Method, device and system for real-time news content generation and avatar-based reporting based on artificial intelligence

Abstract

According to one embodiment, a method for automatically generating real-time news content based on artificial intelligence and providing avatar reporting, performed by a device, comprises: receiving a request for provision of news content from a first user terminal; collecting news-related content newly registered in the external data source during a preset first period from an external data source as a first news target set; extracting text included in the first news target set as first text data; encoding the first text data to generate a first input signal; inputting the first input signal to a first artificial intelligence model trained to generate a script of news content by analyzing keywords, context, or importance of the text; obtaining a first output signal representing a first script generated as an analysis result of the first text data from the first artificial intelligence model; converting the first script into voice data based on the first output signal to generate first voice data; and generating or calling a first avatar, which is a digital human avatar for delivering news content. A method for artificial intelligence-based real-time news content automatic generation and avatar reporting provision is provided, comprising: generating first reporting video data including mouth shape synchronization, facial expression change, and gesture of the first avatar based on the first voice data and the first script; setting a format representing the output format of a news content provision target platform to a first format based on a request for provision of the news content; generating first news content by converting the first reporting video data according to the first format; and transmitting the first news content to the first user terminal.

Description

Technology Field

[0001] The following embodiments relate to technology for artificial intelligence-based real-time news content automatic generation and avatar reporting provision. Background Technology

[0002] With the recent advancement of the digital media environment, the demand for consuming news content in real time through various platforms is rapidly increasing. In particular, due to the proliferation of mobile devices, social media, and short-form video platforms, users are increasingly preferring intuitive and immersive video-based content over text-based news. Consequently, there is a continuously growing demand for technology to rapidly generate news content and provide it in various formats.

[0003] However, conventional news content production methods have the problem of requiring a significant amount of time and manpower, as reporters or producers must sequentially perform processes such as collecting news data, writing articles, editing videos, recording audio, and inserting subtitles based on it. Furthermore, in the case of video-based news, production costs increase due to the need for filming equipment, studio environments, anchors, and editing personnel, and there are limitations in responding immediately to real-time issues.

[0004] Meanwhile, although technologies utilizing artificial intelligence to summarize text or convert it into speech have recently been partially introduced, they often remain at the level of simple text processing or speech conversion, posing limitations in automatically generating reporting content similar to actual broadcasts. In particular, technologies that comprehensively process text analysis, script generation, speech synthesis, digital human-based facial expression and gesture generation, and format conversion suitable for various platforms have not yet been fully implemented.

[0005] Furthermore, despite the fact that the content formats required by each platform differ, conventional technology lacks the functionality to automatically convert and provide the same content optimized for various platforms, and there is also a problem in that keyword-based customization functions for providing user-tailored news are provided in a limited manner.

[0006] Therefore, there is a need to develop technology capable of automatically generating news scripts using artificial intelligence based on news-related information collected from external data sources, automatically creating reporting videos including voice, facial expressions, and gestures through digital human avatars, and converting them into formats suitable for various platforms to provide them in real time. Prior art literature

[0007] Korean Registered Patent No. 10-2934528, Korean Published Patent No. 10-2026-0024888, Korean Published Patent No. 10-2025-0153102, Korean Published Patent No. 10-2024-0171923 The problem to be solved

[0008] According to one embodiment, the purpose is to provide a method, device, and system for automatically generating real-time news content and providing avatar reporting based on artificial intelligence.

[0009] The objectives of the present invention are not limited to those mentioned above, and other unmentioned objectives will be clearly understood from the description below. means of solving the problem

[0010] According to one embodiment, a method for automatically generating real-time news content based on artificial intelligence and providing avatar reporting, performed by a device, comprises: receiving a request for provision of news content from a first user terminal; collecting news-related content newly registered in the external data source during a preset first period from an external data source as a first news target set; extracting text included in the first news target set as first text data; encoding the first text data to generate a first input signal; inputting the first input signal to a first artificial intelligence model trained to generate a script of news content by analyzing keywords, context, or importance of the text; obtaining a first output signal representing a first script generated as an analysis result of the first text data from the first artificial intelligence model; converting the first script into voice data based on the first output signal to generate first voice data; and generating or calling a first avatar, which is a digital human avatar for delivering news content. A method for artificial intelligence-based real-time news content automatic generation and avatar reporting provision is provided, comprising: generating first reporting video data including mouth shape synchronization, facial expression change, and gesture of the first avatar based on the first voice data and the first script; setting a format representing the output format of a news content provision target platform to a first format based on a request for provision of the news content; generating first news content by converting the first reporting video data according to the first format; and transmitting the first news content to the first user terminal.

[0011] The above method for automatically generating real-time news content based on artificial intelligence and providing avatar reporting may further include: a step of, after collecting the first news target set, if a request for providing the news content is confirmed to be a request including a first keyword, a step of classifying data determined to be relevant based on thematic relevance, semantic similarity, or keyword match with the first keyword among the first news target set into a second news target set; and a step of selecting the second news target set from the first news target set and excluding the remaining data, and, after the step of generating the first news content, if the first news content includes a plurality of news segments, a step of sorting the plurality of news segments according to the degree of relevance or importance with the first keyword.

[0012] The above method for providing AI-based real-time news content automatic generation and avatar reporting comprises: a step of dividing the first script into multiple constituent units according to a news topic after the step of acquiring the first output signal; a step of selecting a first constituent unit, which is one of the multiple constituent units, as an analysis target; a step of confirming the number of characters included in the first constituent unit as a first count; a step of confirming the news topic of the first constituent unit as a first topic; a step of confirming the first user's preference for the first topic as a first index based on the first user's preference by topic; a step of calculating the difference between the maximum value of the reference range and the first index as a first difference value if the first index is confirmed to be higher than the maximum value of a preset reference range; a step of setting a first weight to a higher value within the upper section of a preset weight range as the first difference value is higher; a step of calculating a second count by multiplying the first count and the first weight; and a step of detailing the content of the first constituent unit to expand from the first count of characters to the second count of characters. If it is confirmed that the first index is lower than the minimum value of the reference range, the method may further include the step of calculating the difference between the first index and the minimum value of the reference range as a second difference value; the higher the second difference value, the lower the second weight value is set within the lower section of the weight range; the step of calculating a third number by multiplying the first number and the second weight value; and the step of simplifying the content of the first constituent unit to reduce it from the first number of characters to the third number of characters. Effects of the invention

[0013] According to one embodiment, news-related text collected from an external data source is analyzed through an artificial intelligence model to automatically generate a news script, and by converting it into a reporting video based on voice data and a digital human avatar, news content can be quickly generated and provided without a separate filming and editing process.

[0014] In addition, according to one embodiment, by integrally performing text analysis, script generation, speech synthesis, and avatar-based video generation processes, production time and costs can be reduced compared to existing human-centered news production methods, and the speed of response to real-time issues can be improved.

[0015] In addition, according to one embodiment, by generating reporting video data including mouth shape synchronization, facial expression changes, and gestures of the first avatar, immersive video-based news content can be provided to the user, and delivery capabilities similar to those of a real anchor can be secured.

