, Heading 2 is a first-level heading, Heading 1.1 is a second-level heading, and Heading 1.1.1 is a third-level heading. Heading 1.1 is the next-level heading of Heading 1, and Heading 1.1 is the previous-level heading of Heading 1.1.1.

[0241] In some embodiments, the generation parameters include at least one of the following:

[0242] Subject of the document;

[0243] File length;

[0244] Document audience;

[0245] Application scenarios of files;

[0246] File style.

[0247] The present disclosure also provides a model training method for training an AI / ML model capable of executing any of the aforementioned file processing methods. The method may include:

[0248] Based on natural language and alternative models, alternative parameters output by the filing model are obtained;

[0249] The next question output based on the alternative parameters is used to obtain natural language input again;

[0250] When relatively complete candidate parameters are finally obtained, a generated file is obtained;

[0251] Get the score of the generated file;

[0252] The score is used to determine whether the model needs to be trained further.

[0253] If the score of the generated file reaches the specified value, the training can be stopped.

[0254] For example, the score of the generated file is judged from multiple dimensions, and the device calculates the score value. For example, the score is obtained by scoring the file subject, file length, whether the text content of the file has grammatical problems, and whether it meets the expectations of the training user.

[0255] In some embodiments, the method further comprises:

[0256] During the use of the target model, the model is trained online or enhanced based on user feedback to improve the processing accuracy of the model.

[0257] In some embodiments, the model after offline training may be a general model. Subsequently, according to the file editing operation of the target account, the general model is further trained or enhanced to obtain a personalized model for the target account.

[0258] The personality model is an object selected by different contents and can show the personal characteristics and / or style of the user to whom the target account belongs.

[0259] In some embodiments, there are multiple models for processing the target file, and these multiple models may constitute a model set.

[0260] The model set may include: a natural language processing model, which can process natural language to obtain a functional operation specified by a natural language input;

[0261] According to the functional operation, the AI / ML model that can implement the above functional operation is called from the model collection to edit the target file.

[0262] For example, the model set may include but is not limited to at least one of the following:

[0263] Generate parameter acquisition model;

[0264] The target file generation model;

[0265] Outline generation model for target documents;

[0266] The content of the target file populates the model;

[0267] The style processing model of the target file;

[0268] Update model of target file.

[0269] These models can be executed in parallel and can exist as parallel branches of the model processed by the target file. In this case, when the target file is finally generated, on the one hand, the outline generation and logical organization of the file are completed, and on the other hand, the content is filled according to the outline, and the third party sets the style of the target file.

[0270] In this case, the generation parameter acquisition model, the target file generation model, the outline generation model, the content filling model, the style processing model, and the update model can all serve as subsequent models of the natural language processing model. Through model cascading or mutual calling, and with the help of the natural language processing capabilities of the natural language processing model, various tasks can be completed for the target file.

[0271] The update module can be used for minor modifications after the target file is generated.

[0272] In some embodiments, after the initial generation of the target file is completed, the style processing model can be called to set the style of the target file, and the content filling model can be called to update the filling content in the target file.

[0273] In some embodiments, the target file generation model can directly generate the target file according to the generation parameters, that is, skip the generation of the file outline and directly generate the target file. If the generation parameter acquisition model obtains the complete generation parameters, the target file generation model is directly called to generate the target file, or the outline generation model is called to generate the file outline. After the file outline is generated, the content filling model is called to generate the target file.

[0274] Of course, the above is only an example of the model for processing the target file, and the specific implementation is not limited to this example.

[0275] like Figure 4 As shown, the present disclosure provides a file generation device including:

[0276] The first determination module 110 is configured to, in response to the conversation panel receiving a first natural language input indicating generation of a target file, determine an nth first question according to at least part of the first to nth natural language inputs in the case where it is determined that generation parameters are missing according to the first natural language input; wherein n is a positive integer; the nth first question is used to obtain a first parameter that is not yet included in the generation parameters;

[0277] An acquisition module 120, configured to acquire, based on the conversation panel, an (n+1)th natural language input based on the initial output of the (n)th first question;

[0278] A second determination module 130, for determining whether the generated parameters are complete based on the (n+1)th natural language input;

[0279] An output module 140 is configured to initially output the n+1th first question based on at least part of the first to n+1th natural language inputs in response to incomplete generation parameters;

[0280] The generation module 150 is configured to generate a target file according to the generation parameters in response to the generation parameters being complete.

