Article generation method and apparatus and computer-readable storage medium

Abstract

The present disclosure relates to the technical field of computers, and relates to an article generation method, an article generation apparatus and a computer-readable medium. The article generation method comprises generating a target text on the basis of a summary of a plurality of first articles related to a topic, wherein at least one first article among the plurality of first articles comprises a first text paragraph and a first picture; for the at least one first article, determining a first degree of association between the first text paragraph and the first picture comprised therein; on the basis of the target text and the first degree of association, determining a first picture corresponding to at least one second text paragraph of the target text as a second picture; and generating a second article on the basis of the target text and the second picture.

Description

Article generation method and device, and computer readable storage medium TECHNICAL FIELD

[0001] The present disclosure relates to the technical field of computer, and particularly relates to an article generation method, device and computer readable storage medium. BACKGROUND

[0002] With the vigorous development of mobile internet technology, a large amount of articles are gathered in various information flow platforms, providing users with a variety of choices. Sometimes, content creators tend to create articles around the same hot topic to attract traffic, and the content is seriously homogenized. Users need to read multiple articles on the same topic, and then extract and integrate the content to obtain the required information.

[0003] In addition, single text content can make users feel boring and difficult to attract the attention of users.

[0004] SUMMARY

[0005] In view of this, the embodiments of the present disclosure provide an article generation method, device and computer readable storage medium. According to the summary of multiple first articles, a target text is generated, and according to the first correlation degree between the first text paragraph and the first picture included in the first article, a picture is matched for the generated target text, so that a high-quality article with picture-text mixed arrangement is generated, and the interest of users in reading is improved.

[0006] According to a first aspect of some embodiments of the present disclosure, an article generation method is provided, including generating a target text according to a summary of multiple first articles related to a theme, wherein at least one first article of the multiple first articles includes a first text paragraph and a first picture; for the at least one first article, determining a first correlation degree between the first text paragraph and the first picture included in the at least one first article; according to the target text and the first correlation degree, determining a first picture corresponding to at least one second text paragraph of the target text as a second picture; and generating a second article according to the target text and the second picture.

[0007] According to a second aspect of some embodiments of the present disclosure, there is provided an article generation apparatus, comprising: a text generation module configured to generate a target text according to a summary of a plurality of first articles related to a topic, wherein at least one first article of the plurality of first articles comprises a first text passage and a first picture; a relevance determination module configured to determine, for the at least one first article, a first relevance between the first text passage and the first picture comprised by the at least one first article; a picture determination module configured to determine, according to the target text and the first relevance, a first picture corresponding to at least one second text passage of the target text as a second picture; and an article generation module configured to generate a second article according to the target text and the second picture.

[0008] According to a third aspect of the present disclosure, there is provided an article generation apparatus, comprising: a memory; and a processor coupled to the memory, the processor being configured to execute an article generation method according to any of the embodiments of the present disclosure based on instructions stored in the memory.

[0009] According to a fourth aspect of the present disclosure, there is provided a computer readable storage medium having computer program instructions stored thereon, the instructions being executable by a processor to implement an article generation method according to any of the embodiments of the present disclosure.

[0010] According to a fifth aspect of the present disclosure, there is provided a computer program product comprising computer program instructions, the computer program instructions being executable by a processor to implement an article generation method according to any of the embodiments of the present disclosure.

[0011] This summary is provided to introduce some concepts of the present disclosure in a simplified form that are further described below in the detailed description section. This summary is not intended to identify key features or essential features of the claimed technology, nor is it intended to be used to limit the scope of the claimed technology.

[0012] Other features of the present disclosure, and their advantages, will become apparent from the following detailed description of illustrative embodiments of the present disclosure, when considered in conjunction with the accompanying drawings. BRIEF DESCRIPTION OF DRAWINGS

[0013] The preferred embodiments of the present disclosure will be described below with reference to the accompanying drawings. The accompanying drawings are used to provide further understanding of the present disclosure, and together with the following detailed description, form a part of the specification, and are included to further explain the present disclosure. It should be understood that the accompanying drawings only relate to some embodiments of the present disclosure, and do not limit the present disclosure. In the drawings:

[0014] FIG. 1 shows a flowchart of an article generation method according to some embodiments of the present disclosure;

[0015] FIG. 2A shows a method of generating target text according to some embodiments of the present disclosure;

[0016] FIG. 2B shows a schematic diagram of determining a second picture according to some embodiments of the present disclosure;

[0017] FIG. 2C shows a schematic diagram of determining a second picture according to some other embodiments of the present disclosure;

[0018] FIG. 3 shows a schematic diagram of a user interface according to some embodiments of the present disclosure;

[0019] FIG. 4 shows a schematic diagram of generating a summary of a paragraph according to some embodiments of the present disclosure;

[0020] FIG. 5 shows a schematic diagram of determining a second picture according to still other embodiments of the present disclosure;

[0021] FIG. 6 shows a block diagram of an article generation apparatus according to some embodiments of the present disclosure;

[0022] FIG. 7 shows a block diagram of an article generation apparatus according to some other embodiments of the present disclosure;

[0023] FIG. 8 shows a block diagram of an electronic device according to some embodiments of the present disclosure.

[0024] It should be understood that the dimensions of the various portions shown in the drawings are shown for purposes of convenience and do not necessarily correspond to the actual proportions. Identical or similar reference numerals are used to indicate identical or similar components throughout the various figures. Thus, once a component is defined in one figure, it can not be discussed further in subsequent figures. DETAILED DESCRIPTION

[0025] The technical solutions in the embodiments of the present disclosure will be described clearly and completely below with reference to the drawings in the embodiments of the present disclosure. Obviously, the described embodiments are only some of the embodiments of the present disclosure, but not all the embodiments. The description of the embodiments below is actually only illustrative, and should not be construed as any limitation on the present disclosure and its application or use. It should be understood that the present disclosure can be implemented in various forms, and should not be construed as being limited to the embodiments described herein.

