Content generation method and apparatus, electronic device, medium, and product

Abstract

Embodiments of the present disclosure relate to a content generation method and apparatus, an electronic device, a computer readable storage medium, and a computer program product. The method comprises, on the basis of at least one label associated with source content, from among a plurality of objects for content generation, selecting at least one candidate object matching the source content. The method further comprises selecting, by a preset model, a target object from among the at least one candidate object on the basis of the source content. The method further comprises generating target content from the target object on the basis of the source content, wherein the target content comprises target information associated with the source content, and the target information is associated with an attribute of the target object.

Description

Method, device, electronic device, medium and product for content generation TECHNICAL FIELD

[0001] Embodiments of the present disclosure relate to the field of computer, and more particularly, to a method, device, electronic device, computer readable storage medium and computer program product for content generation. BACKGROUND

[0002] Traditional content distribution platforms rely on user-generated content, such as writing articles, making videos, and forwarding and sharing. This mode promotes the diversity of content, but is often inefficient because the content creation and distribution speed is limited by the user's personal time and interest. At the same time, due to the lack of instant content screening and recommendation mechanism, real-time performance is difficult to meet, and the interestingness of the content is uneven, which is difficult to accurately meet the diversified needs of users.

[0003] SUMMARY

[0004] Embodiments of the present disclosure provide a method, device, electronic device, computer readable storage medium and computer program product for content generation.

[0005] According to a first aspect of the present disclosure, a method for content generation is provided. The method comprises selecting, based on at least one label associated with source content, at least one candidate object matching the source content from a plurality of objects for content generation. The method further comprises selecting, by a preset model, a target object from the at least one candidate object based on the source content. The method further comprises generating, by the target object, target content based on the source content, wherein the target content comprises target information associated with the source content, the target information being associated with an attribute of the target object.

[0006] According to a second aspect of the present disclosure, a device for content generation is provided. The device comprises a first selection module configured to select, based on at least one label associated with source content, at least one candidate object matching the source content from a plurality of objects for content generation. The device further comprises a second selection module configured to select, by a preset model, a target object from the at least one candidate object based on the source content. The device further comprises a content generation module configured to generate, by the target object, target content based on the source content, the target content comprising target information associated with the source content, the target information being associated with an attribute of the target object.

[0007] According to a third aspect of the present disclosure, an electronic device is provided. The electronic device comprises a processor and a memory coupled with the processor, the memory having stored therein instructions that, when executed by the processor, cause the electronic device to perform the method according to the first aspect.

[0008] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided. The computer-readable storage medium has stored thereon computer instructions, the computer instructions being executable by a processor to implement the method according to the first aspect.

[0009] According to a fifth aspect of the present disclosure, a computer program product is provided. The computer program product is tangibly stored on a computer-readable storage medium and includes computer-executable instructions, which, when executed by a device, cause the device to perform the method according to the first aspect.

[0010] The summary is provided to introduce a selection of concepts, in a simplified form, that are further described below in the DETAILED DESCRIPTION. This summary is not intended to identify key or essential features of the claimed subject matter, nor is it intended to limit the scope of the claimed subject matter. BRIEF DESCRIPTION OF DRAWINGS

[0011] The above and other features, aspects and advantages of various embodiments of the present disclosure will become more apparent from the following detailed description, taken in conjunction with the accompanying drawings, in which like reference numbers represent like elements throughout. The drawings are not necessarily to scale, emphasis instead being placed upon illustrating aspects of the present disclosure.

[0012] FIG. 1 illustrates a schematic diagram of an example environment in which a method of content generation can be implemented, according to certain embodiments of the present disclosure;

[0013] FIG. 2 illustrates a schematic diagram of a content distribution platform, according to certain embodiments of the present disclosure;

[0014] FIG. 3 illustrates a schematic diagram of a choreography robot, according to certain embodiments of the present disclosure;

[0015] FIG. 4 illustrates a flowchart of a process of content generation, according to certain embodiments of the present disclosure;

[0016] FIG. 5 illustrates a flowchart of a process for recalling BOTs and generating related tweets based on news, according to certain embodiments of the present disclosure;

[0017] FIG. 6 illustrates a schematic diagram of a process of recalling and filtering BOTs, according to certain embodiments of the present disclosure.

[0018] FIG. 7 illustrates a flowchart of a method of content generation, according to certain embodiments of the present disclosure;

[0019] FIG. 8 illustrates a block diagram of an apparatus of content generation, according to certain embodiments of the present disclosure; and

[0020] FIG. 9 illustrates a block diagram of a device of content generation, according to certain embodiments of the present disclosure.

[0021] The same or similar reference numerals are used in all the drawings to represent the same or similar elements. DETAILED DESCRIPTION

[0022] It can be understood that the data involved in the technical solution (for example, chat messages, and including but not limited to the data itself, acquisition or use of the data) should comply with the requirements of the corresponding laws and regulations and relevant provisions.

[0023] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms, and should not be interpreted as being limited to the embodiments set forth herein, but rather these embodiments are provided to more thoroughly and completely understand the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for exemplary purposes only, and are not intended to limit the scope of protection of the present disclosure.

[0024] In the description of embodiments of the present disclosure, the term "comprising" and similar terms are understood to encompass open-ended inclusion, i.e., "including but not limited to". The term "based on" is understood to mean "at least partially based on". The term "one embodiment" or "the embodiment" is understood to mean "at least one embodiment". The terms "first", "second", and the like can refer to different or the same objects. Other explicit and implicit definitions can also be included below.