[0016] In addition, according to one embodiment, by setting a format corresponding to a platform for providing news content and converting reporting video data accordingly, content can be automatically provided in a form optimized for various platform environments such as short-form, mobile, and metaverse.

[0017] In addition, according to one embodiment, by generating and providing news content based on user requests or keywords, it is possible to provide a user-customized news service and expand the scope of content utilization.

[0018] In addition, according to one embodiment, advertising or sponsor elements can be combined with automatically generated news content, which has the effect of expanding into various business models such as digital newsroom-based SaaS services, brand-specific reporters, and content commercialization.

[0019] Meanwhile, the effects according to the embodiments are not limited to those mentioned above, and other unmentioned effects will be clearly understood by those skilled in the art from the description below. Brief explanation of the drawing

[0020] FIG. 1 is a schematic diagram showing the configuration of a system according to one embodiment. FIG. 2 is a flowchart illustrating the process of artificial intelligence-based real-time news content automatic generation and avatar reporting provision according to one embodiment. FIG. 3 is a flowchart illustrating the process of generating customized news content based on selected keywords according to one embodiment. FIGS. 4 and 5 are flowcharts for explaining the process of adjusting the length of news content according to user preference in one embodiment. FIG. 6 is a flowchart illustrating the process of generating reporting video data by reflecting emotional expressions according to one embodiment. FIGS. 7 and 8 are flowcharts for explaining the process of displaying news content according to one embodiment. FIG. 9 is an example diagram of the configuration of a device according to one embodiment. Specific details for implementing the invention

[0021] Hereinafter, embodiments are described in detail with reference to the attached drawings. However, various modifications may be made to the embodiments, and thus the scope of the patent application is not limited or restricted by these embodiments. It should be understood that all modifications, equivalents, and substitutions to the embodiments are included within the scope of the rights.

[0022] Specific structural or functional descriptions of the embodiments are disclosed for illustrative purposes only and may be modified and implemented in various forms. Accordingly, the embodiments are not limited to the specific disclosed forms, and the scope of this specification includes modifications, equivalents, or substitutions that fall within the technical concept.

[0023] Terms such as "first" or "second" may be used to describe various components, but these terms should be interpreted solely for the purpose of distinguishing one component from another. For example, the first component may be named the second component, and similarly, the second component may be named the first component.

[0024] When it is stated that a component is "connected" to another component, it should be understood that it may be directly connected to or coupled with that other component, or that there may be other components in between.

[0025] The terms used in the embodiments are for illustrative purposes only and should not be interpreted as intended to be limiting. Singular expressions include plural expressions unless the context clearly indicates otherwise. In this specification, terms such as "comprising" or "having" are intended to indicate the existence of the features, numbers, steps, actions, components, parts, or combinations thereof described in the specification, and should be understood as not precluding the existence or addition of one or more other features, numbers, steps, actions, components, parts, or combinations thereof.

[0026] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which the embodiments pertain. Terms such as those defined in commonly used dictionaries should be interpreted as having a meaning consistent with their meaning in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.

[0027] In addition, when describing with reference to the attached drawings, identical components are assigned the same reference numeral regardless of drawing symbols, and redundant descriptions thereof are omitted. In describing the embodiments, if it is determined that a detailed description of related prior art could unnecessarily obscure the essence of the embodiments, such detailed description is omitted.

[0028] The embodiments can be implemented in various forms of products such as personal computers, laptop computers, tablet computers, smartphones, televisions, smart home appliances, intelligent automobiles, kiosks, and wearable devices.

[0029] In the embodiments, the Artificial Intelligence (AI) system is a computer system that implements human-level intelligence, and unlike existing rule-based smart systems, it is a system in which the machine learns and makes judgments on its own. As the AI ​​system improves in recognition rate and becomes capable of understanding sellers' preferences more accurately with use, existing rule-based smart systems are gradually being replaced by deep learning-based AI systems.

[0030] Artificial intelligence technology consists of machine learning and component technologies that utilize machine learning. Machine learning is an algorithmic technology that autonomously classifies and learns the characteristics of input data, while component technologies are technologies that mimic the functions of the human brain, such as cognition and judgment, by utilizing machine learning algorithms like deep learning; these technologies comprise fields such as linguistic understanding, visual understanding, reasoning / prediction, knowledge representation, and motion control.

[0031] The various fields where artificial intelligence technology is applied are as follows. Linguistic understanding refers to the technology of recognizing, applying, and processing human language and text, including natural language processing, machine translation, dialogue systems, question answering, and speech recognition / synthesis. Visual understanding refers to the technology of perceiving and processing objects like human vision, including object recognition, object tracking, image search, person recognition, scene understanding, spatial understanding, and image enhancement. Inference and prediction refers to the technology of logically reasoning and predicting by judging information, including knowledge / probability-based inference, optimization prediction, preference-based planning, and recommendation. Knowledge representation refers to the technology of automatically processing human experiential information into knowledge data, including knowledge construction (data generation / classification) and knowledge management (data utilization). Motion control refers to the technology of controlling autonomous driving of vehicles and the movement of robots, including motion control (navigation, collision, driving) and manipulation control (behavior control).

[0032] Generally, to apply machine learning algorithms to real-world situations, training is performed using a trial-and-error method due to the inherent characteristics of the fundamental methodologies. In particular, deep learning requires hundreds of thousands of iterations. Since it is impossible to execute this in a real physical external environment, training is instead performed through simulations that virtually recreate the actual physical environment on a computer.

[0033] FIG. 1 is a schematic diagram showing the configuration of a system according to one embodiment.

[0034] Referring to FIG. 1, a system according to one embodiment may include a plurality of user terminals (100) and devices (200) capable of communicating with each other through a communication network.

[0035] First, the communication network can be configured regardless of the mode of communication, such as wired or wireless, and can be implemented in various forms to enable communication between servers and between servers and terminals.

[0036] Each of the multiple user terminals (100) can be implemented as a computing device equipped with a communication function, and, for example, can be implemented as a mobile phone, desktop PC, laptop PC, tablet PC, smartphone, etc., but is not limited thereto, and can be implemented as various types of communication devices that can be connected to an external server.

[0037] A plurality of user terminals (100) are terminals used by users who wish to receive news content, and may include a first user terminal (110) used by a first user, a second user terminal (120) used by a second user, etc.

[0038] For convenience of explanation, the operation of the first user terminal (110) is described below, but it is obvious that other user terminals, such as the second user terminal (120), can perform the operation of the first user terminal (110) instead.

[0039] Each of the multiple user terminals (100) can access the device (200) through a web page, an application, etc. provided by the device (200). To this end, an application for providing news content may be installed on each of the multiple user terminals (100), and the application installed on each of the multiple user terminals (100) may operate in conjunction with the device (200).

