KR20220088369A - Method for recommending AI automatic hashtag based on trend analysis - Google Patents

Abstract

The present invention relates to a trend analysis-based AI automatic hashtag recommendation method, and the trend analysis-based AI automatic hashtag recommendation method according to an embodiment comprises the steps of extracting at least one word from a produced image content, the produced image content Analyzing metadata for image contents selected from among images that correspond to a category corresponding to and previously registered on SNS, and comparing the extracted word with the analyzed metadata, It may include the step of recommending a hashtag to be used for the video content.

Description

{Method for recommending AI automatic hashtag based on trend analysis}

The present invention relates to a trend analysis-based AI automatic hashtag recommendation method, and more specifically, to automatically recommend a hashtag in order to be able to express the subject and personality of one's content and to be searched after posting the content. It's about function.

Looking at today's Internet networks, as Web 2.0, in which the user is the subject, has become an issue, the active participation of users in the Internet has become a subject of epochal interest.

In other words, the existing user, who stayed only in the role of a consumer, became a prosumer who plays an active role in 　 content 　 production 　 It has become a being that encompasses the production, distribution, sharing, and sale of content. User-created content (UCC) is actively being produced.

In particular, thanks to the video trend throughout the digital content market, users 　 produced 　 contents have also entered the multimedia era centered on video rather than text or image. Content is becoming more active.

On the other hand, the MZ generation recognizes the way of life as an option, shares information through constant communication, and always pursues new things.

As the MZ generation with these characteristics is currently meeting with an aging society, senior content is riding the trend, and seniors are showing off not only in content but also in advertising and marketing.

Korea Patent No. 2160600 "Hash tag recommendation method, device and computer-readable recording medium for improving user response" Korea Patent Publication No. 2020-0045310 "Hash tag-based information recommendation method and terminal for performing the same"

An object of the present invention is to provide a technology for automatically recommending hashtags in order to help users more easily author and distribute content by developing a one-solution to solve the cultural and social needs of seniors.

An object of the present invention is to provide an automatic hashtag function in order to be able to express the subject and personality of one's own content and to be searched after posting the content.

An object of the present invention is to provide a user's content-based hashtag recommendation function using unsupervised learning, collaborative filtering, deep learning algorithm, and the like.

An object of the present invention is to satisfy the needs of users as much as possible by configuring the top 10 popular hashtags and recommending hashtags using an image recognition weighting method.

An object of the present invention is to derive key keywords by collecting metadata (title, subtitle, video) of video content produced by users using a video content authoring platform.

An object of the present invention is to rank keywords by calculating the degree of association between keywords derived from titles and subtitles and objects detected in images.

An object of the present invention is to facilitate the use of hashtags by identifying key keywords from data input by a user and images uploaded by a user, and recommending appropriate hashtags to the user.

An object of the present invention is to provide a video content authoring platform administrator with a result that can determine the status of hashtags used in the corresponding platform.

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment includes extracting at least one word from the produced image content, the images corresponding to the category corresponding to the produced image content, and pre-registered on SNS The method may include analyzing metadata for the image contents selected from among them, and recommending a hashtag to be used for the produced image content by comparing the extracted word with the analyzed metadata.

According to an embodiment, the extracting of at least one word from the produced image content includes receiving image metadata for the produced image content, uploading the image content, and an image from the uploaded image content. extracting an object based on recognition, selecting a word included in a title or subtitle for the video content from the image metadata, and extracting words corresponding to the extracted object from among the selected words may include steps.

According to an embodiment, the step of comparing the extracted word with the analyzed metadata and recommending a hashtag to be used for the produced video content corresponds to a category corresponding to the uploaded video content, and determining whether information extracted from a pre-registered image matches a word extracted from the extracted object and the image metadata; It may include a step of recommending a hashtag, and the step of recommending a hashtag for the images previously registered in the SNS when the determination result does not match.

According to an embodiment, when the determination result does not match, the step of recommending a hashtag for the images previously registered in the SNS considers the frequency of hashtags for the images previously registered in the SNS and may include a step of recommending it.

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment may further include uploading the selected hashtag from among the recommended hashtags to image content.

The step of extracting an object based on image recognition from the uploaded image content according to an embodiment includes extracting features from the uploaded image content, and detecting harmful content, celebrity identification, label recognition, facial property analysis, and scene detection. , facial recognition, and extracting an object including at least one of object recognition, and the number of objects per frame from the extracted object, location, size, object classification accuracy, emotion, age, gender, celebrity identification result, The method may include analyzing at least one of the object labeling result, gender, age group, glasses, and scene classification.

