Method, system, device and storage medium for automatically updating a presentation

Abstract

The application discloses a kind of presentation automatic updating method, system, equipment and storage medium, its technical scheme main point is in, response presentation update request;First text paragraph of presentation is extracted, after each first text paragraph is structured and vectorization processing, obtain first paragraph vector;Second text paragraph of the update information of extraction is extracted, after each second text paragraph is structured and vectorization processing, obtain second paragraph vector;The semantic similarity between first paragraph vector and second paragraph vector is calculated, and similarity matrix is obtained;Diffusion matrix is constructed according to similarity matrix, the propagation of update information between first paragraph vector is simulated by diffusion matrix, and updated first paragraph vector is obtained, and presentation is updated according to updated first paragraph vector.The application realizes automatic updating presentation, ensures that presentation content is optimized according to update content, and improves the efficiency and quality of making presentation.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of automated office, and particularly relates to a presentation automatic updating method, system, device and storage medium. BACKGROUND

[0002] As one of the report carriers for daily meetings and summaries, presentations (PowerPoint, PPT) are widely used in daily work.

[0003] In today's information explosion era, information is updated every day. Due to the continuous updating of information, the presentation also needs to be updated continuously, especially in the field of teaching, the teaching presentation of teachers usually needs to follow the latest information. However, it is very tedious and inefficient to manually update the presentation according to the latest information. SUMMARY

[0004] The purpose of the present application is to provide a presentation automatic updating method, system, device and storage medium, which can realize automatic updating of the presentation and improve the efficiency and quality of making the presentation.

[0005] The first aspect of the present application provides a presentation automatic updating method, comprising:

[0006] responding to a request for updating the presentation;

[0007] extracting all first text paragraphs of the presentation, and obtaining first paragraph vectors after structuring and vectorizing each first text paragraph;

[0008] extracting all second text paragraphs of the updated information, and obtaining second paragraph vectors after structuring and vectorizing each second text paragraph;

[0009] calculating the semantic similarity between all first paragraph vectors and all second paragraph vectors to obtain a similarity matrix;

[0010] constructing a diffusion matrix according to the similarity matrix, simulating the propagation of the updated information between the first paragraph vectors through the diffusion matrix to obtain updated first paragraph vectors, and updating the presentation according to all updated first paragraph vectors.

[0011] In some embodiments, the calculation of the semantic similarity between all first paragraph vectors and all second paragraph vectors comprises:

[0012] calculating the semantic similarity between all first paragraph vectors and all second paragraph vectors according to a similarity calculation formula, wherein the similarity calculation formula is:

[0013] E ij =(Vi *V j ) / (||V i ||*||V j ||)

[0014] wherein, the V i represents the i-th first passage vector, V j represents the j-th second passage vector, E ij represents the semantic similarity between the i-th first passage vector and the j-th second passage vector.

[0015] In some embodiments, the constructing a diffusion matrix according to the similarity matrix comprises:

[0016] calculating the propagation proportion of all second passage vectors on all first passage vectors according to a propagation proportion calculation formula, the propagation proportion calculation formula being:

[0017]

[0018] wherein, the E ij represents the semantic similarity between the i-th first passage vector and the j-th second passage vector, D ij represents the propagation proportion of the i-th first passage vector on the j-th second passage vector.

[0019] The propagation proportions of all second passage vectors on all first passage vectors constitute a diffusion matrix.

[0020] In some embodiments, the simulating the propagation of updated information between first passage vectors through the diffusion matrix to obtain updated first passage vectors comprises:

[0021] In the propagation process, the first passage vectors are updated according to the diffusion matrix, and after multiple rounds of propagation, the change amplitude between the first passage vectors after each round of propagation and the first passage vectors of the previous round is calculated, until the change amplitude of the first passage vectors is lower than a preset threshold, the propagation is stopped, and the first passage vectors obtained after the last round of propagation are taken as the updated first passage vectors.

