KR20230168534A - Contents Recommending method - Google Patents

Abstract

In addition to the content provided according to the category to which the user belongs, the present invention provides content recommendation that allows the user to access and utilize user interest information and corresponding content of others belonging to the same category as the user, thereby allowing the user to access a wider range of actual interests. suggest a method To this end, the present invention is performed in a content server connected to a user terminal and a network, and a) a first hashtag assigned to content created or consumed in the user terminal and an object analysis of the image included in the content are determined. Extracting a second hashtag; b) mapping the user terminal to a Maslow category using the first hashtag and the second hashtag; c) extracting user interest information about the user terminal by referring to at least one of a first hashtag and a second hashtag of another user terminal classified into the same Maslow category as the user terminal; and d) extracting new interest content corresponding to the user interest information, generating matching content including the new interest content and existing interest content according to the Maslow category of the user terminal, and providing the matching content to the user terminal; It may be configured to include.

Description

Content Recommending Method {Contents Recommending method}

The present invention relates to a content recommendation method. More specifically, the present invention relates to a content recommendation method, and more specifically, after extracting content based on hashtags, new interests and corresponding new content are provided to the user by referring to user interest information of other users classified in the same Maslow category. It concerns a content recommendation method that provides users with access to various content and experiences.

Typically, a content extraction method using a hashtag refers to a hashtag given by the user to content, such as a web page, web image, or text, and extracts other content corresponding to the hashtag to the user. The extracted content is registered as the user's interest information, and when the content server (or web server, hereinafter omitted) extracts content that matches the user's interest information based on the user's interest information, the user's interest is extracted from the content server. Information has the characteristic that its categories rarely change and have a certain tendency. Additionally, the hashtags that users send to the content server accumulate over a long period of time, so their trends may not change easily.

For example, in the case of a user who is mainly active in a club, the contents of the posts or comments on the posted posts are focused on the content of the user's activities in the club, so most of the user's interest information is related to the club. It can be.

In addition, users who upload beauty-related content on social platforms such as Facebook mainly create or consume very narrow content such as cosmetics or clothes, so when viewed from the content server side such as Facebook, there is no beauty-related content. It can be seen that the user's category remains almost constant.

At first glance, the user's interest information extracted based on the amount of activity the user uploads to the content server or communicates through other users' content is reasonable. However, it is impossible for a user who is active in a club to be indifferent to anything other than club activities, and it is impossible for a woman who is interested in beauty-related content on a social platform such as Facebook to be indifferent to anything other than beauty. For example, a user related to beauty may be interested in good restaurants or may have interest and affection for pets.

Accordingly, the present applicant considered a method of extracting various areas of interest desired by actual users by utilizing user interest information of similar types of users and providing content to users according to the extracted user interest information.

Republic of Korea Patent No. 10-2007504 (System and method for providing customized content information for each user) Republic of Korea Registered Patent No. 10-1754371 (Method of providing tagged social network service posting content)

The purpose of the present invention is to provide a content recommendation method that provides new user interest information to users with certain tendencies by referring to the interest information of other users belonging to the same category, and allows users to encounter new content through this.

Another purpose of the present invention is to provide users with a mixture of new content and matching content so that user interest information is calculated cumulatively over time so that users are not fixed on a specific type, and content that finds new interests of users based on user responses to this. It provides recommended methods.

According to the present invention, the above purpose is performed in a content server connected to a user terminal and a network, and a) object analysis of the first hashtag assigned to the content created or consumed in the user terminal and the image included in the content. Extracting a second hashtag determined through; b) mapping the user terminal to a Maslow category using the first hashtag and the second hashtag; c) extracting user interest information about the user terminal by referring to at least one of a first hashtag and a second hashtag of another user terminal classified into the same Maslow category as the user terminal; and d) extracting new interest content corresponding to the user interest information, including it in matching content according to the Maslow category, and providing it to the user terminal.

Here, step d) may further include reclassifying the new interest content into the matching content when the new interest content is repeatedly created or consumed in the user terminal more than a preset number of times.

Here, after the step of reclassifying the new content into the matching content, the step of extracting new new content through steps a) and b) may be further included.

Here, the new interest information may be extracted by referring to the first and second hashtags commonly included in at least one other user terminal that matches the Maslow category of the user terminal among the Maslow categories.