[0025] New media operation is a professional responsible for planning, creating, publishing and promoting content on new media platforms. They create content to attract users, maintain platform interaction, analyze data to optimize strategies, and plan activities to enhance user engagement. In addition, new media operators can also be responsible for planning content strategies to attract users, collecting, editing and optimizing various content to ensure that the quality of the content meets the platform positioning.

[0026] Currently, new media operations mainly rely on manual writing of tweets. In this context, a tweet can refer to a short message on a social media platform that expresses opinions, shares information or sparks discussions in the form of concise text, images or links. Tweets are usually limited by the number of characters, such as 300 characters, requiring concise content. As a real-time communication tool, tweets have the characteristics of fast dissemination and strong interaction, and are widely used for information dissemination, opinion expression and marketing promotion. It is an important way for people to quickly communicate, share and access information in the Internet era.

[0027] Artificially writing tweets requires a lot of time and effort. In situations where a large amount of content needs to be generated quickly to respond to emergencies, hot topics, or promotional activities, manual writing may not meet the timeliness requirements. This may result in missing the best dissemination opportunity and affecting the dissemination effect. In addition, long-term reliance on manual writing of tweets may lead to creative exhaustion. Even the most talented copywriters may struggle to continuously produce novel and interesting content over a long period of time. This may result in a decline in content quality, affecting the reading experience of the audience and the brand image.

[0028] To solve the above-mentioned shortcomings, embodiments of the present disclosure provide a scheme for content generation. The scheme can select a suitable object from a plurality of objects for content generation for writing scripts and having different styles according to the characteristics of the source content. Thus, while generating tweets in a timely manner, the cost of manual work is reduced, and the styles of the tweets are different, thereby improving the user experience and to some extent improving the interest and reading experience of the news.

[0029] The present disclosure will be described below in conjunction with FIGS. 1-9. However, it should be understood that this is only to enable those of ordinary skill in the art to better understand the principles and ideas of the embodiments of the present disclosure, and is not intended to limit the scope of the present disclosure in any way.

[0030] FIG. 1 shows a schematic diagram of an example environment 100 in which a method of content generation according to certain embodiments of the present disclosure can be implemented. As shown in FIG. 1, the example environment 100 can include a large model 110. The large model 110 can also be referred to as a large language model, which can be a deep learning model in the field of artificial intelligence. The large model 110 is capable of capturing and learning complex patterns and relationships in data, thereby exhibiting excellent performance in multiple fields such as natural language processing, computer vision, etc.

[0031] For example, the large model 110 integrates a rich plugin system covering multiple fields such as information, tourism, office work, etc., supporting both built-in plugins and user customization, thereby enabling unlimited expansion of BOT capabilities. The knowledge base function of the large model 110 is diverse and supports data in multiple formats and sources, ensuring that the BOT can accurately answer user questions. The long-term memory function of the large model 110 allows the BOT to remember user preferences and key information in the conversation, providing personalized services. The timed task function can be set without programming, improving work efficiency. Users can use workflow design on the large model 110 to flexibly, efficiently, and easily build complex processes. The multi-agent (Agent) mode supports parallel or serial processing of multiple tasks to meet complex scenario requirements. The large model 110, with its comprehensive functions and outstanding advantages, helps users easily create intelligent and efficient BOT services.

[0032] The example environment 100 can further include a computing device 140. The computing device 120 can implement the method of content generation presented by the present disclosure. For example, the computing device 140 can have installed thereon a recall platform 142. The recall platform 142 can be a program or a software module. The recall platform 142 can be communicatively connected with the large model 110. The recall platform 142 can also be connected with a publishing platform 120 of content for sending the tweets generated by the recalled BOTs to the publishing platform 120. The publishing platform can be an application (APP) of, for example, news, social media, etc.

[0033] On the large model 110, a plurality of BOTs can be established. For example, a first BOT 112, a second BOT 114, a third BOT 116, and a fourth BOT 118, etc. It can be understood that the number of BOTs shown in FIG. 1 is only an example and is not a limitation. The large model 110 can have fewer or more BOTs. The first BOT 112, the second BOT 114, the third BOT 116, and the fourth BOT 118 can be arranged as BOTs with different styles, and the outputs of the BOTs have different styles of tweets.

[0034] The establishment of the BOT can be based on a prompt word. Writing a prompt word is an indispensable core link for building an intelligent BOT, which directly shapes the role positioning, interaction style, and function implementation method of the BOT. In the creation process, first, the identity setting of the BOT needs to be clarified, such as a news broadcaster, a data analysis expert, etc., and the matching reply style to ensure that each response of the BOT can reflect its unique personality and professionalism. The function and workflow of the BOT need to be described, which includes the coping strategies in different scenarios, which can be configured to call large-scale internal tools (such as a search news tool) and specific steps for handling specific tasks (such as data cleaning, analysis). Clear function description not only guides the behavior pattern of the BOT, but also enhances its response ability to user requests.

[0035] For complex scenarios, structured prompts are particularly crucial. They not only improve the readability and maintainability of prompts, but also enhance the constraints on BOT behavior, ensuring the accuracy and consistency of responses. In addition, writing prompts also needs to consider iterative optimization. With the actual application of BOT, according to user feedback and performance, continuously adjust and optimize the prompts, is a necessary way to improve the intelligence level of BOT and user experience. Through this process, BOT can more accurately understand user intent and give more expected responses, thus establishing a more natural and smooth interaction experience. It should be understood that the BOT mentioned in this paper can refer to a robot, an agent, or other forms that may appear in future large models to complete a specific task. Therefore, BOT and future possible implementations can be referred to as objects. In the following, BOT will be used as an example to describe the object.

