Method and system for intelligently generating PowerPoint based on AI

Abstract

The invention discloses an AI-based method and system for intelligently generating a PowerPoint, and the method comprises the steps: obtaining user input data, and generating a demand text based on an AI large language model; the intention category and the situation category are analyzed to complete theme customization; extracting a multi-layer content structure from the demand text based on the intention category and the context category; generating a dynamic browsing template according to the theme customization and the multi-layer content structure; and adjusting the dynamic browsing template, and automatically optimizing the multi-layer content structure based on the adjustment feedback of the user to generate the presentation. The theme customization is completed by analyzing the intention category and the situation category of the text, the recognized multi-layer content structure can automatically generate the presentation file template meeting the required text, the multi-layer content structure can be more reasonably filled into the presentation file template, and in the template generation process, a user can perform adjustment in real time, so that the user experience is improved, and the user experience is improved. And the satisfaction degree of the user on the finally generated presentation file can be greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent generation of presentations, and in particular to an AI-based intelligent generation method and system for presentations. Background Art

[0002] With the rapid development of AI technology, more and more AI-based tools are being rapidly introduced to the market, including presentation generation tools. However, existing tools are extremely dependent on user input text when generating presentations. AI tools can only automatically fill text into preset templates, and such tools lack independent understanding and data processing of user input text, resulting in the generated presentation either not conforming to the meaning of the user's input text, or the template selection of the presentation cannot fit the real scene and intention of the input text. Even after the presentation is generated, users are required to make frequent adjustments. In other words, the existing AI generation tools for presentations cannot meet the user's intelligent generation needs. Summary of the invention

[0003] In view of this, the present invention proposes an AI-based intelligent presentation generation method and system, which can generate a required text based on the user's input content, and automatically generate a presentation template that fits the required text content, so as to intelligently output a presentation that meets the user's requirements. The present invention provides the following technical solutions: An AI-based intelligent presentation generation method, the method comprising: obtaining user input data, generating a demand text based on an AI large language model; parsing intent categories and context categories based on the demand text to complete theme customization; extracting a multi-layer content structure from the demand text based on the intent categories and context categories; generating a dynamic browsing template based on theme customization and the multi-layer content structure; adjusting the dynamic browsing template, and automatically optimizing the multi-layer content structure based on user adjustment feedback to generate a presentation.

[0004] Optionally, a method for parsing intent categories and context categories based on a demand text to complete topic customization includes: semantically parsing and segmenting the demand text through a natural language processing model; determining the intent categories of the segmented data through an intent classification model; quantifying the intent categories corresponding to the segmented paragraphs into priority scores and sorting them; assigning a presentation style to each segmented paragraph based on the intent category and priority score sorting, and generating a content structure framework corresponding to the presentation style; determining the context category of the segmented data through a semantic context classification model; and generating a customized topic based on the intent category and context category data.

[0005] Optionally, the method for extracting a multi-layer content structure from a demand text based on intent categories and context categories includes: calculating and sorting the topic intent scores corresponding to the segmented paragraphs of each demand text, defining the segmented paragraph corresponding to the highest score as a first-level topic to construct a topic layer; obtaining relevant paragraphs of the segmented paragraph corresponding to the first-level topic, and using a semantic clustering algorithm to form sub-topics to construct a sub-topic layer; based on the sub-topic content, reducing it to key points to complete the key point layer extraction.

[0006] Optionally, the calculation formula for quantifying the intent category into a priority score is as follows: ; in, Represents the content of the i-th segmented paragraph, Indicates paragraph The probability value of being classified in the jth category of intent, Represents the weight of the j-th intent.

[0007] Optionally, the method of determining the context category of the segmented data through a semantic context classification model includes: vectorizing each segmented paragraph text through a feature extraction model, and calculating the similarity score between it and a predefined context, so as to determine the closest context category according to the score, and the calculation formula of the similarity score is: ; in, Represents the entire text, Represents text with predefined context categories The similarity of Represents the weight of the situation category.

[0008] Optionally, the calculation formula for the topic intention score is: ; in, Indicates the current segment The relevance score to the complete text T, Indicates the current segment The independent importance of , α, β are weight coefficients.

[0009] Optionally, the method of obtaining relevant paragraphs of the segmented paragraphs corresponding to the primary topic and forming sub-topics using a semantic clustering algorithm to construct a sub-topic layer includes: converting the segmented paragraphs corresponding to the primary topic into a feature vector V T ; Convert other segmented segments into feature vectors V si ; Calculate each eigenvector V si With the eigenvector V T The similarity between them is calculated as follows: ; A threshold θ is set, and the segmented paragraphs with a corresponding similarity greater than θ are retained as related paragraphs; a clustering algorithm is selected to cluster the related paragraphs, and multiple sub-topics are generated to complete the construction of the sub-topic layer.