Here, the new interest information is preferably extracted through synonym comparison of the first and second hashtags extracted from the other user terminals.

Here, the Maslow category may include at least one subcategory with reference to user information of the user terminal.

Here, the user information may include at least one of age, gender, and activity area.

According to the present invention, in addition to the content provided according to the user's own category, the user accesses user interest information and corresponding content of others belonging to a similar category to the user, thereby enabling the user to access a wider range of actual interests.

According to the present invention, when new content is repeatedly selected by the user, it is classified as existing content of interest, and the user's new interest information and corresponding new content are continuously extracted and provided to continuously expand the user's experience and interests. there is.

Figure 1 shows a conceptual diagram of a content recommendation method according to an embodiment of the present invention.
Figure 2 shows a conceptual diagram of a method for generating a second hashtag based on an image included in content.
Figure 3 shows a conceptual diagram of a content server according to an embodiment.
Figure 4 shows a conceptual diagram of a method for renewing new content in a content server according to an embodiment of the present invention.

The user terminal referred to in this specification may refer to a device that is capable of wired and wireless network communication, including a smartphone, tablet, laptop, or personal computer, and that can display and play content on a screen (or touch screen). In addition to those mentioned in this specification, any device capable of wired and wireless network communication and playback of video and audio may be a user terminal referred to in this specification. However, for convenience of explanation and understanding, the user terminal is shown and explained as a portable device such as a smartphone.

User interest information referred to in this specification is described as either “user interest information” or “user interest information” depending on the position and context in the sentence, and may be used interchangeably.

The content server referred to in this specification may include a social network service (SNS) server, a portal server, a web server, and a content provision server that provides web- or app-based image or video content. A content server refers to any type of server that can provide the content requested by the user terminal to the user terminal. The content desired by the user includes text, images, videos, links, and streaming links, and is created by selective combination of these. It could be content. Accordingly, the content server mentioned in this specification may be any type of server that provides such content, and is not specifically limited.

Hereinafter, the present invention will be described in detail with reference to the drawings.

Figure 1 shows a conceptual diagram of a content recommendation method according to an embodiment of the present invention.

Referring to FIG. 1, the content recommendation method according to the embodiment is implemented through the interaction between user terminals 10a to 10n and a content server 100 connected to a network, and the content server 100 is a user connected to a wired or wireless network. User terminals 10a to 10n can be classified according to Maslow categories with reference to what content the terminals 10a to 10n generate and consume.

Maslow's categories divide human needs into seven categories, and users can be classified into at least one or two of the seven Maslow categories.

Maslow's categories are:

1) Biological and physical needs

2) Need for safety

3) Need for belonging and love

4) Need for respect

5) Cognitive needs

6) Aesthetic needs

7) It is divided by the desire for self-actualization, and the content server 100 classifies each user terminal (10a to 10n) into Maslow categories through the content consumed and generated by the user terminals (10a to 10n).

For example, assuming that content with hashtags about food and restaurants is created on a user terminal (any one of 10a to 10n), and that feedback (comments) for the content and responses to the feedback (replies) are active, The content server 100 may classify users of the corresponding user terminal (any one of 10a to 10n) into Maslow's categories: 1) biological and physical needs, and 3) needs for belonging and love.

The act of creating content about food or restaurants and uploading it to the content server, as well as providing feedback and responses to the uploaded content, correspond to the act of interacting with others who have similar interests in food or restaurants, item 3) , and the content about food or restaurants itself falls under item 1). In addition, depending on the action taken through the user terminal (any one of 10a to 10n), the content may further include the remaining Maslow categories. For example, 4) or 7) of Maslow's categories may be further included depending on the relationship between the content creator and the content consumer (other person) or the act of mutual respect (recommendation of the other's content) and the frequency of the act. According to this method, the content server 100 maps each user terminal (10a to 10n) to a plurality of Maslow categories. After mapping, the content server 100 combines Maslow categories + user interest information and filters it with user information such as activity area, age, and gender for each user terminal 10a to 10n to extract similarities between users. You can.

For example, even for users who are very interested in clothes, men's clothes and women's clothes are different, even for the same summer clothes, the shapes of clothes people are interested in are different for each age group, and even if the interest in clothes stores is the same, there are two stores located in Seoul and Busan. Since the locations and inventory products of outlets or stores are differentiated, filtering to match similarities for each user terminal 10a to 10n may be required in order to match similar interests between different users.