A trend analysis-based AI automatic hashtag recommendation method according to an embodiment includes collecting produced image content and metadata of the produced image content, including objects included in the produced image content and the collected metadata It may include extracting the extracted word, and recommending a hashtag in consideration of the frequency of the extracted word corresponding to the extracted object.

According to an embodiment, the step of extracting an object included in the produced image content and a word included in the collected metadata includes selecting a word included in a title or subtitle for the image content from the image metadata. and extracting words corresponding to the extracted object from among the selected words.

According to an embodiment, the step of recommending a hashtag in consideration of the frequency of the extracted word corresponding to the extracted object may include scaling the frequency of the extracted word, the word extracted from the object, and the Determining whether a word extracted from image metadata corresponds or not, recommending words as a hashtag for the image content if the words match as a result of the determination, and if the determination result does not match, the steps It may include the step of recommending a keyword that appears with a frequency greater than or equal to a selected criterion in the title or subtitle as a hashtag.

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment may further include uploading the selected hashtag from among the recommended hashtags to image content.

The step of extracting the object included in the produced image content according to an embodiment may include extracting features from the uploaded image content, detecting harmful content, celebrity identification, label recognition, facial property analysis, scene detection, face recognition , and extracting an object including at least one process among object recognition, the number of objects per frame, location, size, object classification accuracy, emotion, age, gender, celebrity identification result, object labeling result from the extracted object, It may include analyzing at least one of gender, age group, glasses, and scene classification.

The hashtag recommendation server according to an embodiment includes a data collection unit that collects produced image content and metadata of the produced image content, a preprocessor that performs preprocessing including stopword processing, and at least one or more from the produced image content. Extracting a word, analyzing the meta data for image content selected from images previously registered on SNS, corresponding to a category corresponding to the produced image content, and analyzing the extracted word and the analyzed meta data It may include a hashtag recommendation unit that compares the data and recommends a hashtag to be used for the produced image content.

The hashtag recommendation unit according to an embodiment extracts an object from the uploaded image content based on image recognition, selects a word included in a title or subtitle for the image content from the image metadata, and selects the Words corresponding to the extracted object may be extracted from among the extracted words.

The hashtag recommendation unit according to an embodiment includes information extracted from an image that corresponds to a category corresponding to the uploaded image content, and is pre-registered on SNS, and a word extracted from the extracted object and the image metadata. Determines whether or not matches with, and if the determination result matches, recommends words as a hashtag for the video content. You can recommend hashtags for

The hashtag recommendation unit according to an embodiment extracts a feature from the uploaded video content, and performs at least one of harmful content detection, celebrity identification, label recognition, face attribute analysis, scene detection, face recognition, and object recognition. Extracting an object including, and at least among the number of objects per frame, location, size, object classification accuracy, emotion, age, gender, celebrity identification result, object labeling result, gender, age group, glasses, and scene classification from the extracted object One can analyze

The hashtag recommendation server according to an embodiment includes a data collection unit that collects produced image content and metadata of the produced image content, a preprocessor that performs preprocessing including stopword processing, and and a hashtag recommender for extracting an object and a word included in the collected metadata, and recommending a hashtag in consideration of the frequency of the extracted word corresponding to the extracted object.

The hashtag recommendation unit according to an embodiment may select a word included in a title or subtitle for the video content from the image metadata, and extract words corresponding to the extracted object from among the selected words. have.

According to one embodiment, a technology for automatically recommending hashtags can be provided to help users more easily author and distribute content by developing a one-solution to solve the cultural and social needs of seniors.

According to an embodiment, it is possible to express the subject and individuality of one's own content, and an automatic hashtag function may be provided in order to be searchable after the content is posted.

According to an embodiment, it is possible to provide a user's content-based hashtag recommendation function using unsupervised learning, collaborative filtering, deep learning algorithm, and the like.

According to one embodiment, it is possible to satisfy the user's needs as much as possible by configuring the top 10 popular hashtags and hashtag recommendations based on the image recognition weighting method.

According to an embodiment, it is possible to derive key keywords by collecting metadata (title, subtitle, video) of video content produced by users using a video content authoring platform.