[0036] The recall platform 142 can obtain source content 127. The source content 127 can be an article, such as an article about entertainment, literature, sports, etc., or multimedia content. Assuming that the source content 127 is an article about sports, the recall platform 142 can select the most suitable BOT for writing a tweet of the sports type from the first BOT 112, the second BOT 114, the third BOT 116, and the fourth BOT 118, for example, the first BOT 112. Then, the first BOT 112 can generate a tweet based on the source content 127 as the target content 126, which can also be included and sent to the publishing platform 120 after being reviewed.

[0037] In the display area of the publishing platform 120, the target content 126 and the source content 127 can be displayed. In some embodiments, the target content 126 can include the source content 127 and the related recommendation or summary information (which can be referred to as target information), and the target content 126 can be displayed in the card 122. In the display area of the publishing platform 120, multiple cards such as the card 122 and the card 124 can also be displayed simultaneously. The card 124 can display source content 129 different from the source content 127 and target content 128 different from the target content 126, and the target content 128 can include the source content 129 and the related recommendation or summary information (which can be referred to as target information), and the generated target content 128 is associated with the attributes of the BOT 112 itself (for example, related to the information of the prompt words of the BOT 112).

[0038] FIG. 2 illustrates a schematic diagram of a content distribution platform 200, according to certain embodiments of the present disclosure. FIG. 2 takes the display area 202 of the content distribution platform 200 running on a mobile device (e.g., a smartphone) as an example. In the display area 202, an application interface is presented that integrates news content of technology, entertainment, sports, etc. Specifically, in the top information bar 204, tabs of interest, discovery, and daily are included, and the tab of discovery is selected. These tabs can represent content that the user has personalized interest in, content that the platform recommends, and daily curated news or information.

[0039] Below the top information bar 204, since the tab of discovery is selected, tabs of technology 206, entertainment 208, and sports 210 are displayed, and the tab of entertainment 208 is selected. It can be appreciated that the user can also select to view other tabs according to interests. These tabs can belong to sub-tabs under discovery. If the tab of interest is selected, the content here will change accordingly. In the middle and lower part of the display area 202, two cards are displayed, which are card 212 and card 224. Both card 212 and card 224 correspond to the category of entertainment 208, and their content belongs to entertainment information.

[0040] As an example, the upper part of card 212 includes an avatar 214, and to the right of the avatar 214, the name ABC of the user who published the card 212 is displayed, and the publishing time of the card is 1 hour ago, which enhances the timeliness and attractiveness of the information. To the right of the card 212, a button of interest 216 is displayed, which can be clicked by other users to follow the user who published the card. At the bottom of the card 212, an article 220 can be displayed, and to the left of the article 220, a representative thumbnail 222 of the article 220 can be displayed to make the article 220 more attractive. In the middle of the card, a comment 218 on the article 220 can be displayed, which can be a summary, a recommendation, or a further creation based on the article 220, etc.

[0041] Below card 212, other similar cards, such as card 224, can also be displayed. Since the publishing time of card 224 is 1 day ago, it is displayed below card 212. The display area 202 can also display more cards, which can depend on the size of the screen, user settings, etc.

[0042] FIG. 3 shows a schematic diagram 300 of a bot orchestration according to certain embodiments of the present disclosure. In a display area 302 for bot orchestration, the name of the bot can be edited at 304, for example, the name of the bot can be ABC (hereinafter also referred to as BOT 304). At 304, there can be buttons embodying the current operation function, for example, the current operation is orchestration. After the orchestration button, an area 308 for the prompt words of the bot is displayed. In the area 308, the style and reply logic can be created for the bot. Specifically, a character setting (abbreviated as person setting) can be created for the bot. The person setting is a setting of an individual or a character image, personality, and characteristics, which can attract attention through a specific image and behavior mode. The person setting can be determined by setting the prompt words of the role, character, requirements, and format, etc. for the bot. For example, the BOT 304 can be set as a poet. Then, the input prompt word can be set as: “You are a literary youth who likes poetry”. Similarly, the task of the BOT 304 can be set through the prompt word: “Write a script for a news of film and television entertainment, and the script can be used to forward the news”. The following requirements can be made for the written script: “1. The number of words is not more than XX; 2. The main content of the news can be summarized; 3. The poem is used as the end of the script”. The format of the written script can be required, for example, “Format requirement: summarize the main content of the news, and fill in the created poem after “Poem says:”. Through these prompt words, a number of robots with different styles can be established.

[0043] In subsequent use, the prompt words can also be modified to optimize the effect of the output script. Optimizing the prompt words of the robot can make the prompt words clear and direct, directly expressing the specific requirements of writing the script task. At the same time, the prompt words can be structured as much as possible, and the structured prompt words can help the robot better understand the task and reduce misunderstandings by separating different parts (such as role, target, and limitation condition). Emphasizing important information, using emphasis words and priority ranking, can ensure that the robot pays attention to and prioritizes key elements. Providing context and background information is also an important part of optimizing the prompt words, which helps the robot generate more practical responses according to the context. In addition, the introduction of examples and cases can provide intuitive references for the robot, making its output more in line with user expectations. Throughout the process, it is crucial to continuously collect user feedback and adjust the prompt words accordingly. Through continuous iteration and optimization, the accuracy and efficiency of the robot can be gradually improved, making it better serve the needs of users.