[0040] The device (200) may be a private server owned by a person or organization providing a service using the device (200), a cloud server, or a peer-to-peer (P2P) set of distributed nodes. The device (200) may be configured to perform all or part of the computational functions, storage / reference functions, input / output functions, and control functions that a conventional computer possesses. The device (200) may be equipped with at least one artificial intelligence model that performs inference functions.

[0041] The device (200) may include a first artificial intelligence model trained to generate a script of news content by analyzing keywords, context, or importance of text, and may be configured to generate a script based on the inference results of the first artificial intelligence model and to generate voice data and reporting video data using the generated script.

[0042] Additionally, the device (200) may be configured to collect news-related content from an external data source, generate a news script based on the collected text data, convert the generated script into voice data, and generate reporting video data using a digital human avatar. Here, the digital human avatar may refer to a virtual character created or predefined to express mouth shape, facial expression changes, and gestures corresponding to the voice data.

[0043] Additionally, the device (200) may be configured to convert the generated reporting video data according to the output format of the news content provision target platform and to transmit the converted news content to a plurality of user terminals (100).

[0044] The device (200) may be configured to include a news content collection module, a script generation module, a voice generation module, an avatar control module, and a video generation module.

[0045] The device (200) may be configured to communicate with a plurality of user terminals (100) via wired or wireless means, control the operation of each of the plurality of user terminals (100), and control which information to display on the screen of each of the plurality of user terminals (100).

[0046] The device (200) is implemented as a server for providing AI-based real-time news content automatic generation and avatar reporting, and can provide various related services.

[0047] The device (200) can operate in conjunction with a database that stores data necessary for the automatic generation of AI-based real-time news content and the provision of avatar reporting.

[0048] Meanwhile, for convenience of explanation, only the first user terminal (110) and the second user terminal (120) among the plurality of user terminals (100) are shown in FIG. 1, but the number of terminals can vary depending on the embodiment. As long as the processing capacity of the device (200) allows, there is no particular limit to the number of terminals.

[0049] According to one embodiment, the device (200) can generate a script of news content by analyzing keywords, context, or importance of text based on artificial intelligence, and convert the generated script into voice data to generate news content.

[0050] In the present invention, Artificial Intelligence (AI) refers to a technology that imitates human learning ability, reasoning ability, and perceptual ability, and implements them on a computer, and may include concepts such as machine learning and symbolic logic. Machine Learning (ML) is an algorithmic technology that classifies or learns the characteristics of input data on its own. AI technology can analyze input data as a machine learning algorithm, learn from the results of the analysis, and make judgments or predictions based on the results of the learning. Furthermore, technologies that mimic the functions of the human brain, such as cognition and judgment, by utilizing machine learning algorithms can also be understood as falling within the category of AI. For example, technological fields such as linguistic understanding, visual understanding, reasoning / prediction, knowledge representation, and motion control may be included.

[0051] Machine learning can refer to the process of training neural network models using experience in processing data. It implies that through machine learning, computer software improves its own data processing capabilities. A neural network model is constructed by modeling the correlations between data, and these correlations can be expressed by multiple parameters. A neural network model extracts and analyzes features from given data to derive correlations between them; machine learning can be defined as the process of optimizing the model's parameters by repeating this process. For example, a neural network model can learn the mapping (correlation) between inputs and outputs for data given as input-output pairs. Alternatively, even when only input data is provided, a neural network model can derive regularities between the given data and learn those relationships.

[0052] An artificial intelligence learning model or neural network model can be designed to implement the structure of the human brain on a computer and may include multiple network nodes that have weights and simulate neurons of a human neural network. The multiple network nodes may have interconnected relationships by simulating the synaptic activity of neurons, where neurons exchange signals through synapses. In an artificial intelligence learning model, multiple network nodes may be located in layers of different depths and exchange data according to convolutional connections. The artificial intelligence learning model may be, for example, an Artificial Neural Network (ANN) or a Convolutional Neural Network (CNN). As an embodiment, the artificial intelligence learning model may be machine learned according to methods such as supervised learning, unsupervised learning, and reinforcement learning. Machine learning algorithms for performing machine learning may include Decision Tree, Bayesian Network, Support Vector Machine, Artificial Neural Network, Ada-boost, Perceptron, Genetic Programming, and Clustering.

[0053] Among these, CNNs are a type of multilayer perceptron designed to use minimal preprocessing. CNNs consist of one or more convolutional layers and standard artificial neural network layers stacked on top, additionally utilizing weights and pooling layers. Thanks to this structure, CNNs can fully utilize two-dimensional input data. Compared to other deep learning architectures, CNNs demonstrate good performance in both image and audio fields. CNNs can also be trained using standard backpropagation. CNNs have the advantage of being easier to train than other feedforward artificial neural network techniques and using a small number of parameters.

[0054] Convolutional networks are neural networks comprising sets of nodes with bounded parameters. Many computer vision tasks have been significantly improved, driven by the increased size of available training data and the availability of computational power, combined with algorithmic advancements such as discriminative linear units and dropout training. In the case of massive datasets, such as those available for many tasks today, outfitting is not critical, and increasing the network size improves test accuracy. Optimal utilization of computing resources becomes a limiting factor. To address this, distributed, scalable implementations of deep neural networks can be employed.

[0055] FIG. 2 is a flowchart illustrating the process of artificial intelligence-based real-time news content automatic generation and avatar reporting provision according to one embodiment.

[0056] According to one embodiment, each step illustrated in FIG. 2 can be performed by a device (200), and the device (200) can be configured to automatically generate news content while communicating with a first user terminal (110) and provide it in the form of a report through a digital human avatar.

[0057] Referring to FIG. 2, first, in step S201, the device (200) may receive a request for provision of news content from the first user terminal (110). Here, the provision request may refer to a signal requesting news content for a specific topic, keyword, category, time range, or field of interest. For example, if a specific keyword is entered or a news category is selected on the screen of the first user terminal (110), the corresponding request information may be transmitted to the device (200). That is, the device (200) may initiate a news content creation procedure based on the user request.

[0058] In step S202, the device (200) may collect news-related content newly registered in the external data source during a first period from an external data source as a first news target set. Here, the external data source may include a news website, social media, a blog, an RSS feed, a content platform, or an API-based data providing server. Additionally, the external data source may include a news providing server, a portal site, a social network service (SNS), a content providing platform, or an open API, and may collect data through at least one of crawling, API calling, or data streaming methods. Additionally, the first period may be set differently depending on the embodiment, such as the last hour, one day, or one week. Additionally, the first news target set may be defined as a data set containing a plurality of news-related content collected during the first period. That is, the device (200) may collect news data to reflect the latest issues and classify it as a first news target set.

[0059] According to one embodiment, collection from an external data source may be performed in a streaming manner or by repeating according to a preset period. For example, the device (200) may receive new content as soon as it is registered from a news API, RSS feed, or real-time content provider server, or may periodically check for and collect new content according to a preset period. That is, the device (200) may reflect the latest news content in real-time or near-real-time.