According to an embodiment, the keywords derived from the title and subtitles may be ranked by calculating the degree of association between the keywords and the objects detected in the image.

According to an embodiment, it is possible to identify key keywords from data input by the user and images uploaded by the user, and recommend appropriate hashtags to the user, thereby facilitating the use of hashtags.

According to one embodiment, the video content authoring platform administrator may be provided with a result for identifying the status of hashtags used in the corresponding platform.

1 is a diagram illustrating a hashtag recommendation server according to an embodiment.
2 is a diagram for explaining the characteristics of a trend analysis-based AI automatic hashtag recommendation system according to an embodiment.
3 is a diagram for explaining an idea for recommending a hashtag according to an embodiment.
4 is a view for explaining a trend analysis-based AI automatic hashtag recommendation method according to an embodiment.
5 is a view for explaining an embodiment of recognizing an object from image content and recommending a hashtag.
6 is a view for explaining a trend analysis-based AI automatic hashtag recommendation method according to another embodiment.
7 shows an MS Azure image recognition test result according to an embodiment.
8 shows a trend analysis-based AI automatic hashtag recommendation method according to another embodiment.

Specific structural or functional descriptions of the embodiments according to the concept of the present invention disclosed herein are only exemplified for the purpose of explaining the embodiments according to the concept of the present invention, and the embodiment according to the concept of the present invention These may be embodied in various forms and are not limited to the embodiments described herein.

Since the embodiments according to the concept of the present invention may have various changes and may have various forms, the embodiments will be illustrated in the drawings and described in detail herein. However, this is not intended to limit the embodiments according to the concept of the present invention to specific disclosed forms, and includes changes, equivalents, or substitutes included in the spirit and scope of the present invention.

Terms such as first or second may be used to describe various elements, but the elements should not be limited by the terms. The above terms are used only for the purpose of distinguishing one component from other components, for example, without departing from the scope of rights according to the concept of the present invention, a first component may be named a second component, Similarly, the second component may also be referred to as the first component.

When an element is referred to as being “connected” or “connected” to another element, it is understood that it may be directly connected or connected to the other element, but other elements may exist in between. it should be On the other hand, when it is said that a certain element is "directly connected" or "directly connected" to another element, it should be understood that the other element does not exist in the middle. Expressions describing the relationship between elements, for example, “between” and “between” or “directly adjacent to”, etc. should be interpreted similarly.

The terminology used herein is used only to describe specific embodiments, and is not intended to limit the present invention. The singular expression includes the plural expression unless the context clearly dictates otherwise. In the present specification, terms such as "comprise" or "have" are intended to designate that the described feature, number, step, operation, component, part, or combination thereof exists, and includes one or more other features or numbers, It should be understood that the possibility of the presence or addition of steps, operations, components, parts or combinations thereof is not precluded in advance.

Unless defined otherwise, all terms used herein, including technical or scientific terms, have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. Terms such as those defined in a commonly used dictionary should be interpreted as having a meaning consistent with the meaning in the context of the related art, and should not be interpreted in an ideal or excessively formal meaning unless explicitly defined in the present specification. does not

Hereinafter, embodiments will be described in detail with reference to the accompanying drawings. However, the scope of the patent application is not limited or limited by these embodiments. Like reference numerals in each figure indicate like elements.

1 is a diagram for explaining a hashtag recommendation server 100 according to an embodiment.

The hashtag recommendation server 100 according to the present invention can automatically recommend hashtags to help users more easily author and distribute content by developing a one-solution to solve the cultural and social needs of seniors. have.

In this specification, the characteristic configuration of the present invention is described as senior content, but the video content produced by content creators is not limited thereto, and various video content mainly consumed by a specific group such as a specific age group, subject, nationality, race, gender, etc. can be interpreted as

To this end, the hashtag recommendation server 100 according to an embodiment may include a data collection unit 110 , a preprocessor 120 , and a hashtag recommendation unit 130 .

The data collection unit 110 according to an embodiment may collect the produced image content and metadata of the produced image content.

As an example, the data collection unit 110 may collect image metadata of famous senior YouTubers such as Makrye and Milanona through an SNS content platform such as YouTube. The image metadata may include, for example, a title, date, info, hashtag, and comment for image content.

Meanwhile, the data collection unit may collect metadata including titles, subtitles, and other info along with the video content of the video producer.