[0010] The present invention further discloses an AI-based intelligent presentation generation system, comprising: an AI model module, used to obtain user input data and generate a demand text based on an AI large language model; a theme customization module, used to parse intent categories and context categories based on the demand text to complete theme customization; a structure recognition module, used to extract a multi-layer content structure from the demand text based on intent categories and context categories; a dynamic template generation module, used to generate a dynamic browsing template based on theme customization and the multi-layer content structure; a presentation generation module, used to adjust the dynamic browsing template and automatically optimize the multi-layer content structure based on user adjustment feedback to generate a presentation.

[0011] The present invention further discloses a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the method for intelligently generating presentations based on AI is implemented.

[0012] The present invention further discloses an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned AI-based intelligent presentation generation method when executing the program.

[0013] According to the technical solution of the present invention, the theme customization is completed by parsing the intention category and context category of the demand text, and the identified multi-layer content structure can automatically generate a presentation template that meets the demand text, and the multi-layer content structure can be more reasonably filled in the presentation template, and during the template generation process, the user can make real-time adjustments, which can greatly improve the user's satisfaction with the final generated presentation and improve the presentation generation efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] For the purpose of illustration and not limitation, the present invention is now described in conjunction with the embodiments of the present invention and the accompanying drawings, in which: Figure 1 It is a flowchart of an AI-based intelligent presentation generation method in an embodiment of the present invention; Figure 2 It is a structural diagram of an AI-based presentation intelligent generation system in an embodiment of the present invention; Figure 3 It is a schematic diagram of the structure of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION

[0015] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the implementation mode of the present application will be clearly and completely described below in conjunction with the drawings in the implementation mode of the present application. Obviously, the described implementation mode is only a part of the implementation mode of the present application, not all the implementation modes. Based on the implementation mode in the present application, all other implementation modes obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present application.

[0016] It should be noted that the terms "first", "second", etc. in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application described here. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] In the present application, the terms "upper", "lower", "left", "right", "front", "back", "top", "bottom", "inner", "outer", "middle", "vertical", "horizontal", "lateral", "longitudinal" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the drawings. These terms are mainly used to better describe the present application and its implementation methods, and are not used to limit the indicated devices, elements or components to have a specific orientation, or to be constructed and operated in a specific orientation.

[0018] In addition, some of the above terms may be used to express other meanings in addition to indicating orientation or positional relationship. For example, the term "on" may also be used to express a certain dependency or connection relationship in some cases. For those of ordinary skill in the art, the specific meanings of these terms in this application can be understood according to the specific circumstances. In addition, the term "plurality" should mean two or more than two.

[0019] It should be noted that, in the absence of conflict, the embodiments of the present application and the features in the embodiments can be combined with each other. The embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0020] refer to Figure 1 This embodiment discloses an AI-based intelligent presentation generation method, wherein the method specifically includes: S100: Obtain user input data and generate demand text based on the AI ​​big language model.

[0021] The user input data is a summary sentence summarized by the user or the only sentence that can express the main idea of ​​the target presentation. After the user input is completed, the AI ​​big language model is called to expand the user input data and generate the demand text. The demand text is a text expanded around the user input data, and its main idea is the main idea of ​​the user input data. In this embodiment, there is no restriction on the AI ​​big language model, and the user can choose the AI ​​big language model suitable for his own field as needed to improve the matching degree between the generated demand text and the user input data.

[0022] S200: Analyze the intent category and context category based on the demand text to complete the theme customization.

[0023] By analyzing the demand text, the main intention of the text can be identified, such as argumentation, analysis or summary. Similarly, the context of the text can be identified, such as business scenarios, educational scenarios or technology demonstration scenarios. Based on this, the theme that meets the intention and context can be customized. The theme here is an abstract concept, which is optimized in the structure of the demand text and will also serve as the basis for the subsequent content screening, layout design, etc. Specific methods include: S210: semantically parse and segment the demand text through a natural language processing model. That is, semantically segment the demand text through a natural language processing model. The semantic segmentation in this embodiment is based on syntactic analysis, such as punctuation marks, syntactic structures, etc., and the sentence segmentation model in the natural language processing tool library can also be used to segment the demand text.

[0024] S220: Determine the intent category of the segmented data through the intent classification model. After segmentation, the demand text is divided into multiple segmented paragraphs. In this embodiment, the intent classification model uses the Transformer architecture, such as BERT, RoBERTa, etc., and corresponds the intent category structure to the segmented paragraphs.