[0060] In step S203, the device (200) can extract text included in the first news target set as first text data. Here, the first text data may be defined as text information including at least one of the news article body, title, summary, comments, or explanatory phrases. For example, the device (200) can extract only the text area from an HTML document or extract core text by removing unnecessary elements through natural language processing techniques. That is, the device (200) can generate first text data by extracting text from the content included in the first news target set, thereby obtaining text data in an analyzable form.

[0061] In step S204, the device (200) can generate a first input signal by encoding the first text data. Here, the first input signal may refer to the result of converting the text data into an embedding vector, a token sequence, or quantified feature data. For example, the device (200) can perform word embedding or sentence embedding using a natural language processing model. That is, the device (200) can generate a first input signal by converting the first text data into a form that can be processed by an artificial intelligence model.

[0062] In step S205, the device (200) may input a first input signal to a first artificial intelligence model trained to generate a script of news content by analyzing keywords, context, or importance of the text. Here, the first artificial intelligence model may be defined as a model trained to generate a script having an utterance structure suitable for news reporting from input text data. Additionally, the first artificial intelligence model may be defined as a sequence generation model trained to generate a news script containing an utterance structure by summarizing and reconstructing the input text.

[0063] In step S206, the device (200) may obtain a first output signal representing a first script generated as a result of analysis of the first text data from the first artificial intelligence model. Here, the first script may be defined as text containing an utterance structure for conveying news content, and may include a sentence connection structure, emphasis expressions, utterance order, and intonation information. For example, the first script may include a reporting structure such as “headline → detailed description → summary.” That is, the device (200) may generate a structured script for conveying news content.

[0064] According to one embodiment, the first artificial intelligence model may be implemented as a sequence generation model based on natural language processing, and may include, for example, a Transformer structure, an Artificial Neural Network, or a combination thereof.

[0065] The first artificial intelligence model can be configured to learn the relationships between tokens of input text data and perform sentence generation that reflects contextual information.

[0066] In addition, the first artificial intelligence model can be trained through pre-training and fine-tuning processes using training data including news article data, summary data, and reporting script data.

[0067] For example, the first artificial intelligence model can generate an appropriate reporting script for an input text by learning the correspondence between a news article and a summary or broadcast script corresponding to that article. At this time, the loss function can be set to minimize the difference between the generated script and the correct script, and the model parameters can be optimized through iterative learning. That is, the device (200) can generate a natural reporting script that reflects the meaning of the news data through the first artificial intelligence model.

[0068] In addition to the learning method described above, the first artificial intelligence model may also be trained through various learning methods of the known artificial intelligence model.

[0069] In step S207, the device (200) can generate first voice data by converting the first script into voice data based on the first output signal. Here, the first voice data may refer to a digital audio signal converted from text into voice and may be generated using Text-to-Speech technology. Additionally, the first voice data may be generated by reflecting intonation, speed, emotional expression, etc. That is, the device (200) can convert the reporting script into voice in the form of an actual broadcast.

[0070] According to one embodiment, the generation of the first voice data may be performed by dividing it into utterance units reflecting the sentence structure, emotional information, and keyword importance included in the first script, and then adjusting the intonation, speed, and stress corresponding to each utterance unit. For example, the device (200) may generate a natural reporting voice by changing the tone of the voice or giving emphasis to specific keywords according to the result of the emotional analysis.

[0071] In step S208, the device (200) may create or call a first avatar, which is a digital human avatar for delivering news content. Here, the first avatar may be defined as a virtual character configured to express mouth shape, facial expression changes, and gestures corresponding to voice data, and may be called from a pre-generated avatar or dynamically created according to user settings. For example, the first avatar may be configured as an announcer type, a character type, or a brand-specific avatar. That is, the device (200) may set the subject to perform the reporting.

[0072] In step S209, the device (200) can generate first reporting video data including mouth shape synchronization, facial expression changes, and gestures of the first avatar based on the first voice data and the first script. Here, the first reporting video data is video data synchronized with the voice data and may include information on mouth shape, facial expression, and gestures. For example, the first reporting video data may have emphasis gestures applied according to specific keywords or facial expressions changed according to emotions. That is, the device (200) can generate news reporting video through a digital human avatar.

[0073] According to one embodiment, the first reporting video data is video data synchronized with the first voice data, and can be defined as multimodal data including mouth shape information corresponding to the voice data, facial expression change information according to the content of speech, and gesture information.

[0074] Specifically, the first reporting video data may have a data structure including a voice signal aligned along a time axis and mouth shape parameters, facial expression parameters, and gesture parameters corresponding to each time interval.

[0075] For example, when a sentence containing a specific keyword is spoken, facial expression changes or hand gestures indicating emphasis may be applied, and for sentences requiring emotional expression, the intensity of the facial expression may be adjusted.

[0076] That is, the device (200) can combine voice data and visual representation elements to generate image data in a form similar to actual reporting.

[0077] Specifically, the device (200) can divide the time axis based on the speech interval of the first voice data and extract phoneme or pronunciation unit information to determine the mouth shape corresponding to each time interval. Here, the phoneme information may be defined as data representing pronunciation elements included in the voice data, and the device (200) can generate the mouth movement of the first avatar by mapping mouth shape parameters corresponding to the phoneme information. For example, the degree of mouth opening, lip shape, or jaw movement corresponding to a specific vowel or consonant pronunciation may be determined according to a predefined rule or a learned model. That is, the device (200) can implement lip-sync by generating a mouth shape synchronized with the voice data.

[0078] Additionally, the device (200) can generate facial expression change information by analyzing the meaning, emotion, or emphasis section of a sentence included in the first script. Here, the facial expression change information can be defined as parameters representing the degree of change of the avatar's eyes, eyebrows, corners of the mouth, and facial muscles. For example, it can be controlled by generating a smile expression in the case of a sentence containing positive content, and increasing the intensity of the expression when it contains urgent or important content. That is, the device (200) can generate natural facial expression changes based on the semantic information of the text.

[0079] Additionally, the device (200) can generate gesture information based on the sentence structure or keywords of the first script. Here, the gesture information can be defined as data representing the hand movements, arm movements, or changes in the direction of the upper body of the avatar. For example, a hand-raising motion can be generated when emphasizing a specific numerical value, or a gesture using both hands can be generated when explaining comparison content. That is, the device (200) can improve the delivery of the report by generating a gesture corresponding to the utterance content.

[0080] The device (200) can generate first reporting video data by aligning mouth shape parameters, facial expression parameters, and gesture parameters along the time axis and rendering them in synchronization with first voice data. Here, rendering may refer to the process of generating video images on a frame-by-frame basis using a 2D or 3D model of an avatar. That is, the device (200) can generate a digital human reporting video in which voice, facial expressions, and gestures are integrated.

[0081] In step S210, the device (200) may set a format representing the output format of the news content provision target platform as a first format based on a request for provision of news content. Here, the first format may be defined as an output format including video resolution, aspect ratio, playback length, subtitle format, or interface structure. For example, the first format may be set as a mobile shorts video, a square news video, or a content format for a metaverse environment. That is, the device (200) may set output conditions suitable for the content provision environment.