[0044] FIG. 4 shows a flowchart of a process 400 of content generation according to certain embodiments of the present disclosure. The process 400 can be jointly completed by different execution subjects, such as the large model 110, the recall platform 142, and the publishing platform 120 shown in FIG. 1. The process 400 realizes the intelligentization and personalization of news tweet generation by deeply fusing the recall platform and the large model. The process 400 first uses a two-level recall strategy to accurately extract labels from news, and matches some basic candidate robots that meet the requirements. Then, the language ability of the large model is used to filter target robots from the vector library representing the candidate robots. These candidate robots meet the basic types of target news, but also have their own characteristics and can cover various styles and language patterns. Among them, the process 400 uses the large model to evaluate these candidate robots to ensure that the selected target robots can most closely match the news content, thereby generating tweets that are not only accurate but also attractive. This process not only improves the efficiency of tweet generation, but also ensures the stability of tweet quality. In the tweet generation stage, the selected robots present the news to the readers in the corresponding form according to their unique style and language characteristics. These tweets not only cover the key elements of the news, but also incorporate personalized interpretation and creative expression of the robots, making each tweet fresh and interesting.

[0045] In some embodiments, the production system 410 can be a news production system 410 or also referred to as an article production system 410. Specifically, the production system 410 can use artificial intelligence technology to automatically capture and process news data through deep learning, natural language processing, and other technologies, and generate news reports or summaries by imitating the writing style of human reporters. Therefore, the production system 410 significantly improves the efficiency and accuracy of news production, and realizes the full automation from data collection to news publishing. It brings more convenience and innovation to the content creation and dissemination of news, and provides users with more rich and personalized news services.

[0046] The news generated by the production system 410 can be executed for label extraction 412. For example, automatically identifying and extracting key information or category labels from content data. These labels can be predefined categories (such as “sports”, “entertainment” in news classification), entities mentioned in the content (such as names, place names, organization names, etc.), sentiment orientations (such as positive, negative, neutral), or any other metadata that helps understand and organize content. For example, the extracted labels can be: technology, new product, mobile phone, release. Vectors representing their semantics can be generated based on the extracted labels.

[0047] On the other hand, in the large model platform 402, a plurality of BOTs with different personas can be established, and a BOT set 404 can be established. The BOT set 404 can also be registered in the publishing platform 406. In this way, the publishing platform 406 can know which BOTs there are, and for each of the plurality of BOTs, a corresponding vector can be generated based on its corresponding prompt word, which embodies the semantic characteristics of the prompt word of the BOT. The vector set of the BOT set 404 can be maintained in the vector library 408.

[0048] In the vector recall 414, the vector generated by labeling the news extraction can be matched with the vectors in the vector library 408 to obtain some matched vectors. A number of BOTs corresponding to these matched vectors can be determined as candidate BOTs. These candidate BOTs can be filtered again (416) via the large model platform 402. The large model platform 402 can use its powerful language and reasoning capabilities to filter the most suitable BOT for the news from the plurality of candidate BOTs. For example, continuing the above example, the news is about the release of a mobile phone. Then, if there are two persona BOTs in the plurality of candidate BOTs, which are a technology blogger and a lifestyle blogger, respectively, the technology blogger can be selected as the final target BOT. Because the technology blogger is more suitable to write tweets about the release of a mobile phone, and the lifestyle blogger has a wider field than the technology blogger, so it lacks specificity. The target BOT filtered can write (418) a tweet according to the news. The tweet can be reviewed (420). If the review is passed, the tweet can be published (422), for example, to a news APP or the like.

[0049] FIG. 5 shows a flowchart of a process 500 for recalling BOTs and generating related tweets based on news, according to certain embodiments of the present disclosure. In the large model 502, eight BOTs have been established, which are BOT 504, BOT 506, BOT 508, BOT 510, BOT 512, BOT 514, BOT 516, and BOT 518. BOT 504 can focus on artificial intelligence technology. BOT 506 can focus on Chinese literature. BOT 508 can focus on entertainment and poetry. BOT 510 can focus on film and television entertainment and poetry. BOT 512 can focus on football. BOT 514 can focus on overseas technology. BOT 516 can focus on global movies. BOT 518 can focus on software and programming. These BOTs can be registered in the content platform 532. The content platform 532 can be responsible for generating articles and tweets to the news platform 530. The content platform 532 can know what style of BOT there is.

[0050] The news scraping module 522 can obtain various news reports on the same event from various sources, and the article generation module generates an article on the news. Since the article is generated based on the news from various angles, it is more readable and has higher quality. Based on the article, a number of BOTs that meet the article can be recalled in the recall BOT module 526. The number of recalled BOTs can be set as needed. In some embodiments, since the information of the BOTs 504-518 has been registered in the content platform 532, the recall process can also not communicate with the large model 502. As an example, the recalled BOTs can be, for example, BOT 504 and BOT 514.

[0051] After recalling the BOTs 504 and 514, large-scale fine filtering can be performed again. For example, assume that the news here is news of artificial intelligence technology overseas. The BOT 504 can focus on the technology of the artificial intelligence category. The BOT 514 can focus on the technology overseas. Then, relying only on the recall technique of label matching / vector matching, it can not be possible to select which BOT is the most suitable BOT. The large model 502 can use its language ability and reasoning ability to fine filter that the BOT 504 is the most suitable target BOT, i.e., the target. Therefore, the target BOT 520 obtained after filtering can be the BOT 504. The target BOT 504 can write (528) a tweet based on the news, and send it to the news platform 530 after review.