[0025] S230: quantify the intent categories corresponding to the segmented paragraphs into priority scores and sort them. Specifically, the calculation formula of the priority score is: ; in, Represents the content of the i-th segmented paragraph, Indicates paragraph The probability value of being classified in the jth category of intent, Indicates the weight of the j-th type of intent. The weight is set according to the scenario of the requirement text. For example, explanatory intent and suggestion intent can be displayed first.

[0026] S240: Assign a presentation style to each segmented paragraph according to the intent category and priority score, and generate a content structure framework corresponding to the presentation style, such as by highlighting the segmented paragraphs that are displayed first.

[0027] S250: Determine the context category of the segmented data through a semantic context classification model. Specifically, each segmented paragraph text is vectorized through a feature extraction model, and a similarity score between it and a predefined context is calculated to determine the closest context category based on the score. The calculation formula for the similarity score is: ; in, Represents the entire text, Represents text with predefined context categories The similarity of Represents the weight of the context category. After calculating the maximum similarity score, the corresponding context category This is the context category that is closest to the required text.

[0028] S260: Generate a customized theme based on the intent category and context category data. After determining the intent category and context category, generate the main elements of the theme design. Including colors, fonts, icons and backgrounds, etc. For example, color selection: select the main color and auxiliary color from the color matching database according to the intent category and context category. For example, business meetings may prefer stable colors such as blue or gray, while education and training may choose softer colors. A color emotion mapping algorithm can be used to convert intent and context into corresponding color combinations. Font selection: Different intents and contexts will affect the choice of fonts. For example, a more authoritative serif font may be selected for formal reports, while a simpler sans serif font may be selected for teaching and training to ensure reading clarity. Icons and shape elements: Based on the intent category and context category, the system can select suitable icon styles and shape elements from the icon library. For example, data icons can be used in scientific research presentations, while more dynamic visual elements are more suitable for marketing promotion. Background and pattern: Context categories sometimes require specific background patterns or decorative elements. For example, a PPT for internal training can have a company logo watermark, while a presentation for a product launch can use a gradient background or pattern to enhance the visual effect.

[0029] S300: Extracting a multi-layer content structure from the demand text based on the intent category and the context category. Specifically, this embodiment takes the three-layer content structure as an example, calculates and sorts the topic intent scores corresponding to the segmented paragraphs of each demand text, defines the segmented paragraph corresponding to the highest score as the first-level topic, and constructs the topic layer; obtains the relevant paragraphs of the segmented paragraph corresponding to the first-level topic, and uses the semantic clustering algorithm to form sub-topics to construct the sub-topic layer; based on the sub-topic content, reduces it to key points to complete the key point layer extraction. Among them, the calculation formula for the topic intent score is: ; in, Indicates the current segment The relevance score to the complete text T, Indicates the current segment The independent importance of , α, β are weight coefficients. This embodiment provides an exemplary calculation method: the current segmentation paragraph The complete text T is converted into a vector representation and , calculated using cosine similarity and Similarity: The result is a value between -1 and 1, where 1 means completely similar and -1 means completely opposite. The calculation method of the independent importance is as follows: extract a group of keywords that can reflect the main idea of ​​the text from the entire text T. For example, TF-IDF, TextRank or LDA methods can be used to identify the words that best represent the main idea of ​​the text, and then calculate the paragraph The density of the keywords contained in the paragraph is then used as the result of the density. The independent importance of the paragraph is the ratio of the number of keywords to the paragraph length. The calculation method of this embodiment is only exemplary. For example, the position importance, centrality graph algorithm, topic contribution and other values ​​can also be used as independent importance.

[0030] The method for constructing the sub-topic layer includes: converting the segmented paragraphs corresponding to the first-level topic into feature vectors V T ; Convert other segmented segments into feature vectors V si ; Calculate each eigenvector V si With the eigenvector V T The similarity between them is calculated as follows: ; Set a threshold θ and retain the segmented paragraphs with a corresponding similarity greater than θ as related paragraphs; A clustering algorithm is used to cluster related paragraphs and generate multiple subtopics to complete the construction of the subtopic layer. For the extraction of the key point layer, the key point extraction formula is: Point( ) =γ·Rel( , )+δ·Imp( ),in, To split paragraphs, For the subtopic, Rel ( , )for and The correlation of Imp( )for The importance of is, γ and δ are adjustment weights.

[0031] S400: Generate a dynamic browsing template based on theme customization and multi-layer content structure.