[0082] In step S211, the device (200) can generate first news content by converting first reporting video data according to a first format. Here, the conversion may include resizing the video, reconstructing frames, inserting subtitles, or adding UI elements. Additionally, the conversion may include processes of dividing video segments according to platform characteristics, adjusting the position and display method of subtitles, and rearranging user interface elements. That is, the device (200) can generate news content optimized for the platform.

[0083] In step S212, the device (200) can transmit the first news content to the first user terminal (110). For example, the first news content may be provided via a streaming method or a file transfer method. That is, the device (200) can provide the user with automatically generated news reporting content.

[0084] Accordingly, since news scripts can be automatically generated using an artificial intelligence model based on news content collected from external data sources and converted into reporting video data using digital human avatars for provision, real-time news content can be generated without a separate video production process, thereby improving the efficiency of content production. In particular, by automating the processes of text analysis, script generation, voice generation, and avatar-based reporting video generation by linking them into a single pipeline, the consistency and processing efficiency of the entire news content generation process can be enhanced.

[0085] FIG. 3 is a flowchart illustrating the process of generating customized news content based on selected keywords according to one embodiment.

[0086] According to one embodiment, each step illustrated in FIG. 3 may be performed in conjunction after step S202 and after step S211, and the device (200) may be configured to select and sort news content based on keywords input from the first user terminal (110).

[0087] Referring to FIG. 3, first, in step S301, if the device (200) confirms that a request for providing news content is a request containing a first keyword, it may classify data determined to be relevant among the first news target set based on thematic relevance, semantic similarity, or keyword match with the first keyword into a second news target set. Here, the first keyword may be defined as a word or phrase representing a topic, person, brand, or specific event of interest to the user, and thematic relevance may be defined as a concept indicating whether the topic included in a constituent unit is identical or similar to the first keyword. Additionally, semantic similarity may be calculated through semantic distance between texts or embedding-based similarity, and keyword match may be determined based on whether the first keyword is included in the news content. That is, the device (200) may classify news-related content determined to be relevant to the first keyword into a second news target set.

[0088] In step S302, the device (200) may select a second set of news targets from a first set of news targets and exclude the remaining data. Here, the remaining data may refer to news-related content whose relevance to the first keyword is below a standard. That is, the device (200) may perform the subsequent news content generation process based only on the second set of news targets.

[0089] Afterward, the device (200) can generate first news content based on a second news target set according to the steps of FIG. 2.

[0090] In step S303, if the first news content includes multiple news segments, the device (200) can arrange the multiple news segments according to their relevance or importance to the first keyword. Here, a news segment may refer to an individual news item or topic unit distinguished within a single news content, and the relevance may be calculated based on the semantic similarity between the first keyword and each news segment or whether the keyword is included. Additionally, the importance may be calculated based on the news segment's recency, views, frequency of mention, or user preference. That is, the device (200) can arrange the news segments so that those with high relevance or importance to the first keyword are placed preferentially.

[0091] Accordingly, news-related content can be selected based on input keywords to generate news content based only on content that matches user interests, and by sorting multiple news segments according to their relevance or importance to the keywords, news information that should be provided to the user first can be placed at the top, thereby improving the accuracy and delivery efficiency of user-customized news content.

[0092] FIGS. 4 and 5 are flowcharts for explaining the process of adjusting the length of news content according to user preference in one embodiment.

[0093] According to one embodiment, each step illustrated in FIGS. 4 to 5 may be performed after step S206, and the device (200) may be configured to dynamically adjust the length of the news content by reflecting the subject preference of the first user terminal (110).

[0094] Referring to FIGS. 4 and 5, first, in step S401, the device (200) can divide the first script into multiple constituent units according to news topics. Here, a constituent unit can be defined as a paragraph, a group of sentences, or a semantic unit corresponding to a single news topic, and can be divided into units such as, for example, “headline,” “detailed explanation,” or “background information.” That is, the device (200) can divide the first script by topic and configure it into an analyzable structure.

[0095] In step S402, the device (200) may select a first component unit, which is one of a plurality of component units, as the target for analysis. That is, the device (200) may determine whether to sequentially adjust the length for each component unit.

[0096] In step S403, the device (200) can determine the number of characters included in the first configuration unit as the first count. Here, the first count can be defined as a reference value representing the length of the corresponding configuration unit. That is, the device (200) can set a reference value for length adjustment.

[0097] In step S404, the device (200) can identify the news topic of the first constituent unit as the first topic. Here, the first topic may refer to the core content or category handled by the corresponding constituent unit. That is, the device (200) can identify topic information by constituent unit.

[0098] In step S405, the device (200) can determine the first user's preference for a first topic as a first index based on the first user's preference by topic. Here, the first index can be defined as a value that quantifies the user's interest in a specific topic, and can be calculated, for example, based on click history, viewing time, selection frequency, or interaction data. At this time, the first user's preference by topic can be determined through information stored in a database. That is, the device (200) can quantitatively reflect the user's preference.

[0099] In step S406, the device (200) can check whether the first index is included within a reference range. Here, the reference range may be set differently depending on the embodiment.

[0100] If it is confirmed in step S406 that the first index falls within the reference range, the length of the content for the first constituent unit may not be adjusted.

[0101] If it is confirmed in step S406 that the first index is not included within the reference range, in step S501, the device (200) can check whether the first index is higher than the maximum value of the reference range.

[0102] If it is confirmed in step S501 that the first index is higher than the maximum value of the reference range, in step S502, the device (200) can calculate the difference between the maximum value of the reference range and the first index as the first difference value. That is, if the user preference exceeds the reference range, the device (200) can calculate the degree of excess as the first difference value.

[0103] In step S503, the device (200) may set the first weight to a higher value within the upper section of the weight range as the first difference value increases. Here, the weight range may be set differently depending on the embodiment, for example, to a range of 0 to 2, and the upper section of the weight range may be set to a range of 1 to 2. That is, the device (200) may increase the first weight to provide more information as the preference increases.

[0104] For example, the device (200) can set the first weight to 1.1 if the first difference value is confirmed to be 1, and can set the first weight to 1.2 if the first difference value is confirmed to be 2.

[0105] In step S504, the device (200) can calculate a second number by multiplying a first number by a first weight. Here, the second number can be defined as a value representing the extended content length. That is, the device (200) can increase the content length according to user preference.

[0106] In step S505, the device (200) can elaborate on the content of the first constituent unit to expand it from a first number of characters to a second number of characters. Here, elaboratement may mean the process of increasing the number of sentences or the length of sentences by adding relevant information based on the core sentences included in the first constituent unit.

[0107] For example, the device (200) can expand the content by adding background information related to the subject included in the first component unit, explanations of causes or results, related numerical information or case information, or by rewriting existing sentences into more specific expressions.

[0108] Additionally, the device (200) can generate additional sentences using the first artificial intelligence model used to generate the first script or a separate artificial intelligence model and insert them into the existing constituent unit. That is, the device (200) can detail the content to increase the density and volume of information while maintaining the meaning of the first constituent unit.