[0052] FIG. 6 shows a schematic diagram of a process 600 of recalling and filtering BOTs according to certain embodiments of the present disclosure. The starting point of the flow 600 is the source content 602. The source content 602 can include multiple channels derived from news reports, social media, etc. The source content 602 can be tagged (604), in which natural language processing (NLP) techniques and machine learning algorithms can be used to automatically identify and extract key information or features from the source content 602, forming a series of tags. These tags can include topic words, keywords, sentiment orientations, etc., which provide an important basis for subsequent BOT recall and classification. Through tag extraction, the system can more accurately understand the content and lay a solid foundation for subsequent processing. For example, based on the source content 602, the tags 608, 610, and 612 can be extracted. A number of BOTs can be created in the large model 606, such as BOT 616, BOT 618, BOT 620, BOT 622, BOT 624, and more BOTs. These BOTs all have their own personas.

[0053] Based on the label 608, the label 610, and the label 612, a plurality of BOTs that match the labels can be recalled (614) as candidate BOTs, such as BOT 616, BOT 618, and BOT 620. A predetermined number of candidate BOTs can be recalled from the BOT 616, the BOT 618, the BOT 620, the BOT 622, and the BOT 624 according to the similarity of the label 608, the label 610, and the label 612 to the feature vectors of the respective BOTs (which can be generated based on the prompt words). The similarity of the news label to the respective BOT vectors can be calculated using methods such as cosine similarity, Euclidean distance, etc., and a predetermined number of BOTs with the highest ranking are selected.

[0054] The large model can deeply analyze and evaluate the matching degree between the source content 602 and the candidate BOT 616, the BOT 618, and the BOT 620, rather than just based on the label 608, the label 610, and the label 612. For example, the large model uses previously extracted labels and source content information to conduct in-depth analysis and evaluation. This process can involve various NLP tasks such as classification, sentiment analysis, entity recognition, etc., to comprehensively interpret the source content 602 from multiple dimensions. Through the processing of the large model, the system can more accurately determine the relevance between the content and the BOT, providing strong support for filtering. After filtering, a final BOT is selected to perform the final processing on the source content 602. For example, the BOT 618 is selected as the target BOT.

[0055] Therefore, a two-layer recall strategy is achieved to label extract news. A batch of candidate BOTs are recalled from the BOT vector library through the label, and the most suitable BOT is selected from the candidate BOTs through the large model to be responsible for the writing of the tweet. This can improve the matching degree of the BOT and the news, because different BOTs have their own focus points, thereby improving the effect of BOT writing tweets and improving user experience. In addition, it can also reduce the labor cost of generating tweets.

[0056] FIG. 7 shows a flowchart of a method 700 of content generation according to certain embodiments of the present disclosure. The method 700 can be performed by one or more of the recall model, the large model, the publishing platform. At 702, based on at least one label associated with source content, at least one candidate object matching the source content is selected from a plurality of objects for content generation. At 704, by a preset model (e.g., a large model), a target object is selected from the at least one candidate object based on the source content. At 706, by the target object, target content is generated based on the source content, wherein the target content includes target information associated with the source content, and the target information is associated with an attribute of the target object.

[0057] By implementing the method 700, the uniqueness of the source content can be utilized to select the most suitable object from a collection of variously-styled objects for content generation to create a tweet. The method 700 not only ensures that the tweet can be generated quickly, effectively reducing the cost of manual intervention, but also greatly enriches the user's reading experience due to the diversity of tweet styles. The unique style brought by each object not only enhances the interest of the news, but also enhances the freshness and appeal during the reading process, thereby improving the overall news reading experience to some extent.

[0058] FIG. 8 shows a block diagram of an apparatus 800 for content generation according to certain embodiments of the present disclosure. As shown in FIG. 8, the apparatus 800 includes a first selection module 802 configured to select at least one candidate object matching the source content from a plurality of objects for content generation based on at least one label associated with the source content. The apparatus 800 further includes a second selection module 804 configured to select a target object from the at least one candidate object based on the source content by a preset model. The apparatus 800 further includes a content generation module 806 configured to generate target content based on the source content by the target object, the target content including target information related to the source content, the target information being associated with an attribute of the target object. The apparatus 800 can further include other modules to implement the functions as the method 700, which will not be described herein for brevity.

[0059] It can be understood that by the apparatus 800 of the present disclosure, at least one of the many advantages that can be achieved by the method or process described above can be achieved. For example, among a large number of objects for content generation that are good at copywriting and have different styles, the most suitable object for generating a tweet is intelligently selected. This method not only significantly reduces the burden of manual operation, but also brings users a more colorful reading experience through the diversity of tweet styles. Each unique style of tweet adds readability to the news, thereby improving the overall reading experience.

[0060] FIG. 9 shows a block diagram of an instant messaging presentation device 900 according to certain embodiments of the present disclosure, which can be a device or apparatus described in embodiments of the present disclosure. As shown in FIG. 9, the device 900 includes a central processing unit (CPU) 901, which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) 902 or loaded into a random access memory (RAM) 903 from a storage unit 908. Various programs and data required for operation of the device 900 can also be stored in the RAM 903. The CPU 901, the ROM 902, and the RAM 903 are connected to each other through a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904. Although not shown in FIG. 9, the device 900 can also include a coprocessor.

[0061] Various components in the device 900 are connected to the I / O interface 905, including an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a magnetic disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunication networks.