[0022] In some embodiments, the updating the first passage vectors according to the diffusion matrix in the propagation process comprises:

[0023] In the propagation process, the first passage vectors are updated according to a passage update formula, the passage update formula being:

[0024]

[0025] wherein, the D represents the diffusion matrix, represents the set of first passage vectors of the k-th round, a set of first paragraph vectors representing the first paragraph vectors of the k+1th round.

[0026] In some embodiments, the calculating the change amplitude between the first paragraph vectors after each round of propagation and the first paragraph vectors of the previous round comprises:

[0027] The change amplitude between the first paragraph vectors after each round of propagation and the first paragraph vectors of the previous round is calculated according to a change amplitude calculation formula, and the change amplitude calculation formula is:

[0028]

[0029] The A represents the change amplitude between the first paragraph vectors after each round of propagation and the first paragraph vectors of the previous round.

[0030] In some embodiments, the extracting all first text paragraphs of the presentation, and after the structural processing and vectorization processing of each first text paragraph, obtaining the first paragraph vector, comprises:

[0031] According to the page order of the presentation, all first text paragraphs of each page are extracted in turn to obtain a first paragraph set of each page.

[0032] After the structural processing and vectorization processing of the first paragraph set of each page, a first vector set of each page is obtained.

[0033] The second aspect of the present application provides a presentation automatic updating system, comprising:

[0034] A data acquisition module is configured to respond to an updating request of a presentation.

[0035] A first processing module is configured to extract all first text paragraphs of the presentation, and after the structural processing and vectorization processing of each first text paragraph, obtain first paragraph vectors.

[0036] A second processing module is configured to extract all second text paragraphs of the updated information, and after the structural processing and vectorization processing of each second text paragraph, obtain second paragraph vectors.

[0037] A similarity calculation module is configured to calculate the semantic similarity between all first paragraph vectors and all second paragraph vectors to obtain a similarity matrix.

[0038] An updated information propagation module is configured to construct a diffusion matrix according to the similarity matrix, simulate the propagation of the updated information between the first paragraph vectors through the diffusion matrix, obtain updated first paragraph vectors, and update the presentation according to all updated first paragraph vectors.

[0039] A third aspect of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described above.

[0040] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described above.

[0041] The technical solution provided by this invention has the following advantages and effects: by calculating the semantic similarity between the presentation and the update information, a similarity matrix is ​​obtained; by using the similarity matrix and multi-round information propagation processing, an update report is obtained, thereby realizing automatic updating of the presentation, ensuring that the presentation content is optimized according to the update content, and improving the efficiency and quality of presentation production. Attached Figure Description

[0042] Figure 1 This is a flowchart illustrating the automatic presentation update method provided by the present invention;

[0043] Figure 2 This is a structural block diagram of the presentation automatic update system provided by the present invention;

[0044] Figure 3 This is an internal structural diagram of a computer device provided in an embodiment of the present invention. Detailed Implementation

[0045] To facilitate understanding of the present invention, specific embodiments of the present invention will be described in more detail below with reference to the accompanying drawings.

[0046] Unless otherwise specified or defined, the terms "first," "second," etc., used in this document are for distinguishing names only and do not represent a specific number or order.

[0047] Unless otherwise stated or defined, the term “and / or” as used herein includes any and all combinations of one or more of the associated listed items.

[0048] It should be noted that in this article, "fixed to" or "connected to" can mean directly fixed to or connected to a component, or indirectly fixed to or connected to a component.

[0049] like Figure 1 As shown, this embodiment provides a method for automatically updating presentations, including the following steps S1 to S5:

[0050] Step S1: Respond to the presentation update request.

[0051] In practical applications, the presentation can be a locally stored presentation or a presentation in a remote server, and the user uploads or specifies the presentation and then issues an update request. The update information can be obtained by crawling from the specified information material or industry news. The update information can be pre-crawled or crawled during the process of updating the presentation. The specified information material can be an academic database, an electronic journal, an online library, data and announcements published by governments or institutions, etc. A crawler tool can be used to crawl the update information from the specified information material or industry news.