[0281] In some embodiments, the file generating device may correspond to various electronic devices.

[0282] In some embodiments, the first determination module 110 , the acquisition module 120 , the second determination module 130 , the output module 140 , and the generation module 150 may all be program modules, which can implement the above operations after being executed by a processor.

[0283] In other embodiments, the first determination module 110, the acquisition module 120, the second determination module 130, the output module 140 and the generation module 150 can all be soft-hard combination modules, which include but are not limited to various programmable arrays; the programmable arrays include but are not limited to field programmable arrays and / or complex programmable arrays.

[0284] In some other embodiments, the first determination module 110, the acquisition module 120, the second determination module 130, the output module 140 and the generation module 150 may all be pure hardware modules, which include but are not limited to application specific integrated circuits.

[0285] In some embodiments, the output module 140 is configured to query a question bank, wherein the question bank contains one or more second questions constructed in natural language form, wherein the second questions are alternative questions for obtaining generation parameters required to generate a target file; based on the 1st to nth natural language inputs and the association relationship between different second questions in the question bank, the nth first question output in the session panel is selected from the second questions that have not yet been output.

[0286] In some embodiments, the output module 140 is configured to determine the second parameter obtained in the generation parameter based on the 1st to nth natural language inputs; based on the second parameter, select part or all of the second questions that have not yet been output in the question bank to construct a set of alternative questions; and select the nth first question based on the correlation between the second questions in the set of alternative questions.

[0287] In some embodiments, the output module 140 is configured to generate the nth first question based on the 1st to nth natural languages ​​and the generation parameters to be completed.

[0288] In some embodiments, the apparatus further comprises:

[0289] A third determination module is configured to determine whether a preset event occurs before selecting the nth first question;

[0290] The stopping module is configured to stop the selection and output of the nth first question in response to detecting a preset event, so as to change the way of obtaining the missing generation parameters.

[0291] In some embodiments, after stopping the selection and output of the nth first question, the apparatus further comprises:

[0292] The generating module 150 is configured to input the first to nth natural languages ​​and generate prompt words;

[0293] The completion module is configured to complete the generation parameters based on the prompt word by the first model.

[0294] In some embodiments, the preset event includes at least one of the following:

[0295] whether the number of first questions output after receiving the first natural language input is equal to or greater than a specified number;

[0296] Whether the duration of the multi-round conversation based on the conversation panel reaches the first duration;

[0297] A stop operation is detected, wherein the stop operation includes: an operation acting on a stop questioning control in a conversation panel;

[0298] When the duration of multiple rounds of conversations based on the conversation panel has reached a second duration and a complete third parameter has been obtained, wherein the second duration is less than the first duration; the generated parameters include a third parameter and a fourth parameter, wherein the importance of the third parameter is stronger than that of the fourth parameter.

[0299] In some embodiments, the apparatus further comprises:

[0300] The fourth determination module is configured to determine the second question with the same meaning but different description as the mth second question as the m+1th first question when the m+1th natural language input for the mth first question does not answer the mth first question; wherein the mth first question is the second question for obtaining the third parameter; and m is a positive integer.

[0301] In some embodiments, the generation module 150 is configured to obtain account information of an account associated with the target file; and generate prompt words using the first to nth natural language inputs and the account information.

[0302] In some embodiments, the output module 140 is further configured to output the recommended parameters of the missing generation parameters after stopping the selection and output of the nth first question;

[0303] The device also includes:

[0304] The completion module is configured to complete the generation parameters based on the feedback operation acting on the generation parameters.

[0305] In some embodiments, the generation module 150 is configured to write a file outline according to generation parameters; and generate a target file according to the file outline.

[0306] In some embodiments, the apparatus further comprises:

[0307] A display module, configured to display a document outline in a session panel;

[0308] The update module is configured to update the file outline based on a partial or complete modification operation of the file outline.

[0309] In some embodiments, the generation parameters include at least one of the following:

[0310] Subject of the document;

[0311] File length;

[0312] Document audience;

[0313] Application scenarios of files;

[0314] File style.

[0315] In some embodiments, the association relationship includes:

[0316] Jump relationship when multiple second questions are output;

[0317] Output priority of multiple second questions.