This filtering is required when the content server 100 provides content according to other people's interests to the user terminal (any one of 10a to 10n) and will be described in detail later.

The Maslow categories and user interest information described above can be extracted through hashtags included in content consumed by users and content created by users. Hashtags are given when creating content, and in the present invention, the hashtag given when creating content will be referred to as a first hashtag or content hashtag and described.

The first hashtag assigned to the content includes information about the main content of the content. However, it is difficult to view that the first hashtag assigned according to the user's subjectivity in the user terminals 10a to 10n contains everything about the content. Since the first hashtag is assigned from the user's perspective, various information included in the content may be omitted, and there is always a possibility that only what the user thought of will be registered or omitted.

Accordingly, the present applicant considered a method of extracting another hashtag through images included in the content. Hashtags extracted through images are referred to as second hashtags or image hashtags and explained.

The content server 100 may extract a second hashtag from an image included in the content and classify the user terminals 10a to 10n according to Maslow categories by referring to the first and second hashtags.

In one embodiment, the content server 100 uses hashtags for a plurality of users as learning data, and determines the Maslow category for each of the plurality of users as a result of a psychological test for each of the plurality of users. An artificial neural network is trained using the training data as a label to create a Maslow category classification model, and the first and second hashtags are extracted for the user terminals 10a to 10n from the created Maslow category classification model. By entering , the user terminals 10a to 10n can be classified according to Maslow categories.

Therefore, the content server 100 provides the user with new content extracted based on the hashtags of others belonging to the same category as well as content based on the user's own hashtags of the user terminals 10a to 10n, thereby increasing the user's interest. You can continue to be exposed to other content worth having. This will be explained in more detail with reference to FIG. 2.

Figure 2 shows a conceptual diagram of a method for generating a second hashtag based on an image included in content.

Referring to FIG. 2 together, images (image 1 to image 3) included in the content may include various information that the user is not aware of and has not registered as the first hashtag.

For example, assuming that the image (image 1) is an image taken of an iPhone with a bed in the background, and that the iPhone is registered as the first hashtag in the image (image 1), the content server 100 Through image analysis, the bed can be extracted as a second hashtag. Even if iPhone is selected as the first hashtag, it can be determined that the user is a person who works at a bed furniture store or is a user interested in purchasing a bed, and the content server 100 uses “bed” as the second hashtag. It is desirable to estimate the Maslow category by extracting and combining it with the first hashtag.

Image extraction is derived by the process of extracting objects included in the image, and this object extraction program can be implemented by a currently commercialized artificial intelligence (AI) object recognition program. The point is to extract the objects included in the images (image 1 to image 3) without specifying the user's interest information by relying on the first hashtag assigned to the image (image 1 to image 3) to additionally extract the user's interest information. I'm doing it.

If an object corresponding to a bed is continuously extracted from images provided from the user terminals 10a to 10n to the content server 100, the content server 100 assigns weight to the second hashtag, “bed” and It may be determined that the user terminal (any one of 10a to 10n) is closely related to the furniture. Of course, it is preferable that weights are assigned when the same object (such as a bed) is identified at different times, and the weight value is increased in proportion to the number of times the same object (such as a bed) is repeated over a long period of time. It is desirable to be

Meanwhile, hashtags can be recognized as the same word through synonym comparison. For example, assuming that the hashtag “bed” is obtained through the second hashtag, and that the user terminal (any one of 10a to 10m) assigns the hashtag “furniture” to the content, the user content server 100 The first hashtag (furniture) and the second hashtag (bed) given by the user terminal (any one of 10a to 10m) may be recognized and processed as equivalent.

Hashtag comparison through this equivalent comparison unifies hashtags that can be identified differently into furniture, which is a higher concept compared to beds, thereby converting the user's interest information from the perspective of furniture rather than individual beds and other furniture (tables, chairs, desk, etc.).

In FIG. 2, it can be seen that user terminals 10a to 10n are mapped to users represented by A to I, and user terminals defined by A to D belong to the same category (category 1).

Among the user terminals (A to D) belonging to the same category, the user terminal (A) includes hashtags such as Gangnam, tteokbokki, and ramen, and the user terminal (C) includes hashtags such as Gangnam, restaurant, and delicious restaurant.