[0026] It should be understood that the various steps recited in the method embodiments of the present disclosure can be performed in different orders and / or in parallel. In addition, the method embodiments can include additional steps and / or omit the steps shown. The scope of the present disclosure is not limited in this respect. Unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions, and numerical values set forth in these embodiments should be interpreted as merely exemplary, and not limiting to the scope of the present disclosure.

[0027] The term "include" and variations thereof, used in the present disclosure, mean a non-exclusive inclusion, such that a process, method, apparatus, article, or apparatus that comprises a list of elements is not necessarily limited to those elements, but can include other elements not expressly listed or inherent to such process, method, apparatus, article, or apparatus. Further, the term "consisting of" is intended to be synonymous with "consisting only of" such that a process, method, apparatus, article, or apparatus consisting of a list of elements includes only those elements.

[0028] Reference throughout this specification to "one embodiment", "some embodiments" or "an embodiment" means that a particular feature, structure or characteristic described in connection with the embodiment is included in at least one embodiment of the application. The appearances of the phrases "in one embodiment", "in some embodiments" or "in an embodiment" in various places in the specification are not necessarily all referring to the same embodiment, although it can. Furthermore, the terms "a" or "an", as used herein, are defined as "one or more" when used in the context of expressing a quantity of objects.

[0029] It should be noted that the terms "first", "second", etc. used in the present disclosure are merely intended to distinguish different devices, modules or units, and do not imply the order or sequence of the functions of the devices, modules or units. Unless otherwise specified, the terms "first", "second", etc. are not intended to imply the order or sequence of the functions of the objects so described.

[0030] It should be noted that the terms "one", "multiple", etc. used in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise explicitly specified in the context, "one" or "multiple" should be understood as "one or more".

[0031] The names of the messages or information exchanged between the devices in the embodiments of the present disclosure are merely for illustrative purposes, and are not intended to limit the scope of the messages or information.

[0032] The embodiments of the present disclosure will be described in detail below with reference to the accompanying drawings, but the present disclosure is not limited to these specific embodiments. The following specific embodiments can be combined with each other, and for the same or similar concepts or processes, some embodiments can not be described again. In addition, in one or more embodiments, specific features, structures or characteristics can be combined by any suitable means from the present disclosure that will be clear to those skilled in the art.

[0033] Currently, there is a lack of research on article automatic image matching technology, and it is difficult to meet the reading needs of users. Embodiments of the present disclosure provide an article generation method, device and computer readable storage medium. According to the summary of a plurality of first articles, a target text is generated, and a first image is matched for the generated target text according to a first correlation degree between a first text paragraph and the first image included in the first article, so that a more high-quality picture-text mixed article is generated, and the interest of users in reading is improved.

[0034] FIG. 1 shows a flowchart of an article generation method according to some embodiments of the present disclosure.

[0035] As shown in FIG. 1, the article generation method includes: step S1, generating a target text according to a summary of a plurality of first articles related to a theme, wherein at least one first article of the plurality of first articles includes a first text paragraph and a first image; step S2, determining a first correlation degree between the first text paragraph and the first image included in the at least one first article; step S3, determining a first image corresponding to at least one second text paragraph of the target text as a second image according to the target text and the first correlation degree; and step S4, generating a second article according to the target text and the second image.

[0036] The text generation method of this embodiment can be executed on a client or partially on a server.

[0037] The first article is, for example, an article written by an author. At least one first article includes both a text part (i.e., a first text paragraph) and an image (i.e., a first image). That is, the first image is the original image in the first article.

[0038] The first article is, for example, an article in the fields of science and technology, games, fashion, sports, etc. Taking the field of sports as an example, the theme of the first article is, for example, the World Cup, the Olympic Games, the championship, etc.

[0039] The first correlation degree between the first text paragraph and the first image represents the degree of correlation between the content of the first text paragraph and the content of the first image.

[0040] The second article includes the target text and the second image. According to the embodiments of the present disclosure, the text in the second article is generated according to the text in the first article, and the second article is matched with the image through the correlation of the original image in the first article, which improves the correlation of the image and the text in the second article, and a more high-quality picture-text mixed second article can be obtained.

[0041] At the same time, by integrating the text summary of a plurality of first articles to generate the text of the second article, multi-text fusion and creation are realized, and the needs of users for efficient information acquisition are met.

[0042] The method of integrating multiple first articles and generating a target text according to some embodiments of the present disclosure will be introduced below in combination with FIG. 2A.

[0043] The summary of the multiple first articles includes a summary of a first text paragraph of the multiple first articles, as shown in FIG. 2A. The generating of the target text according to the summary of the multiple first articles related to the theme in step S1 includes: generating a summary of a first text paragraph of the multiple first articles by using a first machine learning model in step S11; and generating the target text by using a second machine learning model according to the summary of the first text paragraph in step S12.

[0044] For example, for a first article with multiple first text paragraphs, a summary of each first text paragraph is generated. The summary of the first text paragraph can include part of the key content of the first text paragraph, or can be a re-expression of the main idea or key information of the first text paragraph after extraction. By summarizing each first text paragraph, the content of the multiple first articles can be integrated and compressed while reducing the loss of key information of the first articles.

[0045] In some embodiments, the maximum input length of the first machine learning model is greater than the maximum input length of the second machine learning model.

[0046] For example, the second machine learning model has a limit on the length of the input text. If the original text of the multiple first articles is directly input into the second machine learning model after splicing, it may exceed the maximum input length allowed by the second machine learning model. By using the first machine learning model to generate a summary of the first text paragraph and extract the core content, the information of the multiple first articles is compressed and simplified, which helps to generate the target text by using the second machine learning model subsequently.

[0047] The first machine learning model is, for example, a large language model (LLM), and the second machine learning model is, for example, a dialogue generation model.

[0048] In some embodiments, the generating of the target text by using the second machine learning model according to the summary of the first text paragraph includes: generating an outline of a second article by using a third machine learning model according to the summary of the first text paragraph; and generating the target text by using the second machine learning model according to the outline of the second article.