[0052] Step S2, extracting all first text paragraphs of the presentation, and obtaining a first paragraph vector after structuring and vectorizing each first text paragraph.

[0053] In practical applications, the title, paragraphs, charts, and note content in the presentation are identified, and then the paragraph content is extracted. Each piece of text in the paragraph content is identified as a first text paragraph. The first text paragraph is then structured. After extracting key information, annotating semantic roles, and establishing data relationships, structured text data is obtained. The structured text data corresponding to the first text paragraph is then vectorized. In this application, a locally pre-trained Sentence-BERT model is used for reasoning to obtain a vectorized representation of the structured text data, i.e., a first paragraph vector. The Sentence-BERT model generates fixed-length text paragraph embedding vectors, making the subsequent semantic similarity calculation between text paragraphs more efficient.

[0054] Further, the extracting all first text paragraphs of the presentation, and obtaining a first paragraph vector after structuring and vectorizing each first text paragraph, comprises:

[0055] According to the page order of the presentation, all first text paragraphs of each page are extracted in turn to obtain a first paragraph set of each page.

[0056] After structuring and vectorizing the first paragraph set of each page, a first vector set of each page is obtained.

[0057] In practical applications, by sequentially extracting all first text paragraphs of each page of the presentation, the first paragraph set of all pages of the presentation can be obtained, and then the first paragraph set is sequentially structured and vectorized according to the page number order to obtain the first vector set of all pages, facilitating subsequent semantic similarity calculation of the first vector set of each page and the paragraph vector corresponding to the update information, obtaining the corresponding similarity matrix of each page, that is, processing the presentation page by page, which can decompose the content of the presentation into more detailed units, each page usually revolves around a core theme, facilitating accurate content analysis, semantic extraction and structured processing. Processing page by page can also avoid processing a large amount of content at one time, reducing memory usage and computing resource consumption. For large presentations, processing page by page can significantly improve processing efficiency.

[0058] Step S3, extracting all second text paragraphs of the captured update information, and after structured processing and vectorization of each second text paragraph, obtaining a second paragraph vector.

[0059] In practical applications, the second text paragraph is extracted from the update information using the same method as in the presentation, and the structured processing and vectorization of the second text paragraph are performed using the same method as the first text paragraph. Specifically, the second text paragraph is first structured to obtain the corresponding structured data of the second text paragraph, and then the structured data of the second text paragraph is vectorized to obtain the second paragraph vector.

[0060] Step S4, calculating the semantic similarity between all first paragraph vectors and all second paragraph vectors to obtain a similarity matrix.

[0061] In practical applications, by calculating the semantic similarity between the first paragraph vector and the second paragraph vector, a quantitative basis for the association between the presentation and the update information can be provided, thereby supporting the construction of an intelligent recommendation system and a knowledge graph, and in this application, providing an update basis for updating the presentation. All second paragraph vectors form a second vector set. In the case of processing the presentation page by page, the first vector set of each page is obtained, the semantic similarity between all first paragraph vectors in the first vector set of each page and all second paragraph vectors in the second vector set is calculated, and thus the similarity matrix of each page is obtained, facilitating subsequent page-by-page processing of the presentation according to the similarity matrix of each page.

[0062] Further, the calculation of the semantic similarity between all first paragraph vectors and all second paragraph vectors comprises:

[0063] According to the similarity calculation formula, the semantic similarity between all first paragraph vectors and all second paragraph vectors is calculated, and the similarity calculation formula is:

[0064] E ij =(V i *V j ) / (||V i ||*||V j ||)

[0065] wherein, the V i represents the i-th first passage vector, V j represents the j-th second passage vector, E ij represents the semantic similarity between the i-th first passage vector and the j-th second passage vector.