[0032] During the presentation generation process, after the intent categories and context categories are parsed, a basic multi-level content structure and layout is generated based on the parsing results, and the user can further customize it on this basis. When the user adjusts the content, the theme and layout are automatically updated, and the effect of the user's editing is fed back in real time through the page. Since the themes generated by intent and context constitute a structural framework, whenever the user modifies or supplements new text content, only the relevant paragraphs or modules are updated without redrawing the entire template.

[0033] S500: Adjust the dynamic browsing template, and automatically optimize the multi-layer content structure based on the user's adjustment feedback to generate a presentation. The automatic optimization method includes: Capturing adjustment behavior: Recording the user's operation type on each content module, such as text adjustment, paragraph position movement, picture addition and deletion, chart modification, etc. Operation frequency statistics: Counting the user's operation frequency in different modules, such as the user repeatedly adjusting the size and color of the title module, which may indicate that the user has a specific preference for the module. Inference of operation intention: By analyzing the user's adjustment behavior, the system infers the user's preference. For example, if the user repeatedly increases the title font, it may indicate that the user wants the title to be more eye-catching. In this embodiment, the weight values ​​of different contents are dynamically adjusted according to the user's adjustment behavior, so that when the presentation is finally generated, the content layout and focus are more in line with the user's preferences, and the content blocks are sorted according to the weight value, and the high-weight content is preferentially displayed in a more prominent position, and the low-weight content is reduced or placed in a secondary position. Further exemplary, if the user merges or simplifies certain content paragraphs multiple times, the system automatically divides these contents into the same sub-topic or merges them into a primary topic. For example, if the user merges "income" and "expenditure" into "financial overview", the system will automatically merge this part of the content in subsequent generation. If the user edits certain paragraphs independently multiple times, the system will refine them into independent sub-topics. For example, if the user adjusts "sales revenue" and "non-sales revenue" separately, the system will automatically divide them into two sub-topics in subsequent optimization. After adjusting the dynamic browsing template, adaptive optimization is used to make the visual effect more in line with user expectations. The above completes the generation of the presentation.

[0034] refer to Figure 2 This embodiment further discloses an AI-based presentation intelligent generation system, including: AI model module 21, used to obtain user input data and generate demand text based on AI large language model; The theme customization module 22 is used to parse the intent category and context category based on the demand text to complete the theme customization; it is also used to semantically parse and segment the demand text through a natural language processing model; determine the intent category of the segmented data through an intent classification model; quantify the intent category corresponding to the segmented paragraphs into priority scores and sort them; assign a presentation style to each segmented paragraph based on the intent category and priority score sorting, and generate a content structure framework corresponding to the presentation style; determine the context category of the segmented data through a semantic context classification model; and generate a customized theme based on the intent category and context category data.

[0035] The structure recognition module 23 is used to extract the multi-layer content structure from the demand text based on the intention category and the context category; it is also used to calculate and sort the subject intention scores corresponding to the segmented paragraphs of each demand text, define the segmented paragraph corresponding to the highest score as the first-level topic, so as to construct the topic layer; obtain the relevant paragraphs of the segmented paragraph corresponding to the first-level topic, and use the semantic clustering algorithm to form sub-topics to construct the sub-topic layer; specifically: convert the segmented paragraph corresponding to the first-level topic into a feature vector V T ; Convert other segmented segments into feature vectors V si ; Calculate each eigenvector V si With the eigenvector V T The similarity between them is calculated as follows: ; Set the threshold θ and retain the segmented paragraphs with the corresponding similarity greater than θ as related paragraphs; select a clustering algorithm to cluster the related paragraphs and generate multiple sub-topics to complete the construction of the sub-topic layer; based on the sub-topic content, reduce it to key points to complete the key point layer extraction.

[0036] A dynamic template generation module 24, used to generate a dynamic browsing template according to theme customization and multi-layer content structure; The presentation generation module 25 is used to adjust the dynamic browsing template and automatically optimize the multi-layer content structure based on the user's adjustment feedback to generate a presentation.

[0037] Figure 3 A schematic diagram of the physical structure of an electronic device provided in an embodiment of the present invention, such as Figure 3 As shown, the electronic device 50 includes: a processor 501 (processor), a memory 502 (memory) and a bus 503; The processor 501 and the memory 502 communicate with each other via the bus 503 ; the processor 501 is used to call program instructions in the memory 502 to execute the methods provided by the above-mentioned method implementation methods.

[0038] This embodiment provides a non-transitory computer-readable storage medium, which stores computer instructions. The computer instructions enable a computer to execute the methods provided by the above-mentioned method embodiments.