[0049] For example, the summaries of the paragraphs of the multiple first articles are spliced together, and the outline of the second article is generated by using the third machine learning model according to the spliced summaries of the paragraphs. The target text is generated by using the second machine learning model according to the outline.

[0050] According to some of the above embodiments of the present disclosure, the information of the multiple first articles is integrated first, and then the body is perfected around the outline. This creation process is closer to the process of people creating articles, and can make the structure of the generated second article more reasonable, and enhance the coherence and logic of the content.

[0051] By refining the outline from the model, the different content, viewpoints, perspectives, etc. of the multiple first articles are thoroughly understood, and the content is re-created, so that the second article provides more comprehensive information. The total number of words of the outline is less than the total number of words of the summaries of the paragraphs of the multiple first articles, so that the article architecture is determined while further compressing and simplifying the information to meet the requirements of the second machine learning model for the length of the input text.

[0052] In some embodiments, according to the outline of the second article, the target text is generated using the second machine learning model, including: according to the outline of the second article and the specified theme, the target text is generated using the second machine learning model.

[0053] For example, the theme of the target text to be generated is specified in advance. By generating the target text according to the outline of the second article and the specified theme, the content of the generated target text is more closely around the specified theme. The specified theme of the second article can also be taken from the theme of the first article.

[0054] In some embodiments, the article generation method further includes determining the theme of each first article of the multiple first articles before generating the target text according to the summaries of the multiple first articles related to the theme.

[0055] For example, search for related articles around the hot words. Then, content recall is performed to obtain a set of articles that meet the user's interest. The themes of the articles in the set are obtained by text clustering or the like. For first articles of different themes, articles of the same theme can be classified into a category. When screening the first articles, articles of the same theme are selected as the multiple first articles. When screening the first articles, the similarity between different themes can also be calculated, so that the multiple first articles have a similarity between themes that exceeds a threshold.

[0056] In some embodiments, the first text paragraph of the multiple first articles is generated using the first machine learning model, including: for each first article of the multiple first articles, the first text paragraph of the current first article is generated using the first machine learning model according to the current first text paragraph and the summary of the previous paragraph before the current first text paragraph.

[0057] For example, when generating the passage summary for each first text passage, not only the text of the first text passage, but also the passage summary of the previous first text passage is considered. Compared with only relying on the current passage, this method provides more information for the first machine learning model, deepens the model's understanding of the current passage, so that the generated passage summary is more accurate, and reduces the information loss in the text compression process from passage to passage summary,

[0058] In some embodiments, generating the passage summary corresponding to the first text passage of each first article by using the first machine learning model comprises: for the first text passage of each first article, generating the passage summary corresponding to the current first text passage by using the first machine learning model according to the current first text passage, the passage summary corresponding to the previous passage, and the partial text of the previous passage.

[0059] For example, when generating the passage summary for each first text passage, not only the text of the first text passage, but also the passage summary of the previous first text passage is considered. Compared with only relying on the current passage, this method provides more information for the first machine learning model, deepens the model's understanding of the current passage, so that the generated passage summary is more accurate, and reduces the information loss in the text compression process from passage to passage summary,

[0060] In some embodiments, for the first text passage of each first article, generating the passage summary corresponding to the current first text passage by using the first machine learning model according to the current first text passage, the passage summary corresponding to the previous passage, and the partial text of the previous passage comprises: generating the passage summary corresponding to the current first text passage by using the first machine learning model according to the current first text passage, the summary of the previous passage, and the partial text of the previous passage adjacent to the current first text passage.

[0061] For example, assuming that the word threshold is set to 1000 words, and the text length of the current first text passage is 800 words, the last 200 words of the previous first text passage are taken and concatenated with the 800 words of the current first text passage as input for the first machine learning model. That is, when generating the passage summary for the adjacent two first text passages, the taken texts are partially overlapped, thereby further reducing the information loss in the text compression process from passage to passage summary.

[0062] The partial text of the previous first text passage can also be randomly taken, for example, assuming that the word threshold is set to 1000 words, and the text length of the current first text passage is 800 words, 200 consecutive words at a random position of the previous first text passage are taken and concatenated with the 800 words of the current first text passage as input for the first machine learning model.

[0063] The method of determining the second picture according to some embodiments of the present disclosure is described below in combination with FIGS. 2B-2C.

[0064] As shown in FIG. 2B, the step S3 of determining the first picture corresponding to the at least one second text paragraph of the target text as the second picture according to the target text and the first correlation degree comprises: a step S31 of determining the first text paragraph having a first correlation degree with the first picture exceeding a first threshold value for the first picture included in the at least one first article; a step S32 of segmenting the determined first text paragraph into one or more first sentences; a step S33 of segmenting the at least one second text paragraph into one or more second sentences; a step S34 of selecting at least one first sentence corresponding to each second sentence of the one or more second sentences from the one or more first sentences; and a step S35 of determining the first picture corresponding to the at least one second text paragraph as the second picture according to the first correlation degree and the first text paragraph to which the at least one first sentence belongs.

[0065] For example, for the at least one first article including the picture and the generated target text, the target text is segmented into sentences respectively. The first sentence corresponding to the second sentence in the target text is selected, and the first picture related to the content of the second text paragraph is determined as the second picture according to the first correlation degree between the first text paragraph from which the corresponding first sentence is derived and the first picture.

[0066] In some embodiments, for the at least one first article of the plurality of first articles, the first correlation degree between the first text paragraph and the first picture included in the first article is determined, comprising: determining the first correlation degree between the first text paragraph and the first picture according to the relative position of the first text paragraph and the first picture in the plurality of first articles.