[0062] The various methods or processes described above can be performed by the CPU 901. For example, in some embodiments, the methods can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 908. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the CPU 901, one or more steps or actions in the methods or processes described above can be performed.

[0063] In some embodiments, the methods and processes described above can be implemented as a computer program product. The computer program product can include a computer readable storage medium having computer readable program instructions embodied therewith.

[0064] A computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. A non-exhaustive list of more specific examples of the computer readable storage medium includes the following: 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), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or raised structures in a groove having instructions recorded thereon, and any suitable combination of the foregoing. A computer readable storage medium, as used herein, is not to be construed as being transitory signals per se, such as radio waves or other freely propagating electromagnetic waves, electromagnetic waves propagating through a waveguide or other transmission media (e.g., light pulses passing through a fiber-optic cable), or electrical signals transmitted through a wire.

[0065] Computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0066] Computer readable program instructions for carrying out operations of the present disclosure can be assembly instructions, instruction set architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including object oriented programming languages and conventional procedural programming languages. The computer readable program instructions 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). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0067] These computer readable program instructions can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks. These computer readable program instructions can also be stored in a computer readable storage medium that can include a non-transitory computer readable storage medium that can direct a computer, a programmable data processing apparatus, and / or other devices to function in a particular manner, such that the computer readable storage medium having instructions stored therein comprises an article of manufacture including instructions which implement aspects of the function / act specified in the flowchart and / or block diagram block or blocks.

[0068] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device 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 flowchart and / or block diagram block or blocks.

[0069] The computer program product of the first aspect can include a computer readable storage medium. The computer readable storage medium can include instructions executable by a processor to receive a request for content from a client device. The computer readable storage medium can include instructions executable by the processor to select, based on at least one tag associated with source content, at least one candidate object from a plurality of objects for content generation that matches the source content. The computer readable storage medium can include instructions executable by the processor to select, by a pre-set model based on the source content, a target object from the at least one candidate object. The computer readable storage medium can include instructions executable by the processor to generate, by the target object based on the source content, target content. The target content can include target information associated with the source content. The target information can be associated with a property of the target object.

[0070] Embodiments of the present disclosure have been described above, with the understanding that these embodiments are exemplary only, and are not restrictive, and are not limited to the disclosed embodiments. Many modifications and changes to the described embodiments are possible by those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of words in this document is intended to best explain the principles of the embodiments, practical application, or technical improvement to the art in the marketplace, or to enable other ordinary skilled persons in the art to understand the embodiments disclosed herein.

[0071] Some example implementations of the present disclosure are listed below.

[0072] Example 1. A method of content generation, comprising:

[0073] selecting, based on at least one tag associated with source content, at least one candidate object from a plurality of objects for content generation that matches the source content;

[0074] selecting, by a pre-set model based on the source content, a target object from the at least one candidate object; and

[0075] generating, by the target object based on the source content, target content. The target content can include target information associated with the source content. The target information can be associated with a property of the target object.

[0076] Example 2. The method of example 1, wherein the source content is news, and the method further comprises, prior to selecting, based on at least one tag associated with source content, at least one candidate object from a plurality of objects for content generation that matches the source content:

[0077] obtaining the source content, wherein the source content is generated based on news associated with the same event obtained from at least one platform; and

[0078] determining the at least one tag associated with the source content.

[0079] Example 3. The method of any of examples 1-2, wherein selecting, based on the at least one tag associated with the source content, at least one candidate object matching the source content from a plurality of objects for content generation comprises:

[0080] extracting a plurality of tags from the source content based on an understanding of the source content;

[0081] generating a first set of vectors associated with the plurality of tags;

[0082] determining, based on a second set of vectors associated with the plurality of objects, a predetermined number of the candidate objects matching the plurality of vectors.

[0083] Example 4. The method of any of examples 1-3, wherein determining, based on a second set of vectors associated with the plurality of objects, a predetermined number of the candidate objects matching the plurality of vectors comprises:

[0084] obtaining a plurality of prompt words of the plurality of objects from the pre-set model;

[0085] generating the second set of vectors based on the plurality of prompt words of the plurality of objects; and

[0086] determining the candidate objects based on the first set of vectors and the second set of vectors.

[0087] Example 5. The method of any of examples 1-4, wherein the prompt words of the plurality of objects comprise at least one of:

[0088] a target role;

[0089] reply logic;

[0090] a language style;

[0091] a target task; and

[0092] a target format.

[0093] Example 6. The method of any one of examples 1-5, wherein the target object is determined by the preset model based on filtering of semantics associated with the plurality of tags, and the filtering comprises: determining, based on a set of semantic similarities between the source content and a prompt word of each of the plurality of candidate objects determined by the preset model, a candidate object having a highest semantic similarity among the plurality of candidate objects as the target object based on the set of semantic similarities.

[0094] Example 7. The method of any one of examples 1-6, the source content is first source content of a first type, and the method further comprises:

[0095] obtaining second source content of a second type, wherein the first type is different from the second type; and

[0096] generating, by the preset model, target content with the target object corresponding to the second source content and based on the second source content.

[0097] Example 8. The method of any one of examples 1-7, wherein prior to generating target content by the target object based on source content, the method further comprises:

[0098] sending, to the preset model, an instruction for generating the target content with the target object.

[0099] Example 9. The method of any one of examples 1-8, wherein generating the target content by the target object based on the source content comprises:

[0100] generating, by the target object, the target content in accordance with a style indicated by a prompt word associated with the target object and in the target format,

[0101] and the method further comprises:

[0102] outputting the target content.