The pre-processing unit 120 according to an embodiment may perform a pre-processing including stop-word processing.

The preprocessor 120 according to an embodiment may process hashtag analysis and stopword processing, stopwords, and profanity (such as profanity).

The hashtag recommendation unit 130 according to an embodiment may extract at least one word from the produced image content.

In addition, the hashtag recommendation unit 130 may analyze metadata for image content selected from among images pre-registered on SNS, corresponding to a category corresponding to the produced image content.

In addition, the hashtag recommendation unit 130 may recommend a hashtag to be used for the produced image content by comparing the extracted word with the analyzed metadata.

The hashtag recommendation unit 130 may extract an object from the uploaded image content based on image recognition, and select a word included in a title or subtitle for the image content from image metadata. Also, words corresponding to the object may be extracted or selected from among the selected words.

For example, the hashtag recommendation unit 130 may perform harmful content detection, celebrity identification, label recognition, face attribute analysis, scene detection, face recognition, object recognition, etc. through feature extraction of preprocessed image content. have.

For this, you can use the feature extraction algorithm provided by portals such as Google, AWS, Azure, NAVER, etc.

In addition, using KETI Basic API Server, the number of objects per frame, location/size, object classification accuracy, person recognition (emotion, age, gender), celebrity identification result, object labeling result, face analysis (gender/age age/glasses), Object analysis such as scene classification can be performed.

The hashtag recommendation unit 130 according to an embodiment checks whether a word extracted from the video content uploaded by the content author matches the hashtag obtainable from the video content uploaded to SNS such as youtube, and the matched video Frequency analysis can be performed on hashtags of

Also, based on the analyzed frequency, a keyword with a high frequency can be recommended.

If there is no matching keyword between the word extracted from the video content uploaded by the content author and the hashtag obtainable from the video content uploaded to the SNS, the hashtag that shows the most frequency in the entire video can be recommended.

As an example, the present invention can help users to more easily author and distribute content by developing a one-solution solution to solve the cultural and social needs of a specific group, such as seniors.

In addition, you can express the subject and personality of your content, and provide an automatic hashtag function to make it searchable after posting the content, and users who use unsupervised learning, collaborative filtering, deep learning algorithms, etc. of content-based hashtag recommendation function can be developed.

In addition, it is possible to satisfy the needs of users as much as possible by composing the top 10 popular hashtags and recommending hashtags using an image recognition weighting method.

2 is a diagram for explaining the characteristics of a trend analysis-based AI automatic hashtag recommendation system according to an embodiment.

Trend analysis-based AI automatic hashtag recommendation system according to an embodiment provides a hash related to the video to the video content creator 240 who creates video content on the video content authoring platform based on the data analysis result in the analysis server 230 By recommending tags, it is possible to provide a hashtag recommendation service that helps to classify videos by subject.

The image content authoring platform 200 to which the present invention is applied uses the server 210 to facilitate the use of hashtags by image content creators through services, and images (image A, image B, image C, etc.) By analyzing the title and subtitle, which are metadata 220 of , and recommending keywords related to the video, it is possible to induce the appropriate use of hashtags and to help search for content from other users.

Using object detection technology, it is possible to recommend hashtags by analyzing the correlation between detected objects and keywords appearing in titles and subtitles, so that hashtags with high relevance to video content can be recommended, and furthermore, video content authoring You can check the status of hashtags being used on the platform.

3 is a diagram for explaining an idea for recommending a hashtag according to an embodiment.

The image authoring platform may be provided through a hashtag recommendation server according to an embodiment.

In addition, the image authoring platform can derive key keywords by collecting meta data (title, subtitle, image, etc.) of the image content produced by the image meta DB 310 for users who use the image content authoring platform.

In addition, the image authoring platform may rank keywords derived from titles and subtitles through a degree of association with an object detected in an image through the object recognition algorithm 320 and natural language processing 330 for metadata.

As a specific example, the image authoring platform may apply an object recognition algorithm utilizing deep learning through the object recognition algorithm 320 , and may extract an object from the image content provided by the image content author.

The image authoring platform may extract nouns from title and subtitle data through morpheme analysis using the natural language processing 330 .

In this process, the image authoring platform may derive a weight through scaling of the frequency if the object detected in the image matches the title and the noun derived from the subtitle, and may assign it to each word or object (340) ).