[0066] In practical applications, i = 1, 2, …, n, j = 1, 2, …, m, n represents the number of first passage vectors in the first vector set, and m represents the number of second passage vectors in the second vector set. In this application, the cosine similarity between the first passage vector and the second passage vector is calculated to obtain the semantic similarity between the first passage vector and the second passage vector. The cosine similarity focuses on the directionality between the vectors, reducing the influence of vector length on similarity calculation. By using the vectors generated by Sentence-BERT, the semantic information of the text can be preserved, so that the cosine similarity can more accurately measure the semantic similarity between the first passage vector and the second passage vector.

[0067] Step S5, constructing a diffusion matrix according to the similarity matrix, simulating the propagation of updated information between the first passage vectors through the diffusion matrix, obtaining updated first passage vectors, and updating the presentation according to all updated first passage vectors.

[0068] In practical applications, the similarity matrix reflects the similarity between the passages, and the diffusion matrix simulates the propagation process of information or features between the passages on this basis. By continuously updating the representation of the passage, it integrates the information of other similar passages, thereby realizing the updating and optimization of the passage. Not limited by the format and content of the passage, it can be applied to various types and fields of text content. Based on the similarity matrix and the diffusion process, the reasons and process of passage updating can be clearly explained, increasing the credibility of the results. It can automatically update according to the similarity between the passages without human intervention, greatly improving the efficiency of updating.

[0069] Further, the constructing a diffusion matrix according to the similarity matrix comprises:

[0070] calculating the propagation proportion of all second passage vectors on all first passage vectors according to a propagation proportion calculation formula, the propagation proportion calculation formula being:

[0071]

[0072] wherein the E ij represents the semantic similarity between the i-th first passage vector and the j-th second passage vector, D ij represents the propagation ratio of the i-th first passage vector on the j-th second passage vector.

[0073] The propagation ratios of all second passage vectors on all first passage vectors constitute a diffusion matrix.

[0074] In practical applications, each row in the diffusion matrix represents the propagation ratio of information from one passage vector to other passage vectors. The first passage vectors are updated according to the propagation ratio, and the presentation is updated according to the updated first passage vectors. The update content of the updated presentation includes content modification, content completion, and format optimization, and the update content is directly embedded into the modified points of the presentation in the form of color annotation to generate an update report. Specifically, by comparing the updated first passage vectors with the first passage vectors before updating, the modified points are screened out, and according to the similarity matrix, the second passage vector with the highest similarity to the modified points is screened out from the second vector set. The second text passage corresponding to the second passage vector is embedded into the corresponding modified point in the form of color annotation.

[0075] For example, the content of a page of the presentation is:

[0076] The optimization algorithms of deep learning include gradient descent, Adam optimizer, etc. With the improvement of computing power, the training of deep neural networks becomes more and more efficient.

[0077] The latest information is:

[0078] Recent research shows that the AdaBelief optimizer has faster convergence speed than the Adam optimizer, especially when dealing with sparse data.

[0079] The final output is:

[0080] The optimization algorithms of deep learning include gradient descent, Adam optimizer, etc. Recent research shows that the AdaBelief optimizer has faster convergence speed than the Adam optimizer, especially when dealing with sparse data. With the improvement of computing power, the training of deep neural networks becomes more and more efficient.

[0081] Further, the updating information between the first passage vectors is simulated by the diffusion matrix to obtain updated first passage vectors, including:

[0082] The first paragraph vector is updated according to the diffusion matrix during the propagation process. After multiple rounds of propagation, the change amplitude between the first paragraph vector after each round of propagation and the first paragraph vector of the previous round is calculated, and the propagation is stopped when the change amplitude of the first paragraph vector is lower than a preset threshold, and the first paragraph vector obtained after the last round of propagation is taken as the updated first paragraph vector.

[0083] In practical applications, multiple rounds of propagation allow information to interact multiple times between different paragraphs or nodes, thereby more comprehensively integrating knowledge. This step-by-step propagation approach ensures that the information of each paragraph or node is fully considered and updated, avoiding information omissions that can occur with single propagation. Multiple rounds of propagation optimize the propagation path and intensity of information step by step, allowing more accurate identification and transmission of important information, and effectively improving the accuracy and reliability of information propagation. When the change amplitude of the first paragraph vector is lower than the preset threshold, the propagation is stopped, and the preset threshold is usually set to a small value (such as 1e-4, representing 0.0001).