[0067] For example, the closer the first text paragraph is to the first picture a, the higher the first correlation degree between the first text paragraph and the first picture a. If there are multiple pictures after a first text paragraph (without text in between), the distance between the multiple pictures and the first text paragraph can no longer be distinguished. That is, the first correlation degree between the first text paragraph and the multiple pictures is the same, which is the highest. Alternatively, the distance between the first text paragraph and the multiple pictures behind can be further distinguished, for example, the first correlation degree between the first picture after the first text paragraph and the first text paragraph is the highest, the first correlation degree between the second picture after the first text paragraph and the first text paragraph is the second highest, and so on.

[0068] Generally, when creating the first article, the author has already evaluated the layout of the text and pictures, so that the inserted pictures between paragraphs not only revolve around the theme of the article, but also combine the content discussed by the paragraphs near the pictures. Generally, a picture and the nearest paragraph of the picture are related and most closely related.

[0069] Some embodiments of the present disclosure establish a mapping relationship between the picture and the paragraph in the first article through the first correlation degree. Then, by comparing the paragraph in the first article with the paragraph in the second article, and combining the mapping relationship between the picture and the paragraph in the first article, the corresponding relationship between the paragraph in the second article and the picture in the first article is found, thereby providing the second article with pictures.

[0070] Compared with analyzing the semantics of the picture by using the artificial intelligence picture model, the above-mentioned way of providing the second article with pictures converts the requirement of analyzing the picture into the requirement of analyzing the text, thereby reducing the calculation cost and improving the efficiency.

[0071] In some embodiments, selecting, from the one or more first sentences, at least one first sentence corresponding to each second sentence of the one or more second sentences comprises: determining the similarity between each second sentence and the one or more first sentences; and selecting, according to the similarity between each second sentence and the one or more first sentences, at least one first sentence corresponding to each second sentence.

[0072] For example, the first sentence with a similarity to the second sentence exceeding a threshold value is determined as the first sentence corresponding to the second sentence. As described above, by establishing the mapping relationship between the picture and the paragraph in the first article, and comparing the similarity between the paragraph in the first article and the sentence in the second article to provide the picture, the requirement of analyzing the picture is converted into the requirement of analyzing the text. Compared with analyzing the semantics of the picture by using the artificial intelligence picture model, the calculation speed of analyzing the text is faster, and the resource consumption is lower. Moreover, the picture understanding ability of the current artificial intelligence model is not ideal, and in comparison, the code engineering means of comparing and analyzing between texts has better stability and accuracy.

[0073] In some embodiments, determining the similarity between each second sentence and the one or more first sentences comprises: determining the vector of each second sentence; determining the vector of each first sentence of the one or more first sentences; and determining the similarity between each second sentence and each first sentence according to the vector of each second sentence and the vector of each first sentence.

[0074] For example, the first sentence and the second sentence are respectively converted into vectors, the vector of the first sentence is stored in a vector database, the vector of the second sentence is used to query in the vector database, and the first Y sentences with high similarity are selected as the corresponding first sentences. The similarity between the first sentence and the second sentence is determined by calculating the similarity between the vectors. Y is a positive integer.

[0075] Compared with converting each paragraph as a whole into a vector, vectorizing each sentence separately can preserve more information. Because the dimension of the vector that the model can calculate is limited, in the limited dimension, the vector obtained by cutting the text and then converting contains more complete semantics, which can more accurately explain the original text, so that the similarity calculated is more accurate, which can more accurately match the second text with the picture.

[0076] At the same time, compared with using an artificial intelligence model to analyze the semantics of the picture, the text vector analysis code engineering method has better stability and accuracy.

[0077] FIG. 2C shows a schematic diagram of determining a second picture according to some embodiments of the present disclosure.

[0078] As shown in FIG. 2C, the step S35 of determining the first picture corresponding to each second text paragraph as the second picture according to the first correlation degree and the first text paragraph to which the at least one first sentence belongs includes: a step S351 of determining a second correlation degree between each second sentence and the first picture according to the similarity between each second sentence and the corresponding at least one first sentence, the first correlation degree between the first text paragraph to which the at least one first sentence belongs and the first picture; a step S352 of determining a third correlation degree between each second text paragraph and the first picture according to the second correlation degree between one or more second sentences of each second text paragraph and the first picture for each second text paragraph of the at least one second text paragraph; and a step S353 of determining the first picture corresponding to each second text paragraph as the second picture according to the third correlation degree between each second text paragraph and the first picture.

[0079] For example, for a first picture a and b, a second sentence A in a second text paragraph p1, the second correlation degree between the second sentence A and the first picture a is determined according to the similarity between the second sentence A and the corresponding first sentence, and the first correlation degree between the first text paragraph from which the corresponding first sentence of the second sentence A originates and the first picture a. Then, the second correlation degrees between all second sentences of the second text paragraph p1 and the first picture a are counted, so as to obtain the third correlation degree between the second text paragraph p1 and the first picture a. Similarly, the third correlation degree between the second text paragraph p1 and the first picture b can be obtained.

[0080] In some embodiments, the second correlation degree is positively correlated with the similarity; the second correlation degree is positively correlated with the first correlation degree; and / or the third correlation degree of each second text paragraph is a weighted sum of the second correlation degrees between one or more second sentences of each second text paragraph and the first picture.

[0081] For example, the second sentence A corresponds to the first sentence 1, the first sentence 2, and the first sentence 3, and the product of the similarity between the second sentence A and the first sentence 1 and the first association degree between the paragraph where the first sentence 1 is located and the first picture a is calculated. For the second sentence A, the products corresponding to the first sentence 1, the first sentence 2, and the first sentence 3 are calculated respectively, and the weighted sum of these products is calculated to obtain the second association degree between the second sentence A and the first picture a.

[0082] Then, the weighted sum of the second association degrees between all the second sentences of the second text paragraph p1 and the first picture a is calculated to obtain the third association degree between the second text paragraph p1 and the first picture a.

[0083] In some embodiments, determining the first picture corresponding to each second text paragraph as the second picture according to the third association degree between each second text paragraph and the first picture comprises: in the first pictures with the third association degree exceeding the second threshold value with each second text paragraph, the first pictures ranked in the top N according to the third association degree with each second text paragraph are determined as the second pictures corresponding to each second text paragraph, where N is a specified positive integer.