In particular, the video authoring platform can identify key keywords from the data entered by the user and the uploaded video and recommend appropriate hashtags to the user to facilitate the use of hashtags, and for the video content authoring platform administrator, the hashes used on the platform It is possible to provide a result that can grasp the tag status.

4 is a view for explaining a trend analysis-based AI automatic hashtag recommendation method according to an embodiment.

In the trend analysis-based AI automatic hashtag recommendation method according to an embodiment, image metadata for the produced image content may be input (step 401).

Next, in the trend analysis-based AI automatic hashtag recommendation method according to an embodiment, video content may be uploaded (step 402).

The AI automatic hashtag recommendation method based on trend analysis according to an embodiment may select a word included in a title or subtitle for an image content from image metadata (step 403).

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment may extract an object based on image recognition from the uploaded image content (step 404).

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment extracts features from uploaded video content, and includes at least one of harmful content detection, celebrity identification, label recognition, face attribute analysis, scene detection, face recognition, and object recognition. An object can be extracted by including one process.

In addition, at least one of the number of objects per frame, location, size, object classification accuracy, emotion, age, gender, celebrity identification result, object labeling result, gender, age group, glasses, and scene classification may be analyzed from the extracted object.

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment may extract words corresponding to the extracted object from among the selected words. In addition, it is possible to analyze metadata for image contents selected from among images that correspond to a category corresponding to the produced image content and are pre-registered on SNS.

To this end, the trend analysis-based AI automatic hashtag recommendation method according to an embodiment may perform a process of analyzing a hashtag for an image previously registered in the SNS (step 405).

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment corresponds to a category corresponding to the uploaded video content, information extracted from an image previously registered in SNS, and the extracted object and the image metadata. It may be determined whether the extracted word matches or not (step 406).

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment may recommend as a hashtag for the video content for words if they match as a result of the determination (step 407), and if they do not match, the SNS It is possible to recommend hashtags for images previously registered in . (step 409). For example, the trend analysis-based AI automatic hashtag recommendation method according to an embodiment may recommend hashtags for images in consideration of the frequency of hashtags for images previously registered on SNS.

On the other hand, the trend analysis-based AI automatic hashtag recommendation method according to an embodiment may receive a selection input for at least one or more hashtags among the recommended hashtags (step 408), and select input from the recommended hashtags. The hashtag may be applied to the video content and uploaded (step 410).

5 is a view for explaining an embodiment of recognizing an object from image content and recommending a hashtag.

According to the trend analysis-based AI automatic hashtag recommendation method according to an embodiment, in order to recognize an object included in image content, a model for image analysis may be preferentially selected ( 510 ). For example, the trend analysis-based AI automatic hashtag recommendation method according to an embodiment may use an object recognition API provided by Google, AWS, AZURE, NAVER, and the like.

For this, a feature extraction algorithm provided by portals such as Google, AWS, Azure, NAVER, etc. can be used.

In addition, using KETI Basic API Server, the number of objects per frame, location/size, object classification accuracy, person recognition (emotion, age, gender), celebrity identification result, object labeling result, face analysis (gender/age age/glasses), Object analysis such as scene classification can be performed.

Next, an object may be analyzed using the selected model and a hashtag may be recommended based on this ( 520 ).

The video content authoring platform to which the AI automatic hashtag recommendation method based on trend analysis according to an embodiment is applied may receive the title and subtitle input from the user who produced the video, and upload the video content.

Accordingly, the image content authoring platform may extract (NLP) a word, for example, a noun from the title/subtitle data, and perform object detection from the image data.

In addition, by using the extracted noun and the extracted object, a hashtag can be recommended and inputted.

To this end, the trend analysis-based AI automatic hashtag recommendation method according to an embodiment may recommend a weighted top hashtag. In particular, by comparing and analyzing the frequency and match of the title/subtitle data NLP result and the image object recognition result, a hashtag can be finally recommended.

Weight calculation calculates the number from 0 to 1 through min-max scaling for the title/subtitle data, and calculates the number from 0 to 1 through the scaling on the image recognition result data. You can judge whether the value matches or not. Also, it is possible to calculate a value between 0 and 1 by multiplying the scaling result of the matched dictionary, and recommend top 10 hashtags.

6 is a diagram 600 for explaining a trend analysis-based AI automatic hashtag recommendation method according to another embodiment.

In the trend analysis-based AI automatic hashtag recommendation method according to an embodiment, video contents such as images and videos are input ( 610 ).