[0318] An embodiment of the present disclosure further provides an electronic device, including:

[0319] processor;

[0320] a memory for storing processor-executable instructions;

[0321] The processor is configured to implement the file processing method of any embodiment of the present disclosure when running the executable instructions.

[0322] The electronic device can at least be a server and / or a terminal. One or more servers can constitute a service end.

[0323] The memory may include various types of storage media, which are non-transitory computer storage media that can continue to remember information stored thereon after the communication device loses power.

[0324] The processor can be connected to the memory through a bus, etc., to read the executable program stored in the memory, for example, to implement the following Figure 1 or Figure 3 Any of the file processing methods shown.

[0325] The embodiments of the present disclosure further provide a computer-readable storage medium, wherein the computer-readable storage medium stores an executable program, wherein when the executable program is executed by a processor, the file processing method of any embodiment of the present disclosure is implemented. Figure 1 or Figure 3 Any of the file processing methods shown.

[0326] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0327] The present disclosure provides an electronic device, comprising: a processor; a memory for storing processor executable instructions; wherein the processor is configured to: implement the file processing method provided by any of the above technical solutions when running the executable instructions, for example, to implement the following Figure 1 or Figure 3 Any of the file processing methods shown.

[0328] The electronic device may be the aforementioned first client, second client and / or server.

[0329] like Figure 5 As shown, an embodiment of the present disclosure shows a structure of an electronic device. For example, the electronic device 900 can be provided as a terminal and / or a server having a client installed therein. Figure 5 The electronic device 900 includes a processing component 922, which further includes one or more processors, and a memory resource represented by a memory 932 for storing instructions executable by the processing component 922, such as an application. The application stored in the memory 932 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 922 is configured as an executable program to perform any of the aforementioned methods for applying to the access device, for example, to implement the following method: Figures 1 to 3 Any of the file processing methods shown.

[0330] The electronic device 900 may also include a power supply component 926 configured to perform power management of the electronic device 900, a wired or wireless network interface 950 configured to connect the electronic device 900 to a network, and an input / output (I / O) interface 958. The electronic device 900 may operate based on an operating system stored in the memory 932, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™ or the like.

[0331] In an exemplary embodiment, an embodiment of the present disclosure provides a computer-readable storage medium, which stores an executable program, wherein the executable program, when executed by a processor, implements the file processing method provided by any of the aforementioned technical solutions.

[0332] The computer-readable storage medium may specifically be: a non-temporary computer-readable storage medium.

[0333] Exemplarily, for example, the memory 932 includes an executable program, and the above instructions can be executed by the processing component 922 of the electronic device 900 to complete the above method.

[0334] The non-transitory computer-readable storage medium may be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, and the like.

[0335] Those skilled in the art will readily appreciate other implementations of the disclosed embodiments after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the disclosed embodiments, which follow the general principles of the disclosed embodiments and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and examples are to be considered as exemplary only, and the true scope and spirit of the disclosed embodiments are indicated by the following claims.

[0336] It should be understood that the embodiments of the present disclosure are not limited to the precise structures described above and shown in the drawings, and various modifications and changes may be made without departing from the scope thereof. The scope of the embodiments of the present disclosure is limited only by the appended claims.

Claims

1. A file generation method, characterized in that: The method comprises: In response to the conversation panel receiving a first natural language input indicating generation of a target file, in the case where it is determined according to the first natural language input that generation parameters are missing, determining an nth first question according to at least part of the first to nth natural language inputs; wherein n is a positive integer; the nth first question is used to obtain a first parameter that is not yet included in the generation parameters; Based on the conversation panel, obtaining an n+1th natural language input based on the initial output of the nth first question; Based on the (n+1)th natural language input, determining whether the generation parameters are complete; In response to the generation parameter being incomplete, initially outputting the n+1th first question based on at least part of the first to n+1th natural language inputs; In response to the generation parameters being complete, the target file is generated according to the generation parameters.

2. The method according to claim 1, wherein: The step of determining the nth first question based on at least part of the first to nth natural language inputs comprises: Querying a question bank, wherein the question bank includes one or more second questions constructed in a natural language form, wherein the second questions are candidate questions for obtaining generation parameters required to generate a target file; According to the association relationship between the 1st to the nth natural language inputs and different second questions in the question bank, the nth first question output in the conversation panel is selected from the second questions that have not yet been output.