[0084] For example, a quantity threshold value N is set, and each paragraph is matched with at most N pictures. If the number of the first pictures with the third association degree exceeding the second threshold value with a second text paragraph is greater than N, the N first pictures with the highest third association degree with the second text paragraph (hereinafter referred to as candidate pictures) are selected as the second pictures corresponding to the second text paragraph.

[0085] In the case where multiple second text paragraphs have the same candidate picture, the same candidate picture is taken as the second picture of the second text paragraph with the highest third association degree, and the second pictures corresponding to other second text paragraphs do not include the same candidate picture.

[0086] In the distribution of the second picture, the third association degrees between all the second text paragraphs in the second article and the multiple first pictures can be calculated first. Then, the third association degrees of all the second text paragraphs in the second article are sorted from high to low, and the picture distribution is performed. The first picture already distributed to a certain second text paragraph will not be distributed to other second text paragraphs, so as to avoid picture repetition.

[0087] In some embodiments, generating the second article according to the target text and the second picture comprises: for the other second text paragraph of the target text, in the case where there is no first picture with the third association degree exceeding the second threshold value in the other second text paragraph, generating the third picture corresponding to the other second text paragraph according to the other second text paragraph by using the fourth machine learning model; and generating the second article according to the target text, the second picture, and the third picture.

[0088] The fourth machine learning model is a picture generation model. If the third correlation degree between the second text paragraph p2 and the first picture does not exceed the second threshold, that is, no suitable picture for the second text paragraph p2 is found in the first article, a fourth machine learning model is used to generate a picture according to the text content of the second text paragraph p2 as the picture for the second text paragraph p2. Alternatively, a natural language processing model can be used to extract the core semantics and / or discussion object from the second text paragraph p2, and then a fourth machine learning model is used to generate a third picture around the core semantics and / or discussion object of the paragraph. By supplementing the second picture with the third picture, the content of the generated second article is more abundant.

[0089] In some embodiments, generating the second article according to the target text and the second picture includes: for each second text paragraph of the at least one second text paragraph, determining the position of the corresponding second picture inserted into each second text paragraph according to at least one of the second correlation degree between one or more second sentences and the corresponding first picture, the size of the display page of the second article, the size of the maximum occupied area of the text on the display page, the size of the maximum occupied area of the picture, the number of first pictures corresponding to each second text paragraph, and the size of the first picture corresponding to each second text paragraph.

[0090] For example, according to the screen size of the user's mobile device, the user's personalized settings for font and display, one page of the screen displays at most 500 words, or at most 300 words and one picture at the same time. The second text paragraph has 800 words and corresponds to one picture, so one picture and 300 words are displayed on the same page, and the other 500 words are displayed on another page.

[0091] On a mobile terminal, when performing picture-text layout of the second article, by adjusting the position of the second picture inserted into the second text paragraph, at least part of the content of the second text paragraph and the corresponding second picture are displayed on the same page, so that the user does not need to search for related paragraphs and pictures.

[0092] FIG. 3 shows a schematic diagram of a user interface according to some embodiments of the present disclosure.

[0093] As shown in FIG. 3, one second text paragraph corresponds to two second pictures, and one page of the screen cannot display all the content of the second text paragraph completely, so the second text paragraph is divided into two parts by pictures, and displayed on two pages respectively, so that each of the first page and the second page has text and a picture. When reading, the user will not feel bored because the entire screen is full of text, thereby enhancing the interest in the article.

[0094] In some embodiments, the second article is generated according to the target text and the second picture, including: inserting the corresponding second picture after each of the at least one second text paragraph.

[0095] For example, when the second article is being laid out, the second picture is inserted immediately after the corresponding second text paragraph, so that the user can immediately refer to the picture intuitively after reading the paragraph, deepen the impression, and better understand the main idea or details of the paragraph.

[0096] In some embodiments, the article generation method further includes: obtaining a theme of the target text; and selecting, from the multiple text paragraphs in the second article, at least one text paragraph with a fourth correlation degree with the theme exceeding a threshold value as the at least one second text paragraph.

[0097] For example, the theme of the target text is determined by using a theme extraction model, or a preset theme is obtained. The higher the relevance of the second text paragraph to the theme, the more important the second text paragraph is. Therefore, selecting a paragraph with high relevance to the theme as a paragraph that needs to be illustrated is conducive to the user's understanding of the article and the theme.

[0098] The fourth correlation degree is calculated by, for example, a text similarity calculation method.

[0099] In some embodiments, the article generation method further includes: in response to the user's historical browsing records including multiple first articles and / or articles related to the themes of the multiple first articles, recommending the second article to the user.

[0100] For example, if the user has browsed the first article, or has browsed an article related to the first article, or has browsed an article related to the theme of the second article, the second article is pushed to the user, thereby providing more personalized content for the user.

[0101] FIG. 4 shows a schematic diagram of generating a paragraph summary according to some embodiments of the present disclosure.

[0102] As shown in FIG. 4, the text in the original first article is split by paragraph and sentence to obtain text data.

[0103] According to the word requirement, the text of each paragraph for generating the summary is determined by using the LLM model. The word requirement is, for example, a maximum number of words required for inputting the LLM model, for example, a word threshold of 1000 words is set, and the text length of the current first text paragraph is 800 words, then the last 200 words of the previous first text paragraph are taken and concatenated with the 800 words of the current first text paragraph to be input into the first machine learning model.

[0104] The word count requirement can also be a word count requirement for inter-paragraph overlap, for example, set the overlap word count to 300. Then take the last 300 words of the previous first text paragraph and concatenate them with the current first text paragraph as input to the LLM model.

[0105] For the first paragraph in each first article (i.e., the first text paragraph 1), use the text of the first paragraph to generate a summary of the first paragraph using the LLM model.