To this end, the trend analysis-based AI automatic hashtag recommendation method according to an embodiment may use a feature extraction algorithm provided by the portal server 620 of Google, AWS, Azure, NAVER, and the like.

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment, as shown in reference numeral 630, through feature extraction of pre-processed image content, harmful content detection, celebrity identification, label recognition, facial property analysis, scene detection, Face recognition, object recognition, etc. can be performed.

In addition, the trend analysis-based AI automatic hashtag recommendation method according to an embodiment uses the KETI Basic API Server as shown in reference numeral 640 to determine the number of objects per frame, location/size, object classification accuracy, and person recognition (emotion, age, Gender), celebrity identification results, object labeling results, face analysis (gender/age/glasses), and object analysis such as scene classification can be performed.

Next, as shown in reference numeral 650, the trend analysis-based AI automatic hashtag recommendation method according to an embodiment includes a word extracted from the video content uploaded by the content author, and a hash obtainable from the video content uploaded to SNS such as youtube. It is possible to check whether a tag matches or not, and perform frequency analysis on the hashtag of the matched video.

In addition, keywords with high frequency can be recommended based on the analyzed frequency. You can recommend the hashtags that show the most frequency in the video.

As an example, the present invention can help users to more easily author and distribute content by developing a one-solution solution to solve the cultural and social needs of a specific group, such as seniors.

7 shows an MS Azure image recognition test result 700 according to an embodiment.

The image content authoring platform to which the trend analysis-based AI automatic hashtag recommendation method according to an embodiment is applied may show the MS Azure image recognition test result 700 .

The image content authoring platform may output 29 labels for the input image content as examples of words for the image content uploaded by the user.

In particular, reference numeral 720 denotes a result of outputting image content uploaded to SNS with 29 labels.

In addition, the video content authoring platform can output multiple labels for video content created and uploaded by the user, and the output label can be used as a candidate for a hashtag in the future.

8 shows a trend analysis-based AI automatic hashtag recommendation method according to another embodiment.

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment may collect metadata of the produced video content (step 801).

In the trend analysis-based AI automatic hashtag recommendation method according to an embodiment, the produced video content may be uploaded and collected (step 802).

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment may extract words corresponding to the extracted object from among the selected words (step 803).

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment may extract nouns from titles or subtitles by analyzing metadata (step 804).

The trend analysis-based AI automatic hashtag recommendation method according to an embodiment may scale the frequency of the extracted word corresponding to the extracted object (step 805) in order to recommend the hashtag.

To this end, the trend analysis-based AI automatic hashtag recommendation method according to an embodiment may determine whether a word extracted from an object matches a word extracted from the image metadata (step 806).

Also, as a result of the determination, if the words match, it may be recommended as a hashtag for the video content (step 807).

Also, when the determination result does not match, a keyword that appears with a frequency greater than or equal to a selected criterion in the title or subtitle may be recommended as a hashtag (step 809).

In the trend analysis-based AI automatic hashtag recommendation method according to an embodiment, it is possible to finally recommend a hashtag by comparing and analyzing the frequency and the coincidence of the title/subtitle data NLP result and the image object recognition result. For this, the weight calculation calculates the number from 0 to 1 through min-max scaling for the title/subtitle data, and calculates the number from 0 to 1 through the scaling on the image recognition result data, It is possible to determine whether the two result values match. Also, it is possible to calculate a value between 0 and 1 by multiplying the scaling result of the matched dictionary, and recommend top 10 hashtags.

Thereafter, the trend analysis-based AI automatic hashtag recommendation method according to an embodiment selects a hashtag to match the video content based on the recommended high-frequency keyword or hashtag (step 808), and reflects the selected hashtag An image may be uploaded (step 810).

After all, by using the present invention, a technology that automatically recommends hashtags can be provided to help users more easily author and distribute content by developing a one-solution to solve the cultural and social needs of seniors. have.

In addition, they can express the subject and personality of their content, and provide an automatic hashtag function to make it searchable after content is posted, and users who use unsupervised learning, collaborative filtering, deep learning algorithms, etc. of content-based hashtag recommendation function can be provided.

In addition, it can satisfy users' needs as much as possible by composing the top 10 popular hashtags and recommending hashtags using the image recognition weighting method, and the metadata (title, title, Subtitles, videos) can be collected to derive key keywords.