[0109] If it is confirmed in step S501 that the first index is not higher than the maximum value of the reference range, it can be confirmed that the first index is not included within the reference range and is outside of it, so that the first index is lower than the minimum value of the reference range; and if it is confirmed that the first index is lower than the minimum value of the reference range, in step S506, the device (200) can calculate the difference between the first index and the minimum value of the reference range as the second difference value. That is, if the user preference is low, the device (200) can calculate the degree of deficiency as the second difference value.

[0110] In step S507, the device (200) may set the second weight to a lower value within the lower interval of the weight range as the second difference value increases. Here, the weight range may be set differently depending on the embodiment, for example, to a range of 0 to 2, and the lower interval of the weight range may be set to a range of 0 to 1. That is, the device (200) may reduce the second weight to shorten the content length as the preference decreases.

[0111] For example, the device (200) can set the second weight to 0.9 if the second difference value is confirmed to be 1, and can set the second weight to 0.8 if the second difference value is confirmed to be 2.

[0112] In step S508, the device (200) can calculate a third number by multiplying the first number by the second weight. Here, the third number can be defined as a value representing the reduced content length. That is, the device (200) can reduce the content length according to user preference.

[0113] In step S509, the device (200) can simplify the content of the first constituent unit to reduce it from a first number of characters to a third number of characters. Here, simplification may mean a process of reducing the total number of characters by removing some sentences or abbreviating sentences within the scope of maintaining the core meaning of the sentences included in the first constituent unit.

[0114] For example, the device (200) can reduce the content by removing additional explanations, repetitive expressions, or detailed examples of low importance among the sentences included in the first constituent unit, integrating multiple sentences into one sentence, or reorganizing sentences around key keywords.

[0115] Additionally, the device (200) can generate a summary sentence using the first artificial intelligence model used to generate the first script or a separate artificial intelligence model and replace it with an existing constituent unit. That is, the device (200) can simplify the content to reduce the density and volume of information while maintaining the meaning of the first constituent unit.

[0116] Accordingly, by dynamically adjusting the length of news content based on user preferences by topic, detailed information can be provided for topics of high interest and brief information for topics of low interest, thereby achieving efficiency in providing customized content and optimization of information delivery.

[0117] FIG. 6 is a flowchart illustrating the process of generating reporting video data by reflecting emotional expressions according to one embodiment.

[0118] According to one embodiment, each step illustrated in FIG. 6 can be performed in conjunction with step S209, and the device (200) can be configured to analyze text included in the first constituent unit to derive an emotion, and to reflect facial expressions and gestures corresponding to the derived emotion in the first avatar.

[0119] Referring to FIG. 6, first, in step S601, the device (200) can recognize a key keyword through natural language processing for the first constituent unit. Here, the key keyword can be defined as a word or phrase representing the meaning of the constituent unit and may include nouns, verbs, adjectives, or words expressing emotion. For example, the device (200) can identify the key keyword using morphological analysis, part-of-speech tagging, or importance-based keyword extraction techniques. That is, the device (200) can extract a key word for sentiment analysis.

[0120] In step S602, the device (200) can distinguish which of a plurality of preset emotion types each of the core keywords of the first constituent unit is. Here, the emotion type can be defined as a category representing an emotional state such as positive, negative, neutral, joy, anger, surprise, or sadness, and can be determined based on an emotion dictionary or an emotion classification model. That is, the device (200) can classify the emotion type corresponding to each core keyword.

[0121] In step S603, the device (200) can classify a core keyword classified as a first type, which is one of a plurality of emotion types, into a first group. Here, the first group can be defined as a set of core keywords belonging to the same emotion type. That is, the device (200) can group keywords by emotion type.

[0122] In step S604, the device (200) can determine the frequency of occurrence of each key keyword in the first constituent unit as the first frequency for each key keyword classified into the first group. Here, the first frequency can be defined as a value representing the frequency with which a specific keyword appears within the text. That is, the device (200) can calculate the frequency of keywords that affect emotional expression for each keyword.

[0123] In step S605, the device (200) may set a first score within a first score range to a higher score for each core keyword classified into a first group, as the first occurrence increases. Here, the first score range may be set differently depending on the embodiment, and the first score may be defined as a weight calculated based on the frequency of appearance of the keyword. For example, it may be set by a linear or non-linear function such that the first score increases as the first occurrence increases. At this time, the first score may be calculated according to a function corresponding to the first occurrence within the first score range. That is, the device (200) may highly reflect the influence of keywords that appear repeatedly.

[0124] For example, if the device (200) confirms that the first number of times for the first core keyword is 1 time, it can set the first score for the first core keyword to 10 points, and if the first number of times for the first core keyword is confirmed to be 2 times, it can set the first score for the first core keyword to 20 points.

[0125] In step S606, the device (200) can determine, for each core keyword classified into a first group, a value representing the length of each core keyword as a first numerical value. Here, the first numerical value may be defined as at least one of the number of characters, tokens, or syllables included in each core keyword. That is, the device (200) can calculate keyword length information for each keyword.

[0126] In step S607, the device (200) may set a second score within a second score range to a higher score for each core keyword classified into a first group, as the first value increases. Here, the second score range may be set differently depending on the embodiment, and the second score may be defined as a weight calculated based on the length of the keyword. For example, it may be set by a linear or non-linear function so that the second score increases as the first value increases. At this time, the second score may be calculated according to a function corresponding to the first value within the second score range. That is, the device (200) may set it so that longer keywords have a greater influence on emotional expression.

[0127] For example, if the device (200) confirms that the first value for the first core keyword is 1, it can set the second score for the first core keyword to 10 points, and if the first value for the first core keyword is confirmed to be 2, it can set the second score for the first core keyword to 20 points.

[0128] In step S608, the device (200) can calculate the influence of each key keyword by summing the first score and the second score for each key keyword classified into the first group. Here, the influence can be defined as a value representing the overall influence that the keyword has on emotional expression. That is, the device (200) can calculate the influence representing the importance of the keyword by reflecting multiple criteria.

[0129] In step S609, the device (200) can calculate the influence of the first type by summing the influence of each core keyword classified into the first group. That is, the device (200) can calculate the influence of each emotion type by integrating the influences of keywords belonging to the same emotion type.

[0130] The device (200) can calculate the influence for each of the multiple emotion types through the same method as calculating the influence of the first type.

[0131] In step S610, if the device (200) calculates the influence of each of the multiple emotion types and compares the influence of each of the multiple emotion types, and confirms that the influence of the first type is the highest, the device (200) may set the emotion of the first constituent unit as the first type. That is, the device (200) may determine the most dominant emotion type as the final emotion.

[0132] In step S611, the device (200) can set facial expression changes and gestures of the first avatar so that a first type of emotion is expressed when the first configuration unit is played. Here, the facial expression changes and gestures may be set based on predefined animation parameters or motion patterns corresponding to the emotion type. For example, in the case of positive emotion, a smiling expression and a bright gesture may be applied, and in the case of negative emotion, a lower facial expression intensity or a restricted gesture may be applied. That is, the device (200) can convert the text-based emotion analysis results into visual expression elements and reflect them in the reporting video.