[0103] Example 10. The method of any one of examples 1-10, further comprising:

[0104] sending the target content to a publishing platform for publishing of the target content and the source content.

[0105] Example 11. An apparatus for content generation, comprising:

[0106] a first selection module configured to select, based on at least one tag associated with source content, at least one candidate object matching the source content from a plurality of objects for content generation;

[0107] a second selection module configured to select, by a preset model, a target object from the at least one candidate object based on the source content; and

[0108] a content generation module configured to generate, by the target object, target content based on the source content, the target content comprising target information related to the source content, the target information being associated with an attribute of the target object.

[0109] Example 12. The apparatus of Example 11, wherein the source content is news, and the apparatus further comprises: a first acquisition module configured to, before selecting, based on at least one label associated with the source content, at least one candidate object matching the source content from a plurality of objects for content generation: acquire the source content, wherein the source content is generated based on news associated with a same event acquired from at least one platform; and

[0110] a first determination module configured to determine the at least one label associated with the source content.

[0111] Example 13. The apparatus of any one of Examples 11-12, wherein the first selection module comprises:

[0112] a first extraction module configured to extract a plurality of labels from the source content based on an understanding of the source content;

[0113] a first generation module configured to generate a first vector set associated with the plurality of labels;

[0114] a second determination module configured to determine, based on a second vector set associated with the plurality of objects, a predetermined number of the candidate objects matching the plurality of vectors.

[0115] Example 14. The apparatus of any one of Examples 11-13, wherein the second determination module comprises:

[0116] a second acquisition module configured to acquire, from the preset model, a plurality of prompt words of the plurality of objects;

[0117] a second generation module configured to generate the second vector set based on the plurality of prompt words of the plurality of objects; and

[0118] a third determination module configured to determine the candidate objects based on the first vector set and the second vector set.

[0119] Example 15. The apparatus of any one of Examples 11-14, wherein the prompt words of the plurality of objects comprise at least one of:

[0120] a target role;

[0121] reply logic;

[0122] language style;

[0123] target task; and

[0124] target format.

[0125] Example 16. The apparatus according to any one of examples 11-15, further comprising a filtering module configured to determine, based on a set of semantic similarities between the source content and an utterance of each of the plurality of candidate objects determined by the preset model, a candidate object having a highest semantic similarity among the plurality of candidate objects as the target object based on the set of semantic similarities.

[0126] Example 17. The apparatus according to any one of examples 11-16, the source content is a first source content of a first type, and the apparatus further comprises:

[0127] a third obtaining module configured to obtain a second source content of a second type, wherein the first type is different from the second type; and

[0128] a third generating module configured to generate, by the target object and based on the second source content, target content using the target object corresponding to the second source content.

[0129] Example 18. The apparatus according to any one of examples 11-17, the apparatus further comprising a sending module configured to send, to the preset model, an instruction for generating the target content using the target object, before the target content is generated by the target object based on the source content.

[0130] Example 19. The apparatus according to any one of examples 11-18, wherein the content generating module comprises:

[0131] a fourth generating module configured to generate, by the target object and based on an utterance associated with the target object, the target content in a style indicated by the utterance and in the target format,

[0132] and the apparatus further comprises:

[0133] an output module configured to output the target content.

[0134] Example 20. The apparatus according to any one of examples 11-19, the apparatus further comprising:

[0135] a second sending module configured to send the target content to a publishing platform for publishing of the target content and the source content.

[0136] Example 21. An electronic device, comprising:

[0137] a processor; and

[0138] a memory coupled with the processor, the memory having stored therein instructions that, when executed by the processor, cause the electronic device to perform acts comprising:

[0139] selecting, based on at least one tag associated with source content, at least one candidate object matching the source content from a plurality of objects for content generation;

[0140] selecting, by a preset model, a target object from the at least one candidate object based on the source content; and

[0141] generating, by the target object, target content based on the source content, wherein the target content comprises target information associated with the source content, the target information being associated with a property of a target object.

[0142] Example 22. The electronic device of example 21, wherein the source content is news, and the method further comprises, prior to selecting, based on at least one tag associated with source content, at least one candidate object matching the source content from a plurality of objects for content generation:

[0143] obtaining the source content, wherein the source content is generated based on news associated with a same event obtained from at least one platform; and

[0144] determining the at least one tag associated with source content.

[0145] Example 23. The electronic device of any of examples 21-22, wherein selecting, based on at least one tag associated with source content, at least one candidate object matching the source content from a plurality of objects for content generation comprises:

[0146] extracting a plurality of tags from the source content based on an understanding of the source content;

[0147] generating a first set of vectors associated with the plurality of tags;

[0148] determining, based on a second set of vectors associated with the plurality of objects, a predetermined number of the candidate objects matching the plurality of vectors.

[0149] Example 24. The electronic device of any of examples 21-23, wherein determining, based on a second set of vectors associated with the plurality of objects, a predetermined number of the candidate objects that match the plurality of vectors comprises:

[0150] obtaining, from the preset model, a plurality of prompt words for the plurality of objects;

[0151] generating the second set of vectors based on the plurality of prompt words for the plurality of objects; and

[0152] determining the candidate objects based on the first set of vectors and the second set of vectors.

[0153] Example 25. The electronic device of any of examples 21-24, wherein the prompt words for the plurality of objects comprise at least one of:

[0154] a target role;

[0155] reply logic;

[0156] a language style;

[0157] a target task; and

[0158] a target format.