In addition, through the present invention, it is possible to rank keywords by calculating the correlation between keywords derived from titles and subtitles and objects detected in images, and to identify key keywords from user input data and uploaded images. This makes it easy to use hashtags by recommending appropriate hashtags to users.

In addition, by using the present invention, it is possible to provide an image content authoring platform administrator with a result that can determine the status of hashtags used in the corresponding platform.

The device described above may be implemented as a hardware component, a software component, and/or a combination of the hardware component and the software component. For example, devices and components described in the embodiments may include, for example, a processor, a controller, an arithmetic logic unit (ALU), a digital signal processor, a microcomputer, a field programmable array (FPA), It may be implemented using one or more general purpose or special purpose computers, such as a programmable logic unit (PLU), microprocessor, or any other device capable of executing and responding to instructions. The processing device may execute an operating system (OS) and one or more software applications running on the operating system. A processing device may also access, store, manipulate, process, and generate data in response to execution of the software. For convenience of understanding, although one processing device is sometimes described as being used, one of ordinary skill in the art will recognize that the processing device includes a plurality of processing elements and/or a plurality of types of processing elements. It can be seen that can include For example, the processing device may include a plurality of processors or one processor and one controller. Other processing configurations are also possible, such as parallel processors.

Software may comprise a computer program, code, instructions, or a combination of one or more thereof, which configures a processing device to operate as desired or is independently or collectively processed You can command the device. The software and/or data may be any kind of machine, component, physical device, virtual equipment, computer storage medium or apparatus, to be interpreted by or to provide instructions or data to the processing device. , or may be permanently or temporarily embody in a transmitted signal wave. The software may be distributed over networked computer systems and stored or executed in a distributed manner. Software and data may be stored in one or more computer-readable recording media.

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 in a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc. alone or in combination. The program instructions recorded on the medium may be specially designed and configured for the embodiment, or may be known and available to those skilled in the art of computer software. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floppy disks. - includes magneto-optical media, and hardware devices specially configured to store and execute program instructions, such as ROM, RAM, flash memory, and the like. Examples of program instructions include not only machine language codes such as those generated by a compiler, but also high-level language codes that can be executed by a computer using an interpreter or the like. The hardware devices described above may be configured to operate as one or more software modules to perform the operations of the embodiments, and vice versa.

As described above, although the embodiments have been described with reference to the limited drawings, various modifications and variations are possible by those skilled in the art from the above description. For example, the described techniques are performed in an order different from the described method, and/or the described components of the system, structure, apparatus, circuit, etc. are combined or combined in a different form than the described method, or other components Or substituted or substituted by equivalents may achieve an appropriate result.

Therefore, other implementations, other embodiments, and equivalents to the claims are also within the scope of the following claims.

Claims (17)