3. The method according to claim 2, characterized in that The selecting, according to the first to nth natural language inputs and the association relationship between different second questions in the question bank, the nth first question output in the conversation panel from the second questions that have not been output, comprises: Determine, according to the first to nth natural language inputs, a second parameter obtained in the generation parameters; According to the second parameter, select part or all of the second questions that have not yet been output from the question bank to construct a set of candidate questions; The nth first question is selected according to the relevance of each second question in the candidate question set.

4. The method according to claim 1, characterized in that The nth first question is determined according to the first to nth natural language inputs; wherein n is a positive integer; and the nth first question includes: Based on the first to nth natural languages ​​and the generation parameters to be completed, the nth first question is generated.

5. The method according to any one of claims 1 to 4, characterized in that: The method further comprises: Before determining the nth first question according to the first to nth natural language inputs, determining whether a preset event occurs; In response to detecting a preset event, the selection and output of the nth first question is stopped to change the way of obtaining the missing generation parameter.

6. The method according to claim 5, characterized in that After stopping the selection and output of the nth first question, the method further includes: Generating prompt words from at least part of the first to nth natural language inputs; The generation parameters are completed by a first model based on the prompt word.

7. The method according to claim 5, characterized in that The preset event includes at least one of the following: whether the number of first questions output after receiving the first natural language input is equal to or greater than a specified number; Whether the duration of the multiple rounds of conversation based on the conversation panel reaches a first duration; A stop operation is detected, wherein the stop operation includes: an operation acting on a stop questioning control in the conversation panel; When the duration of multiple rounds of conversations based on the conversation panel has reached a second duration and a complete third parameter has been obtained, wherein the second duration is less than the first duration; the generated parameters include a third parameter and a fourth parameter, wherein the importance of the third parameter is stronger than that of the fourth parameter.

8. The method according to any one of claims 1 to 7, characterized in that: The method further comprises: When the m+1th natural language input for the mth first question does not answer the mth first question, the second question with the same meaning but different description as the mth second question is determined as the m+1th first question; wherein the mth first question is the second question for obtaining the third parameter; and m is a positive integer.

9. The method according to claim 6, characterized in that The step of inputting the first to nth natural languages ​​to generate prompt words includes: obtaining account information of an account associated with the target file; The prompt word is generated by inputting the first to nth natural languages ​​and the account information.

10. The method according to claim 6, characterized in that After stopping the selection and output of the nth first question, the method further includes: Outputting recommended parameters of the missing generation parameters; The generation parameters are completed according to the feedback operation acting on the generation parameters.

11. The method according to any one of claims 1 to 10, characterized in that: In response to the generation parameters being complete, generating the target file according to the generation parameters includes: According to the generation parameters, write a document outline; Generate the target file according to the file outline.

12. The method according to claim 11, characterized in that The method further comprises: In the session panel, displaying the document outline; Based on a portion or all of the modification operations on the document outline, the document outline is updated.

13. The method according to any one of claims 1 to 12, characterized in that: The generation parameters include at least one of the following: The subject of the document; File length; Document audience; Application scenarios of files; File style.

14. A file generating device, characterized in that: The device comprises: A first determination module is configured to, in response to the conversation panel receiving a first natural language input indicating generation of a target file, determine an nth first question according to at least part of the first to nth natural language inputs in the case where it is determined that generation parameters are missing according to the first natural language input; wherein n is a positive integer; and the nth first question is used to obtain a first parameter that is not yet included in the generation parameters; An acquisition module, configured to acquire, based on the conversation panel, an (n+1)th natural language input based on the initial output of the (n)th first question; A second determination module, configured to determine whether the generation parameters are complete based on the (n+1)th natural language input; an output module, configured to, in response to the generation parameter being incomplete, initially output an n+1th first question based on at least a portion of the first to n+1th natural language inputs; A generation module is used to generate the target file according to the generation parameters in response to the generation parameters being complete.

15. An electronic device, characterized in that: include: processor; a memory for storing processor-executable instructions; Wherein, the processor is configured to: implement the file generation method described in any one of claims 1 to 14 when running the executable instructions.

16. A computer-readable storage medium, characterized in that: The readable storage medium stores an executable program, wherein the executable program, when executed by a processor, implements the file generation method according to any one of claims 1 to 14.

Images