[0106] From the second paragraph, the current paragraph, part of the text of the previous paragraph, and the summaries of the previous paragraphs are all input to the LLM model to obtain a summary of the current paragraph. Recursively, the summary corresponding to each paragraph of all paragraphs is obtained.

[0107] The summaries of all paragraphs of all first articles are concatenated together, and a third machine learning model is used to obtain an outline of a second article. The outline of the second article is input to a second machine learning model to generate a target text that integrates the textual content of multiple first articles.

[0108] After generating the target text, pictures are also inserted into the target text. The following describes a method for determining second pictures according to some embodiments of the present disclosure in conjunction with FIG. 5.

[0109] Taking a first article as an example, first determine a first correlation degree between a paragraph of the first article and an adjacent picture, thereby obtaining a mapping relationship between the first text paragraph and the first picture, where x and y can be the same or different. The first text paragraph is split into multiple first sentences, and a vector of each first sentence is calculated. Each second text paragraph of the second article is also split into multiple second sentences, and the vector of the second sentence is used to query the sentence vector library to find similar first sentences, thereby obtaining a mapping relationship between the first text paragraph and the second text paragraph. According to the mapping relationship between the first text paragraph and the first picture and the mapping relationship between the first text paragraph and the second text paragraph, a first picture corresponding to the second text paragraph is determined as a second picture.

[0110] If there is a second text paragraph without a corresponding first picture, a picture generation model can also be used to generate a corresponding third picture according to the text of the second text paragraph.

[0111] After determining the second picture corresponding to the second text paragraph, the second picture is inserted after the corresponding second text paragraph, and the creation of the second article is completed. If there is a third picture, the third picture is also inserted after the corresponding paragraph.

[0112] The above is an article generation method provided by an embodiment of the present disclosure. The following describes an article generation device according to an embodiment of the present disclosure with reference to FIGS. 6-8, which is used to execute any of the above article generation methods.

[0113] FIG. 6 shows a block diagram of an article generation apparatus according to some embodiments of the present disclosure.

[0114] As shown in FIG. 6, the article generation apparatus 6 comprises: a text generation module 61 configured to generate a target text according to a summary of a plurality of first articles related to a topic, wherein at least one first article of the plurality of first articles comprises a first text paragraph and a first picture; a relevance determination module 62 configured to determine, for the at least one first article, a first relevance between the first text paragraph and the first picture comprised by the at least one first article; a picture determination module 63 configured to determine, according to the target text and the first relevance, a first picture corresponding to at least one second text paragraph of the target text as a second picture; and an article generation module 64 configured to generate a second article according to the target text and the second picture.

[0115] The text generation module 61 of the article generation apparatus 6 can be configured to perform the step S1 of FIG. 1. The relevance determination module 62 of the article generation apparatus 6 can be configured to perform the step S2 of FIG. 1. The picture determination module 63 of the article generation apparatus 6 can be configured to perform the step S3 of FIG. 1. The article generation module 64 of the article generation apparatus 6 can be configured to perform the step S4 of FIG. 1.

[0116] In some embodiments, the article generation apparatus 6 further comprises a paragraph selection module configured to: obtain a topic of the target text; and select, from a plurality of text paragraphs in the second article, at least one text paragraph having a fourth relevance to the topic exceeding a threshold value as the at least one second text paragraph.

[0117] In some embodiments, the article generation apparatus 6 further comprises a recommendation module configured to: in response to the user's historical browsing records comprising the plurality of first articles and / or articles related to the topics of the plurality of first articles, recommend the second article to the user.

[0118] FIG. 7 shows a block diagram of an article generation apparatus according to some embodiments of the present disclosure.

[0119] As shown in FIG. 7, the article generation apparatus 7 comprises: a memory 71; and a processor 72 coupled to the memory 71, the processor 72 being configured to perform the article generation method of any of the preceding embodiments based on instructions stored in the memory 71.

[0120] The memory 71 is configured to store one or more computer-readable instructions. The memory 71 can include any combination of various computer-readable storage media, such as volatile memory and / or non-volatile memory, including, but not limited to, random access memory (RAM), dynamic random access memory (DRAM), static random access memory (SRAM), read-only memory (ROM), flash memory. The memory 71 can store, for example, an operating system, an application program, a boot loader, a database, and other programs, and can also store various application programs and various data.

[0121] The processor 72 is configured to run the computer-readable instructions to implement the article generation method of any of the preceding embodiments. For specific implementation of each step of the article generation method, please refer to the above-mentioned embodiments, and the repeated parts will not be described here.

[0122] The processor 72 can be configured to perform steps S1-S4 of FIG. 1 or steps S1'-S3' of FIG. 7. The processor 72 can be embodied as various processing devices, such as a central processing unit (CPU), a network processing unit (NP), etc.; and can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component. The central processing unit (CPU) can be X86 or ARM architecture, etc.

[0123] The processor 72 and the memory 71 can communicate with each other directly or indirectly. For example, the processor 72 and the memory 71 can communicate through a network. The network can include a wireless network, a wired network, and / or any combination of a wireless network and a wired network. The processor 72 and the memory 71 can also communicate with each other through a system bus, and the present disclosure does not limit the communication between the processor 72 and the memory 71.

[0124] It should be noted that the components of the article generation device 7 shown in FIG. 7 are only exemplary and not limiting, and the article generation device 7 can also have other components according to actual application needs. The processor 72 can control other components in the article generation device 7 to perform the desired functions.

[0125] The article generation device can be implemented by software, firmware and / or hardware, and can be integrated in an electronic device installed with a related application program.

[0126] FIG. 8 shows a block diagram of an electronic device according to some embodiments of the present disclosure.

[0127] The electronic device 8 shown in FIG. 8 can be a computer system with a special hardware structure, which can perform corresponding functions when installed with a related application program.

[0128] The electronic device includes, but is not limited to, a mobile terminal such as a smartphone, a notebook computer, a Personal Digital Assistant (PDA), a Tablet Personal Computer (Tablet PC), a PMP (Portable Multimedia Player), a car terminal (e.g., a car navigation terminal), a wearable device, and the like, and a stationary terminal such as a digital television, a desktop computer, and the like.