extracting at least one word from the produced video content;
analyzing metadata for image contents selected from among images previously registered in SNS and corresponding to a category corresponding to the produced image content; and
comparing the extracted word with the analyzed metadata, and recommending a hashtag to be used for the produced video content
Trend analysis-based AI automatic hashtag recommendation method comprising a. According to claim 1,
The step of extracting at least one word from the produced video content,
receiving image metadata for the produced image content;
uploading the video content;
extracting an object from the uploaded image content based on image recognition;
selecting a word included in a title or subtitle for the video content from the video metadata; and
extracting words corresponding to the extracted object from among the selected words;
Trend analysis-based AI automatic hashtag recommendation method comprising a. 3. The method of claim 2,
The step of comparing the extracted word with the analyzed metadata and recommending a hashtag to be used for the produced video content comprises:
determining whether information extracted from an image corresponding to a category corresponding to the uploaded image content and registered in SNS matches the extracted object and words extracted from the image metadata;
as a result of the determination, recommending words as a hashtag for the video content if they match; and
Recommending a hashtag for the images previously registered in the SNS when the determination result does not match
Trend analysis-based AI automatic hashtag recommendation method comprising a. 4. The method of claim 3,
If the determination result does not match, recommending a hashtag for the images previously registered on the SNS includes:
Recommending in consideration of the frequency of hashtags for images previously registered on the SNS
Trend analysis-based AI automatic hashtag recommendation method comprising a. According to claim 1,
Uploading the selected hashtag from among the recommended hashtags to the video content
Trend analysis-based AI automatic hashtag recommendation method that further includes. 3. The method of claim 2,
The step of extracting an object based on image recognition from the uploaded image content,
extracting an object including at least one of harmful content detection, celebrity identification, label recognition, face attribute analysis, scene detection, face recognition, and object recognition by extracting features from the uploaded video content; and
Analyzing at least one of the number of objects per frame, location, size, object classification accuracy, emotion, age, gender, celebrity identification result, object labeling result, gender, age group, glasses, and scene classification from the extracted object
Trend analysis-based AI automatic hashtag recommendation method including collecting the produced image content and metadata of the produced image content;
extracting an object included in the produced image content and a word included in the collected metadata;
Recommending a hashtag in consideration of the frequency of the extracted word corresponding to the extracted object
Trend analysis-based AI automatic hashtag recommendation method comprising a. 8. The method of claim 7,
The step of extracting the object included in the produced image content and the word included in the collected metadata includes:
selecting a word included in a title or subtitle for the video content from the video metadata; and
extracting words corresponding to the extracted object from among the selected words;
Trend analysis-based AI automatic hashtag recommendation method comprising a. 9. The method of claim 8,
The step of recommending a hashtag in consideration of the frequency of the extracted word corresponding to the extracted object,
scaling the frequency of the extracted word;
determining whether the word extracted from the object corresponds to the word extracted from the image metadata;
as a result of the determination, recommending words as a hashtag for the video content if they match; and
Recommending, as a hashtag, a keyword that appears with a frequency greater than or equal to a criterion selected in the title or subtitle when they do not match as a result of the determination
Trend analysis-based AI automatic hashtag recommendation method comprising a. 8. The method of claim 7,
Uploading the selected hashtag from among the recommended hashtags to the video content
Trend analysis-based AI automatic hashtag recommendation method that further includes. 8. The method of claim 7,
The step of extracting the object included in the produced image content,
extracting an object including at least one of harmful content detection, celebrity identification, label recognition, face attribute analysis, scene detection, face recognition, and object recognition by extracting features from the uploaded video content;
Analyzing at least one of the number of objects per frame, location, size, object classification accuracy, emotion, age, gender, celebrity identification result, object labeling result, gender, age group, glasses, and scene classification from the extracted object
Trend analysis-based AI automatic hashtag recommendation method including a data collection unit for collecting the produced image content and metadata of the produced image content;
a pre-processing unit that performs pre-processing including stop-word processing; and
Extracting at least one word from the produced image content, analyzing metadata for image content selected from among images previously registered on SNS, corresponding to a category corresponding to the produced image content, and the extraction A hashtag recommendation unit that compares the analyzed word with the analyzed metadata and recommends a hashtag to be used for the produced video content.
A hashtag recommendation server comprising a. 13. The method of claim 12,
The hashtag recommendation unit,
Extracts an object from the uploaded image content based on image recognition, selects a word included in a title or subtitle for the image content from the image metadata, and selects a word corresponding to the extracted object from among the selected words A hashtag recommendation server that extracts words. 13. The method of claim 12,
The hashtag recommendation unit,
It is determined whether information extracted from an image corresponding to the category corresponding to the uploaded image content and registered in SNS matches the word extracted from the extracted object and the image metadata, and the determination result , recommending as a hashtag for the video content for words when they match, and recommending a hashtag for images previously registered in the SNS when they do not match as a result of the determination Tag recommendation server. 13. The method of claim 12,
The hashtag recommendation unit,
Extracting features from the uploaded image content, extracting an object including at least one process of harmful content detection, celebrity identification, label recognition, face attribute analysis, scene detection, face recognition, and object recognition, Recommendation of a hashtag characterized by analyzing at least one of the number of objects per frame, location, size, object classification accuracy, emotion, age, gender, celebrity identification result, object labeling result, gender, age group, glasses, and scene classification from the object server. a data collection unit for collecting the produced image content and metadata of the produced image content;
a pre-processing unit that performs pre-processing including stop-word processing; and
A hashtag recommendation unit that extracts an object included in the produced image content and a word included in the collected metadata, and recommends a hashtag in consideration of the frequency of the extracted word corresponding to the extracted object
A hashtag recommendation server comprising a. 17. The method of claim 16,
The hashtag recommendation unit,
The hash tag recommendation server, characterized in that the selected words included in the title or subtitle of the video content are selected from the video metadata, and words corresponding to the extracted object are extracted from the selected words.

Images