[0084] Further, the updating of the first paragraph vector according to the diffusion matrix during the propagation process comprises:

[0085] The first paragraph vector is updated according to the paragraph update formula during the propagation process, and the paragraph update formula is:

[0086]

[0087] wherein the D represents the diffusion matrix, represents a set of first paragraph vectors of the kth round, represents a set of first paragraph vectors of the k+1th round.

[0088] In practical applications, simulating the propagation of information between paragraph vectors actually involves multiplying the current first paragraph vector by the diffusion matrix D in each round of propagation to update the first paragraph vector. The diffusion matrix D is fixed and is constructed based on the initial similarity matrix and remains unchanged throughout the propagation process. In this way, information propagates between paragraphs, ultimately making the paragraph vectors more accurate, comprehensive, and coherent after multiple rounds of propagation.

[0089] Further, the calculation of the change amplitude between the first paragraph vector after each round of propagation and the first paragraph vector of the previous round comprises:

[0090] The change amplitude between the first paragraph vector after each round of propagation and the first paragraph vector of the previous round is calculated according to the change amplitude calculation formula, and the change amplitude calculation formula is:

[0091]

[0092] Wherein, the A represents the change amplitude between the first paragraph vector after each propagation and the first paragraph vector of the previous round.

[0093] In practical application, the modulus between the first paragraph vector after propagation and the first paragraph vector of the previous round reflects the overall change amplitude instead of only the local change, which is convenient for measuring the strength and effect of information propagation, and by providing a global change index, it can help to judge whether the information propagation reaches a stable state, and when the change amplitude is lower than the preset threshold, it can be considered that the propagation process has converged, thereby stopping further propagation.

[0094] The presentation automatic updating method provided by the application obtains a similarity matrix by calculating the semantic similarity between the presentation and the updating information, and obtains an updating report by using the similarity matrix and multi-round information propagation processing, so as to realize real-time updating of the presentation, ensure that the content keeps up with the latest research trend, optimize the content of the presentation, and improve the quality of the presentation.

[0095] As shown in Figure 2 The application further provides a presentation automatic updating system, which comprises:

[0096] The data acquisition module 10 is configured to respond to an updating request of the presentation.

[0097] The first processing module 20 is configured to extract all first text paragraphs of the presentation, and obtain first paragraph vectors after performing structured processing and vectorization processing on each first text paragraph.

[0098] The second processing module 30 is configured to extract all second text paragraphs of the updating information, and obtain second paragraph vectors after performing structured processing and vectorization processing on each second text paragraph.

[0099] The similarity calculation module 40 is configured to calculate the semantic similarity between all first paragraph vectors and all second paragraph vectors, and obtain a similarity matrix.

[0100] The propagation updating module 50 is configured to construct a diffusion matrix according to the similarity matrix, simulate the propagation of the updating information between the first paragraph vectors through the diffusion matrix, obtain updated first paragraph vectors, and update the presentation according to all updated first paragraph vectors.

[0101] As an optional implementation, the similarity calculation module comprises the following units not shown in the figure:

[0102] The similarity calculation unit is configured to calculate the semantic similarity between all first paragraph vectors and all second paragraph vectors according to a similarity calculation formula, and the similarity calculation formula is as follows:

[0103] Eij = (V i *V j ) / (||V i ||*||V j ||)

[0104] wherein, the V i represents the i-th first passage vector, V j represents the j-th second passage vector, E ij represents the semantic similarity between the i-th first passage vector and the j-th second passage vector.

[0105] As an optional implementation, the propagation updating module comprises the following units not shown in the figure:

[0106] The matrix construction unit is configured to calculate the propagation proportion of all second passage vectors on all first passage vectors according to a propagation proportion calculation formula, and the propagation proportion calculation formula is:

[0107]

[0108] wherein, the E ij represents the semantic similarity between the i-th first passage vector and the j-th second passage vector, D ij represents the propagation proportion of the i-th first passage vector on the j-th second passage vector.