[0129] As shown in FIG. 8, a central processing unit (CPU) 81 performs various processes according to a program stored in a read only memory (ROM) 82 or a program loaded from a storage section 88 to a random access memory (RAM) 83. In the RAM 83, data required when the CPU 81 performs various processes and the like is stored as necessary. The central processing unit is merely exemplary and can be other types of processors such as the various processors described above. The ROM 82, the RAM 83, and the storage section 88 can be various forms of computer readable storage media. Note that, although the ROM 82, the RAM 83, and the storage section 88 are shown separately in FIG. 8, one or more of them can be combined or located in the same or different memory or storage module.

[0130] The CPU 81, the ROM 82, and the RAM 83 are connected to each other via a bus 84. An input / output interface 85 is also connected to the bus 84.

[0131] The following components are connected to the input / output interface 85: an input section 86 including a touch screen, a touch pad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, and the like; an output section 87 including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), a speaker, a vibrator, and the like; a storage section 88 including a hard disk, a magnetic tape, and the like; and a communication section 89 including a network interface card such as a LAN card, a modem, and the like. The communication section 89 allows communication processing to be performed via a network such as the Internet. It is readily understood that, although the various devices or modules in the electronic device 8 are shown in FIG. 8 as communicating via the bus 84, they can also communicate through a network or other means, where the network can include a wireless network, a wired network, and / or any combination of a wireless network and a wired network.

[0132] A drive 810 is also connected to the input / output interface 85 as necessary. A removable medium 811 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, and the like is mounted on the drive 810 as necessary, so that a computer program read therefrom is installed in the storage section 88 as necessary.

[0133] In a case where the above series of processes are realized by software, the program constituting the software can be installed from a network such as the Internet or a storage medium 811 such as a detachable medium.

[0134] According to an embodiment of the present disclosure, the processes described above with reference to the flowcharts can be implemented as a computer software program. For example, some embodiments of the present disclosure include a computer program product that, when run on a computer, causes the computer to implement the article generation method described in any of the preceding embodiments. The computer program product includes a computer program carried on a computer readable medium, which contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication device 89, or installed from the storage section 88, or installed from the ROM 82. When the computer program is executed by the CPU 81, the article generation method of the embodiments of the present disclosure is executed.

[0135] Note that, in the context of the present disclosure, the computer readable medium can be a tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device.

[0136] The computer readable medium can be a computer readable storage medium, or a computer readable signal medium, or any combination of the two.

[0137] The computer readable storage medium includes, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, the computer readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. The computer readable storage medium stores a computer program that, when executed by a processor, implements the article generation method described in any of the preceding embodiments.

[0138] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or computer readable storage devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other computer readable storage devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flow diagrams and / or diagrams.

[0139] The computer readable medium can be included within the electronic device; or can exist solely at the electronic device.

[0140] In some embodiments, a computer program including instructions which, when executed by a processor, causes the processor to carry out the article generation method of any one of the above embodiments is also provided. For example, the instructions can be embodied as computer program code.

[0141] In an embodiment of the disclosure, the computer program code for carrying out operations of the disclosure can be written in one or more programming languages or combinations of languages including object oriented programming languages such as Java, Smalltalk, C++ or conventional procedural programming languages such as "C" or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0142] The computer program product of the first aspect can include one or more non-transitory computer-readable media storing instructions that, when executed by one or more processors of a computing device, cause the one or more processors to perform the operations of the method of the first aspect. The one or more non-transitory computer-readable media can include one or more of the following: a magnetic storage device, an optical storage device, a solid-state storage device, a hard disk drive, a flash drive, a RAM, a ROM, a database, and / or any other suitable type of non-transitory computer-readable medium.

[0143] The functions described above can be performed, at least in part, by one or more hardware logic components. For example, and without limitation, non-limiting examples of hardware logic components that can be used include field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), system-on-a-chip (SOCs), complex programmable logic devices (CPLDs), etc.

[0144] While certain aspects of the present disclosure have been described with reference to particular examples, those of ordinary skill in the art will understand that various other modifications can be made to the examples and in other aspects, without departing from the scope and spirit of the disclosure. It is not intended that the disclosure be limited as described above, but instead that the scope of the disclosure be measured by the broadest interpretation of the following claims.

Claims

1. A method for generating an article, comprising: Based on a summary of multiple first articles related to the topic, target text is generated, wherein at least one of the multiple first articles includes a first text paragraph and a first image; For the at least one first article, determine the first degree of correlation between the first text paragraphs included therein and the first image; Based on the target text and the first relevance, a first image corresponding to at least one second text paragraph of the target text is determined and used as the second image; A second article is generated based on the target text and the second image.

2. The article generation method according to claim 1, wherein, The summary of the multiple first articles includes paragraph summaries corresponding to the first text paragraphs of the multiple first articles. The step of generating target text based on the summaries of the multiple first articles related to the topic includes: Using the first machine learning model, generate paragraph summaries corresponding to the first text paragraphs of the multiple first articles; Based on the summary of the paragraph, the target text is generated using a second machine learning model.

3. The article generation method according to claim 2, wherein, The step of summarizing the paragraph and using a second machine learning model to generate the target text includes: Based on the summary of the paragraphs, an outline for the second article is generated using a third machine learning model; Based on the outline of the second article, the target text is generated using the second machine learning model.

4. The article generation method according to claim 3, wherein, The step of generating paragraph summaries corresponding to the first text paragraphs of the multiple first articles using the first machine learning model includes: For the first text paragraph of each of the multiple first articles, the first machine learning model is used to generate a paragraph summary corresponding to the current first text paragraph based on the paragraph summary corresponding to the previous paragraph before the current first text paragraph.