[0133] Accordingly, the emotion of each constituent unit can be quantitatively calculated based on the emotion type and influence of key keywords included in the text, and the calculated emotion can be reflected in the avatar's facial expressions and gestures, thereby improving the accuracy of emotional expression and immersion when delivering news content.

[0134] FIGS. 7 and 8 are flowcharts for explaining the process of displaying news content according to one embodiment.

[0135] According to one embodiment, each step illustrated in FIGS. 7 to 8 may be performed in conjunction with step S212, and the device (200) may control the screen configuration so that the first news content is effectively delivered to the first user terminal (110). At this time, the device (200) may be configured to distinguish and set an area where the first avatar performs reporting and an area where the currently playing sentence and the next playing sentence are displayed, and to adjust the font size according to the length of each sentence and the displayable area.

[0136] Referring to FIGS. 7 and 8, first, in step S701, the device (200) can generate a first page for displaying the first news content and transmit it to the first user terminal (110). Here, the first page may be defined as a screen, web page, application screen, or interface screen for visually displaying the first news content. That is, the device (200) can generate a screen configuration for displaying the first news content on the user terminal.

[0137] In step S702, the device (200) may divide the first page into a first area and a second area. Here, the first area may be defined as an area where the first avatar is displayed, and the second area may be defined as an area where a sentence is displayed. For example, the left or top area of ​​the first page may be set as the first area, and the right or bottom area may be set as the second area. That is, the device (200) may distinguish and set the avatar display area and the sentence display area.

[0138] In step S703, the device (200) can control the first avatar to be displayed in the first area to deliver the first news content. Here, the first avatar may be a digital human avatar that performs reporting on the first news content through voice, facial expression changes, and gestures. That is, the device (200) can control the avatar reporting to be performed in the first area.

[0139] In step S704, the device (200) can identify the currently playing sentence as the first sentence when the first avatar delivers the first news content. Here, the first sentence can be defined as a sentence displayed in synchronization with the avatar's speech at the current time. That is, the device (200) can identify the currently playing sentence as the first sentence.

[0140] In step S705, the device (200) can identify the sentence to be played after the first sentence as the second sentence. Here, the second sentence can be defined as a sentence scheduled to be played sequentially after the playback of the first sentence ends. That is, the device (200) can identify the next sentence scheduled to be played as the second sentence.

[0141] In step S706, the device (200) can determine the number of characters included in the first sentence as the fourth number. Here, the fourth number can be defined as a value representing the number of characters constituting the first sentence. That is, the device (200) can quantify the length of the sentence currently being played.

[0142] In step S707, the device (200) can determine the number of characters included in the second sentence as the fifth number. Here, the fifth number can be defined as a value representing the number of characters constituting the second sentence. That is, the device (200) can quantify the length of the next sentence scheduled for playback.

[0143] In step S708, the device (200) can determine a preset font size as a first size to display the sentence currently being played, and a preset font size as a second size to display the sentence to be played next. Here, the first size can be defined as a default font size for displaying the first sentence with emphasis, and the second size can be defined as a default font size for displaying the second sentence in a relatively small size. That is, the device (200) can set the current sentence and the next sentence to different display standards. At this time, the first size can be set to a larger size than the second size.

[0144] In step S709, the device (200) can identify the area of ​​the second region as the first region. Here, the first region may be defined as the total displayable area where a sentence can be placed in the second region. For example, the actual text placement area excluding margins, icons, or fixed UI elements may be set as the first region. That is, the device (200) can identify the total area for displaying the sentence.

[0145] In step S710, the device (200) can determine the minimum area required to display one character of the first size as the second area. Here, the second area can be defined as a unit character area corresponding to the first size. That is, the device (200) can determine the unit character area for displaying the first sentence.

[0146] In step S711, the device (200) can calculate a third area by multiplying the second area by the fourth number. Here, the third area can be defined as the total area required to display the first sentence in characters of the first size. That is, the device (200) can calculate the required area to display the currently playing sentence.

[0147] In step S712, the device (200) can calculate a fourth area by subtracting a third area from a first area. Here, the fourth area can be defined as the remaining area within the second area after the first sentence is displayed. That is, the device (200) can determine the area available for displaying the next sentence.

[0148] In step S713, the device (200) may set a 2-1 area within a second area with the size of a third area, and a 2-2 area within a second area with the size of a fourth area. Here, the 2-1 area may be defined as an area where a first sentence is displayed, and the 2-2 area may be defined as an area where a second sentence is displayed. That is, the device (200) may divide the second area into a current sentence display area and a next sentence display area.

[0149] In step S714, the device (200) can control the first sentence to be displayed in the 2-1 area with a first size of text. That is, the device (200) can display the currently playing sentence with relatively large text so that it is recognized preferentially by the user.

[0150] In step S801, the device (200) can determine the minimum area required to display one character of the second size as the fifth area. Here, the fifth area can be defined as the unit character area corresponding to the second size. That is, the device (200) can determine the unit character area for displaying the following sentence at the basic size.

[0151] In step S802, the device (200) can calculate a sixth area by multiplying the fifth area by the fifth number. Here, the sixth area can be defined as the total area required to display the second sentence in characters of the second size. That is, the device (200) can calculate the area required to display the next sentence in the basic character size.

[0152] In step S803, the device (200) can check whether the fourth area is larger than the sixth area.

[0153] If it is confirmed in step S803 that the fourth area is larger than the sixth area, in step S804, the device (200) can control the second sentence to be displayed in the second size of the text in the second-2 area. That is, the device (200) can reliably display the next sentence with a preset default text size.

[0154] If it is confirmed in step S803 that the fourth area is not larger than the sixth area, in step S805, the device (200) can calculate the seventh area by dividing the fourth area by the fifth number. Here, the seventh area can be defined as the average area that can be allocated to each character of the second sentence within the second-2 area. That is, the device (200) can calculate the area that can be allocated to each character within the limited remaining area.

[0155] In step S806, the device (200) can determine the maximum character size that can be displayed in the seventh area as the third size. Here, the third size may be defined as a character size adjusted so that the entire second sentence is displayed within the second-2 area. Additionally, the third size may be defined as a value in which the height and width of the character are calculated based on the displayable area per character corresponding to the seventh area. That is, the device (200) can dynamically adjust the character size of the next sentence to fit the remaining area.

[0156] In step S807, the device (200) can control the second sentence to be displayed in a third size character in the second-2 area. That is, the device (200) can control the next sentence to be displayed at the maximum size within the range that does not exceed the remaining area.

[0157] Accordingly, the currently playing sentence and the next sentence are displayed separately in different areas, and the font size can be adjusted according to the length of each sentence and the displayable area. This allows for intuitive provision of the current and next reporting content to the user and improves the readability and delivery efficiency of news content even within a limited display area.

[0158] FIG. 9 is an example diagram of the configuration of a device according to one embodiment.