[0109] The propagation proportions of all second passage vectors on all first passage vectors constitute a diffusion matrix.

[0110] As an optional implementation, the propagation updating module comprises the following units not shown in the figure:

[0111] The multi-round propagation unit is configured to update the first passage vector according to the diffusion matrix during the propagation process, and after a plurality of rounds of propagation, calculate the change amplitude between the first passage vector after each round of propagation and the first passage vector of the previous round, until the change amplitude of the first passage vector is lower than a preset threshold, stop the propagation, and take the first passage vector obtained after the last round of propagation as the updated first passage vector.

[0112] As an optional implementation, the multi-round propagation unit comprises the following units not shown in the figure:

[0113] The passage updating unit is configured to update the first passage vector according to a passage updating formula during the propagation process, and the passage updating formula is:

[0114]

[0115] wherein, the D represents the diffusion matrix, represents the set of the first passage vector of the k-th round, representing a set of first paragraph vectors of the k+1th round.

[0116] As an optional implementation, the multi-round propagation unit comprises the following units not shown in the figure:

[0117] An amplitude calculation unit is configured to calculate the amplitude of change between the first paragraph vector after each round of propagation and the first paragraph vector of the previous round according to an amplitude calculation formula, which is:

[0118]

[0119] wherein, the A represents the amplitude of change between the first paragraph vector after each round of propagation and the first paragraph vector of the previous round.

[0120] As an optional implementation, after extracting all the first text paragraphs of the presentation, and performing structural processing and vectorization processing on each first text paragraph to obtain the first paragraph vector, the method comprises the following units not shown in the figure:

[0121] extracting all the first text paragraphs of each page in sequence according to the page order of the presentation to obtain a first paragraph set of each page;

[0122] performing structural processing and vectorization processing on the first paragraph set of each page to obtain a first vector set of each page.

[0123] The above-mentioned modules of the presentation automatic updating system can be realized by software, hardware and their combinations in whole or in part. The above-mentioned modules and units can be embedded in or independent of the processor in the computer device in hardware form, or stored in the memory in the computer device in software form, so as to be called and executed by the processor to perform the operations corresponding to the above-mentioned modules.

[0124] As shown in Figure 3 the embodiment of the present application discloses a computer device, which comprises a memory and a processor, and the memory stores a computer program.

[0125] wherein, the computer device can be a server, and its internal structure diagram can be as shown in Figure 3As shown. The computer device includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the computer device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to implement the presentation automatic updating method described in each of the above embodiments.

[0126] Those skilled in the art can understand that, Figure 3 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0127] The embodiment of the present application also discloses a computer readable storage medium storing a computer program, wherein the computer program causes a computer to execute the presentation automatic updating method described in each of the above embodiments.

[0128] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium used in each embodiment provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM (RDRAM) and the like.

[0129] Any technical features in the above embodiments can be combined, and for the sake of brevity, not all possible combinations are described, however, any combination of the technical features is considered to be within the scope of the present specification.

Claims

1. A method for automatic updating of a presentation, characterized by, The method comprises the following steps: in response to an update request of a presentation; extracting all first text paragraphs of the presentation, and obtaining first paragraph vectors after structuring and vectorizing each first text paragraph; extracting all second text paragraphs of the update information, and obtaining second paragraph vectors after structuring and vectorizing each second text paragraph; calculating semantic similarities between all first paragraph vectors and all second paragraph vectors to obtain a similarity matrix; constructing a diffusion matrix according to the similarity matrix, simulating the propagation of the update information between the first paragraph vectors through the diffusion matrix, obtaining updated first paragraph vectors, and updating the presentation according to all updated first paragraph vectors; the simulation of the propagation of the update information between the first paragraph vectors through the diffusion matrix to obtain the updated first paragraph vectors comprises: updating the first paragraph vectors according to the diffusion matrix during the propagation process, calculating the change amplitude between the first paragraph vectors after each round of propagation and the first paragraph vectors of the previous round after multiple rounds of propagation, and stopping the propagation when the change amplitude of the first paragraph vectors is lower than a preset threshold, and taking the first paragraph vectors obtained after the last round of propagation as the updated first paragraph vectors.