5. The article generation method according to claim 4, wherein, The process of generating paragraph summaries corresponding to the first text paragraphs of the multiple first articles using the first machine learning model includes: For the first text paragraph of each first article, based on the current first text paragraph, the paragraph summary corresponding to the preceding paragraph, and a portion of the text of the preceding paragraph, the first machine learning model is used to generate the paragraph summary corresponding to the current first text paragraph.

6. The article generation method according to claim 5, wherein, For each first text paragraph of the first article, based on the current first text paragraph, the paragraph summary corresponding to the preceding paragraph, and a portion of the text of the preceding paragraph, the first machine learning model is used to generate the paragraph summary corresponding to the current first text paragraph, including: Based on the current first text paragraph and the summary of the preceding paragraph, the text adjacent to the current first text paragraph in the preceding paragraph is used to generate a paragraph summary corresponding to the current first text paragraph using the first machine learning model.

7. The article generation method according to claim 2, wherein, The maximum input length of the first machine learning model is greater than the maximum input length of the second machine learning model.

8. The article generation method according to claim 1, wherein, The step of determining, based on the target text and the first relevance, at least one second text paragraph corresponding to the first image as the second image includes: For a first image included in the at least one first article, determine a first text paragraph whose first relevance to the first image exceeds a first threshold; Divide the identified first text paragraph into one or more first sentences; The at least one second text paragraph is divided into one or more second sentences; From the one or more first sentences, select at least one first sentence that corresponds to each of the one or more second sentences; Based on the first correlation degree and the first text paragraph to which the at least one first sentence belongs, a first image corresponding to the at least one second text paragraph is determined as the second image.

9. The article generation method according to claim 8, wherein, Selecting at least one first sentence from the one or more first sentences that corresponds to each of the one or more second sentences includes: Determine the similarity between each second sentence and one or more first sentences; Based on the similarity between each second sentence and one or more first sentences, select at least one first sentence corresponding to each second sentence.

10. The article generation method according to claim 9, wherein, Determining the similarity between each second sentence and one or more first sentences includes: Determine the vector for each of the second sentences; Determine the vector of each of the one or more first sentences; The similarity between each second sentence and each first sentence is determined based on the vector of each second sentence and the vector of each first sentence.

11. The article generation method according to claim 9, wherein, The step of determining the first image corresponding to the at least one second text paragraph, as the second image, based on the first relevance and the first text paragraph to which the at least one first sentence belongs, includes: A second correlation between each second sentence and the first image is determined based on the similarity between each second sentence and the corresponding at least one first sentence, the first correlation between the first text paragraph to which the at least one first sentence belongs and the first image. For each of the at least one second text paragraphs, a third degree of association between each second text paragraph and the first image is determined based on a second degree of association between the one or more second sentences of each second text paragraph and the first image; Based on the third correlation between each second text paragraph and the first image, a first image corresponding to each second text paragraph is determined as the second image.

12. The article generation method according to claim 11, wherein: The second correlation is positively correlated with the similarity. The second correlation is positively correlated with the first correlation. and / or The third relevance of each second text paragraph is a weighted sum of the second relevances between the one or more second sentences of each second text paragraph and the first image.

13. The article generation method according to claim 11, wherein, The step of determining the first image corresponding to each second text paragraph as the second image based on the third correlation degree between each second text paragraph and the first image includes: Among the first images whose third relevance to each second text paragraph exceeds a second threshold, the N first images with the highest third relevance to each second text paragraph are determined as the second images corresponding to each second text paragraph, where N is a specified positive integer.

14. The article generation method according to claim 11, wherein, The step of generating a second article based on the target text and the second image includes: For other second text paragraphs of the target text, if there is no first image with a third relevance exceeding the second threshold in the other second text paragraphs, a fourth machine learning model is used to generate a third image corresponding to the other second text paragraphs based on the other second text paragraphs; A second article is generated based on the target text, the second image, and the third image.

15. The article generation method according to claim 11, wherein, The step of generating a second article based on the target text and the second image includes: For each of the at least one second text paragraphs, the position to which the corresponding second image is inserted is determined based on at least one of the following: the second correlation between the one or more second sentences and the corresponding first image; the size of the display page of the second article; the size of the maximum occupied area of ​​the text in the display page; the size of the maximum occupied area of ​​the image; the number of first images corresponding to each second text paragraph; and the size of the first image corresponding to each second text paragraph.

16. The article generation method according to claim 1, wherein, The step of generating a second article based on the target text and the second image includes: After each second text paragraph (at least one second text paragraph), insert the corresponding second image.

17. The article generation method according to claim 1, wherein, Determining the first correlation degree between the first text paragraphs and the first image included in at least one of the plurality of first articles includes: Based on the relative positions of the first text paragraph and the first image in the plurality of first articles, a first degree of correlation between the first text paragraph and the first image is determined.

18. The article generation method according to claim 1, further comprising: Obtain the topic of the target text; From a plurality of text paragraphs in the second article, at least one text paragraph whose fourth relevance to the topic exceeds a threshold is selected as the at least one second text paragraph.

19. The article generation method according to claim 1, further comprising: In response to the user's browsing history including the multiple first articles and / or articles related to the topic of the multiple first articles, the second article is recommended to the user.

20. An article generation device, comprising: The text generation module is configured to generate target text based on a summary of multiple first articles related to the topic, wherein at least one of the multiple first articles includes a first text paragraph and a first image; The correlation determination module is configured to determine, for the at least one first article, a first correlation between a first text paragraph and a first image. The image determination module is configured to determine, based on the target text and the first relevance, a first image corresponding to at least one second text paragraph of the target text, and use it as the second image; The article generation module is configured to generate a second article based on the target text and the second image.

21. An article generation device, comprising: Memory; as well as A processor coupled to the memory, the processor being configured to operate based on instructions stored in the memory. Perform the article generation method according to any one of claims 1 to 19.

22. A computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the article generation method according to any one of claims 1 to 19.

23. A computer program product comprising computer program instructions that, when executed by a processor, implement the article generation method according to any one of claims 1 to 19.

Images