[0159] Example 26. The electronic device of any of examples 21-25, wherein the target object is determined by the preset model based on filtering of semantics associated with the plurality of labels, and the filtering comprises: determining, based on a set of semantic similarities between the source content and prompt words for each of the plurality of candidate objects determined by the preset model, a candidate object having a highest semantic similarity among the plurality of candidate objects as the target object based on the set of semantic similarities.

[0160] Example 27. The electronic device of any of examples 21-26, the source content is a first source content of a first type, and the actions further comprise:

[0161] obtaining a second source content of a second type, wherein the first type is different from the second type; and

[0162] generating, by the preset model, target content with a target object corresponding to the second source content and based on the second source content.

[0163] Example 28. The electronic device of any of examples 21-27, wherein prior to generating target content based on source content by the target object, the actions further comprise:

[0164] sending, to the preset model, an instruction for generating the target content by using the target object.

[0165] Example 29. The electronic device of any of examples 21-28, wherein generating, by the target object, the target content based on the source content comprises:

[0166] generating, by the target object, the target content in accordance with the target format based on a prompt word associated with the target object and in a style indicated by the prompt word,

[0167] and the actions further comprise:

[0168] outputting the target content.

[0169] Example 30. The electronic device of any of examples 21-29, the actions further comprising:

[0170] sending the target content to a publishing platform for publishing of the target content and the source content.

[0171] Example 31. A computer-readable storage medium having stored thereon one or more computer instructions, wherein the one or more computer instructions are executed by a processor to implement the method according to any of examples 1-10.

[0172] Example 32. A computer program product tangibly stored in a computer readable medium and comprising computer executable instructions that, when executed by a device, cause the device to perform the method according to any of examples 1-10.

[0173] Although the present disclosure has been described in some detail with specific reference to structure features and / or methodological acts, it is understood that the subject matter defined in the appended claims is not necessarily limited to the particular features or acts described above. Rather, the particular features and acts described above are merely examples of implementing the claims.

Claims

1. A method of content generation, comprising: selecting, based on at least one label associated with source content, at least one candidate object matching the source content from a plurality of objects for content generation; selecting, by a preset model, a target object from the at least one candidate object based on the source content; and generating, by the target object, target content based on the source content, wherein the target content comprises target information associated with the source content, the target information associated with attributes of the target object. Before selecting, based on at least one label associated with source content, at least one candidate object matching the source content from a plurality of objects for content generation:

2. The method of claim 1, wherein the source content is news, and the method further comprises: obtaining the source content, wherein the source content is generated based on news associated with a same event obtained from at least one platform; and determining the at least one label associated with the source content.

3. The method of claim 1, wherein selecting, based on at least one label associated with source content, at least one candidate object matching the source content from a plurality of objects for content generation comprises: extracting a plurality of labels from the source content based on an understanding of the source content; generating a first set of vectors associated with the plurality of labels; determining a predetermined number of the candidate objects matching the plurality of vectors based on a second set of vectors associated with the plurality of objects.

4. The method of claim 3, wherein determining a predetermined number of the candidate objects matching the plurality of vectors based on a second set of vectors associated with the plurality of objects comprises: obtaining a plurality of prompt words of the plurality of objects from the preset model; generating the second set of vectors based on the plurality of prompt words of the plurality of objects; and determining the candidate objects based on the first set of vectors and the second set of vectors.

5. The method of claim 4, wherein the prompt words of the plurality of objects comprise at least one of: a target role; a reply logic; a language style; a target task; and a target format. determining, based on a set of semantic similarities between the source content and prompt words of each of the plurality of candidate objects determined by the preset model, a candidate object having a highest semantic similarity among the plurality of candidate objects as the target object based on the set of semantic similarities.

7. The method of claim 1, the source content is a first source content of a first type, and the method further comprises: 6.The method of claim 1, wherein the target object is determined by the preset model based on filtering of semantics associated with the plurality of tags, and the filtering comprises: obtaining a second source content of a second type, wherein the first type is different from the second type; and generating, by the preset model, target content utilizing a target object corresponding to the second source content and based on the second source content.

8. The method of claim 1, wherein before generating, by the target object, target content based on source content, the method further comprises: sending, to the preset model, an instruction for utilizing the target object to generate the target content.

9. The method of claim 1, wherein generating, by the target object, target content based on source content comprises: ​ ​ ​ generate, by the target object, the target content in accordance with the target format based on a prompt word associated with the target object and in a style indicated by the prompt word, and the method further comprises: outputting the target content.

10. The method of claim 1, further comprising: sending the target content to a publishing platform for publishing of the target content and the source content.

11. An apparatus of content generation, comprising: a first selection module configured to select, from a plurality of objects for content generation, at least one candidate object matching a source content based on at least one label associated with the source content; a second selection module configured to select, by a preset model, a target object from the at least one candidate object based on the source content; and a content generation module configured to generate, by the target object, a target content based on the source content, the target content comprising target information related to the source content, the target information being associated with an attribute of the target object.

12. An electronic device, comprising: a processor; and a memory coupled with the processor, the memory having stored therein instructions that, when executed by the processor, cause the electronic device to perform the method of any one of claims 1-10.

13. A computer-readable storage medium having computer-executable instructions stored therein, wherein the computer-executable instructions, when executed by a processor, perform the method of any one of claims 1-10.

14. A computer program product tangibly stored on a computer-readable storage medium and comprising computer-executable instructions that, when executed by a device, cause the device to perform the method of any one of claims 1-10. ​

Images