2. The method of claim 1, wherein, the calculation of the semantic similarities between all first paragraph vectors and all second paragraph vectors comprises: calculating the semantic similarities between all first paragraph vectors and all second paragraph vectors according to a similarity calculation formula, wherein the similarity calculation formula is: E ij = (V i * V j ) / (||V i || * ||V j ||) Wherein, the V i represents the i-th first paragraph vector, V j represents the j-th second paragraph vector, E ij represents the semantic similarity between the i-th first paragraph vector and the j-th second paragraph vector.

3. The method of claim 1, wherein, the construction of the diffusion matrix according to the similarity matrix comprises: calculating the propagation proportion of all second paragraph vectors on all first paragraph vectors according to a propagation proportion calculation formula, wherein the propagation proportion calculation formula is: Wherein, the E ij represents the semantic similarity between the i th first paragraph vector and the j th second paragraph vector, D ij represents the propagation ratio of the i th first paragraph vector on the j th second paragraph vector; the propagation proportion of all second paragraph vectors on all first paragraph vectors constitutes the diffusion matrix.

4. The method of claim 1, wherein, the updating of the first paragraph vectors according to the diffusion matrix during the propagation process comprises: updating the first paragraph vectors according to a paragraph update formula during the propagation process, wherein the paragraph update formula is: wherein D denotes a diffusion matrix, denotes a set of first passage vectors of the kth round, denotes a set of first passage vectors of the k+1th round.

5. The method of claim 4, wherein, the calculation of the change amplitude between the first paragraph vectors after each round of propagation and the first paragraph vectors of the previous round comprises: calculating the change amplitude between the first paragraph vectors after each round of propagation and the first paragraph vectors of the previous round according to a change amplitude calculation formula, wherein the change amplitude calculation formula is: wherein A represents the change amplitude between the first paragraph vectors after each round of propagation and the first paragraph vectors of the previous round.

6. The method of claim 1-5, wherein, the extraction of all first text paragraphs of the presentation, the structuring and vectorizing of each first text paragraph, and the obtaining of first paragraph vectors comprise: extracting all first text paragraphs of each page in sequence according to the page order of the presentation to obtain a first paragraph set of each page; obtaining a first vector set of each page after structuring and vectorizing the first paragraph set of each page.

7. A presentation automatic updating system, characterized by, The method comprises the following steps: a data acquisition module for responding to an update request of a presentation; a first processing module for extracting all first text paragraphs of the presentation, and obtaining first paragraph vectors after structuring and vectorizing each first text paragraph; The second processing module is configured to extract all second text paragraphs of the captured update information, and obtain second paragraph vectors by performing structured processing and vectorization processing on each second text paragraph. The similarity calculation module is configured to calculate semantic similarities between all first paragraph vectors and all second paragraph vectors, and obtain a similarity matrix. The propagation update module is configured to construct a diffusion matrix according to the similarity matrix, simulate propagation of the update information between the first paragraph vectors through the diffusion matrix, obtain updated first paragraph vectors, and update the presentation according to all the updated first paragraph vectors. The propagation update module is specifically configured to update the first paragraph vectors according to the diffusion matrix in the propagation process, calculate a change amplitude between the first paragraph vectors after each round of propagation and the first paragraph vectors of the previous round of propagation through multiple rounds of propagation, and stop the propagation when the change amplitude of the first paragraph vectors is lower than a preset threshold, and take the first paragraph vectors obtained after the last round of propagation as the updated first paragraph vectors.

8. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to implement the steps of the method in any one of claims 1-6.

9. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to implement the steps of the method in any one of claims 1-6.

Images