[0159] A device (200) according to one embodiment includes a processor (210) and a memory (220). The processor (210) may include at least one device described with reference to FIGS. 1 through 8 or may perform at least one method described with reference to FIGS. 1 through 8. A person or organization using the device (200) may provide services related to some or all of the methods described with reference to FIGS. 1 through 8.

[0160] The memory (220) may store information related to the methods described above or store a program in which the methods described below are implemented. The memory (220) may be volatile memory or non-volatile memory.

[0161] The processor (210) can execute a program and control the device (200). The code of the program executed by the processor (210) can be stored in memory (220). The device (200) is connected to an external device (e.g., a personal computer or a network) through an input / output device (not shown in the drawing) and can exchange data through wired or wireless communication.

[0162] The device (200) can be used to train an artificial intelligence model or to use a trained artificial intelligence model. The memory (220) may contain an artificial intelligence model that is being trained or has been trained. The processor (210) can train or execute an artificial intelligence model algorithm stored in the memory (220). The training device for training the artificial intelligence model and the device (200) for using the trained artificial intelligence model may be the same or separate.

[0163] The embodiments described above may be implemented as hardware components, software components, and / or combinations of hardware and software components. For example, the devices, methods, and components described in the embodiments may be implemented using one or more general-purpose or special-purpose computers, such as, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable gate array (FPGA), a programmable logic unit (PLU), a microprocessor, or any other device capable of executing and responding to instructions. The processing unit may execute an operating system (OS) and one or more software applications executed on said operating system. Additionally, the processing unit may access, store, manipulate, process, and generate data in response to the execution of the software. For ease of understanding, the processing unit may be described as being used as a single unit, but those skilled in the art will understand that the processing unit may include multiple processing elements and / or multiple types of processing elements. For example, the processing unit may include multiple processors or one processor and one controller. Additionally, other processing configurations, such as parallel processors, are also possible.

[0164] The method according to the embodiment may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either alone or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the embodiment, or they may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The hardware devices described above may be configured to operate as one or more software modules to perform the operation of the embodiment, and vice versa.

[0165] Software may include computer programs, code, instructions, or a combination of one or more of these, and may configure a processing unit to operate as desired or command the processing unit independently or collectively. Software and / or data may be permanently or temporarily embodied in any type of machine, component, physical device, virtual equipment, computer storage medium or device, or transmitted signal wave so as to be interpreted by the processing unit or to provide instructions or data to the processing unit. Software may be distributed over networked computer systems and may be stored or executed in a distributed manner. Software and data may be stored on one or more computer-readable recording media.

[0166] Although the embodiments have been described above with reference to the limited drawings, those skilled in the art can apply various technical modifications and variations based on the above. For example, suitable results may be achieved even if the described techniques are performed in a different order than described, and / or if the components of the described system, structure, device, circuit, etc. are combined or assembled in a form different from described, or replaced or substituted by other components or equivalents.

[0167] Therefore, other implementations, other embodiments, and equivalents to the claims also fall within the scope of the claims set forth below.

Claims

Claim 1 A method for automatically generating real-time news content based on artificial intelligence and providing avatar reporting, performed by a device, comprising: receiving a request for provision of news content from a first user terminal; collecting news-related content newly registered in the external data source from an external data source during a preset first period as a first news target set; extracting text included in the first news target set as first text data; encoding the first text data to generate a first input signal; inputting the first input signal to a first artificial intelligence model trained to generate a script of news content by analyzing keywords, context, or importance of the text; obtaining a first output signal representing a first script generated as an analysis result of the first text data from the first artificial intelligence model; converting the first script into voice data based on the first output signal to generate first voice data; generating or calling a first avatar, which is a digital human avatar for delivering news content; generating first reporting video data including mouth shape synchronization, facial expression change, and gesture of the first avatar based on the first voice data and the first script; and the news Based on a request for provision of content, the method comprises: a step of setting a format representing the output format of a news content provision target platform as a first format; a step of generating first news content by converting the first reporting video data according to the first format; and a step of transmitting the first news content to the first user terminal, wherein after the step of collecting the first news target set, if the request for provision of news content is confirmed to be a request including a first keyword, the method comprises a step of classifying data determined to be relevant among the first news target set based on thematic relevance, semantic similarity, or keyword match with the first keyword into a second news target set.and further includes the step of selecting the second news target set from the first news target set and excluding the remaining data; after the step of generating the first news content, if the first news content includes a plurality of news segments, further includes the step of sorting the plurality of news segments according to the relevance or importance with the first keyword; after the step of obtaining the first output signal, the step of dividing the first script into a plurality of constituent units according to the news topic; the step of selecting a first constituent unit, which is one of the plurality of constituent units, as an analysis target; the step of confirming the number of characters included in the first constituent unit as a first count; the step of confirming the news topic of the first constituent unit as a first topic; the step of confirming the first user's preference for the first topic as a first index based on the first user's preference by topic; the step of calculating the difference between the maximum value of the reference range and the first index as a first difference value if the first index is confirmed to be higher than the maximum value of a preset reference range; the step of setting a first weight to a higher value within the upper section of a preset weight range as the first difference value is higher; the first count and the first weight A step of calculating a second number by multiplying the values; a step of detailing the content of the first constituent unit to expand from the first number of characters to the second number of characters; a step of calculating the difference between the first index and the minimum value of the reference range as a second difference value if it is confirmed that the first index is lower than the minimum value of the reference range; a step of setting a second weight to a lower value within the lower section of the weight range as the second difference value is higher; a step of calculating a third number by multiplying the first number and the second weight;The method further includes the step of simplifying the content of the first constituent unit to reduce it from the first number of characters to the third number of characters, and the step of generating first reporting video data including mouth shape synchronization, facial expression change, and gesture of the first avatar comprises: a step of recognizing a core keyword through natural language processing for the first constituent unit; a step of distinguishing which of a plurality of preset emotion types each core keyword of the first constituent unit is; a step of classifying the core keyword classified as a first type, which is one of the plurality of emotion types, into a first group; a step of confirming the frequency of occurrence of each core keyword in the first constituent unit as a first frequency for each core keyword classified into the first group, the more the first frequency is, the higher the first score within a preset first score range; a step of confirming a value representing the length of each core keyword classified into the first group as a first numerical value for each core keyword classified into the first group, the more the preset second A step of setting a second score to a higher score within a score range; a step of calculating the influence of each core keyword by summing the first score and the second score for each core keyword classified into the first group; a step of calculating the influence of the first type by summing the influence of each core keyword classified into the first group; and a step in which, when the influence of each of the plurality of emotion types is calculated, if the influence of the first type is confirmed to be the highest as a result of comparing the influences of each of the plurality of emotion types, a step of setting the emotion of the first constituent unit as the first type.A method for providing AI-based real-time news content automatic generation and avatar reporting, comprising the step of setting facial expression changes and gestures of the first avatar so that the first type of emotion is expressed when the first constituent unit is played. Claim 2 delete Claim 3 delete

Images