[0039] A person skilled in the art can understand that all or part of the steps for implementing the above-mentioned method implementation method can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium, which, when executed, executes the steps of the above-mentioned method implementation method; and the aforementioned storage medium includes: ROM, RAM, magnetic disk or optical disk, etc., various storage media that can store program codes.

[0040] The device implementation described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the present implementation scheme. Those of ordinary skill in the art may understand and implement it without creative effort.

[0041] Through the description of the above implementation modes, those skilled in the art can clearly understand that each implementation mode can be implemented by means of software plus a necessary general hardware platform, or of course by hardware. Based on such an understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each implementation mode or some parts of the implementation mode.

[0042] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions may occur depending on design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. An AI-based presentation intelligent generation method, characterized in that: The method comprises: Obtain user input data and generate demand text based on the AI ​​big language model; Parse the intent category and context category based on the demand text to complete the theme customization; Extracting multi-layer content structure from requirement text based on intent categories and context categories; Generate dynamic browsing templates based on theme customization and multi-layer content structure; Adjust dynamic browsing templates and automatically optimize multi-layer content structures based on user adjustment feedback to generate presentations.

2. The AI-based presentation intelligent generation method according to claim 1, characterized in that: Methods for parsing intent categories and context categories based on demand text to complete topic customization include: Semantically parse and segment the demand text through the natural language processing model; Determine the intent category of the segmented data through the intent classification model; Quantify the intent categories corresponding to the segmented paragraphs into priority scores and sort them; According to the intent category and priority score sorting, a presentation style is assigned to each segmented paragraph, and a content structure framework corresponding to the presentation style is generated; Determine the context category of the segmented data through a semantic context classification model; Generate customized topics based on intent category and context category data.

3. The AI-based presentation intelligent generation method according to claim 2, characterized in that: The method for extracting a multi-layer content structure from a requirement text based on an intention category and a context category comprises: Calculate and sort the subject intention scores corresponding to the segmented paragraphs of each requirement text, and define the segmented paragraphs corresponding to the highest score as the first-level topic to construct the topic layer; Obtain relevant paragraphs of the segmented paragraphs corresponding to the first-level topic, and use semantic clustering algorithm to form sub-topics to construct a sub-topic layer; Based on the subtopic content, reduce it to key points to complete the key point level extraction.

4. The AI-based presentation intelligent generation method according to claim 2, characterized in that: The calculation formula for quantifying the intent category corresponding to the segmented paragraph into a priority score is as follows: ； in, Represents the content of the i-th segmented paragraph, Indicates paragraph The probability value of being classified in the jth category of intent, Represents the weight of the j-th intent.

5. The AI-based presentation intelligent generation method according to claim 2, characterized in that: The method for determining the context category of the segmented data through the semantic context classification model includes: vectorizing each segmented paragraph text through the feature extraction model, and calculating the similarity score between it and the predefined context, so as to determine the closest context category according to the score, and the calculation formula of the similarity score is: ； in, Represents the entire text, Represents text with predefined context categories The similarity of Represents the weight of the situation category.

6. The AI-based presentation intelligent generation method according to claim 3, characterized in that: The calculation formula for the topic intention score is: ； in, Indicates the current segment The relevance score to the complete text T, Indicates the current segment The independent importance of , α, β are weight coefficients.

7. The AI-based presentation intelligent generation method according to claim 3 is characterized in that: The method of obtaining the relevant paragraphs of the segmented paragraphs corresponding to the primary topic and forming subtopics using a semantic clustering algorithm to construct a subtopic layer includes: Convert the segmented paragraphs corresponding to the first-level topic into feature vectors V T ; Convert other segmented segments into feature vectors V si ; Calculate each eigenvector V si With the eigenvector V T The similarity between them is calculated as follows: ； Set a threshold θ and retain the segmented paragraphs with a corresponding similarity greater than θ as related paragraphs; A clustering algorithm is used to cluster related paragraphs and generate multiple sub-topics to complete the construction of the sub-topic layer.

8. An AI-based presentation intelligent generation system, characterized in that: include: AI model module, used to obtain user input data and generate demand text based on AI large language model; Theme customization module is used to parse the intent category and context category based on the demand text to complete theme customization; A structure recognition module is used to extract multi-layer content structures from requirement texts based on intent categories and context categories; Dynamic template generation module, used to generate dynamic browsing templates based on theme customization and multi-layer content structure; The presentation generation module is used to adjust the dynamic browsing template and automatically optimize the multi-layer content structure based on the user's adjustment feedback to generate a presentation.

9. A computer-readable storage medium, characterized in that: The storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method described in any one of claims 1 to 7 is implemented.

Images