Hashtags included in user terminals (A to D) include at least one first hashtag and one second hashtag.

The user terminals (A, C) included in category 1 have the same activity area (Gangnam) and can be seen as being interested in food. If user interest information is extracted only for the user terminal (A), the Gangnam area Content for the user terminal (A) may be extracted mainly from the snack restaurant and provided to the user terminal (A).

However, it can be seen that the hashtags extracted from other user terminals (C) within the same category, Category 1, include hashtags that are slightly different from snack foods but are related to food.

Considering that the user terminal (C) is included in the same category as the user terminal (A), the content server 100 may include some of the user interest information included in the user terminal (C) in the interest information of the user (A). You can.

In this case, the user terminal (A) can receive from the content server 100 information on restaurants or restaurants in the Gangnam area in addition to snack-related content such as tteokbokki or ramen in the Gangnam area.

This was not included in the hashtag of the user terminal (A), and even if the user terminal (A) did not request it from the content server 100, when it was provided to the user terminal (A), the user of the user terminal (A) was sufficiently interested. It can be viewed as accessible content. In the present invention, the process of acquiring content based on user interest information of others belonging to a similar category rather than based on one's own hashtag is referred to as content expansion.

By expanding the content described above, the possibility of being exposed to other content similar to the user and therefore of interest to the user increases in addition to the hashtags corresponding to the user (the first hashtag and the second hashtag).

This grouping technique expands one user's interest information by exchanging influence with various user interest information belonging to the group, allowing the user's interest to continuously expand and attract the user's attention. This will be explained with reference to Figure 3.

Figure 3 shows a conceptual diagram of the content server 100 according to an embodiment.

Referring to FIG. 3 together, the content server 100 may be configured to include an image extraction unit 110, a hashtag extraction unit 120, and a category classification unit 140.

The image extraction unit 110 is performed for each user terminal 10a to 10n and can extract an object from an image included in content created or consumed by each user terminal 100. After object extraction, a name is given to the extracted object, and the extracted name corresponds to the second hashtag.

The hashtag extraction unit 120 extracts the first hashtag assigned to the content by each user through each user terminal (10a to 10n) and obtains the second hashtag extracted from the image extraction unit 110. The first and second hashtags extracted from the hashtag extraction unit 120 are provided to the category classification unit 140, and the category classification unit 140 provides unique information (A1 to A1) to the user terminals 10a to 10n. By assigning N3), information about the user and the user terminals 10a to 10n can be managed in a hidden state so that personal information is not exposed.

The filter unit 130 divides each user terminal (10a to 10n) into a primary hashtag and a secondary hashtag, and then divides the user terminals (10a to 10n) into user information, such as age, gender, and activity area. Accordingly, the category classification for user terminals 10a to 10n can be subdivided.

For example, even if a plurality of user terminals 10a to 10n belong to one category (P1), they can be grouped according to user information and subdivided into P1, P1', and P1". That is, the categories (P1', P1 ") corresponds to a subcategory comprised within a single category (P1). The subcategory (P1', P1") is essentially the same as the upper category (P1), but corresponds to an expanded type to form the same group. do.

The category classification unit 140 groups the user terminals 10a to 10n with reference to the upper category and sub-category filtered by the filter unit 130. At this time, the grouping includes the first hashtag, the second hashtag, and each It is processed with reference to the behavior patterns of the user terminals 10a to 10n, and each user terminal 10a to 10n can be given unique information (for example, reference numerals A1 to N3).

Each user terminal (10a to 10n) is given unique information (A1 to N3), each unique information is classified into upper categories (P1 to Pn), and the user information of the user using the user terminals (10a to 10n) is provided. For reference, a higher-level category may include at least one sub-category.

Each user terminal (10a to 10n) belonging to the upper category (P1 to Pn) and the subcategories belonging to each upper category (P1 to Pn) includes its own unique hashtag and user interest information.

User terminals (e.g., A1 to A3) belonging to the same category (parent category and subcategories for each parent category) are likely to be a group of users with similar interests, and the similarity between these groups means that other people's interests are my interests. It means that it can be.

For example, when user interest information and corresponding content of the user terminal A1 are provided to the user terminal A2, there is a high possibility that the user terminal A2 will be interested in consuming them. On the other hand, when the user interest information and the corresponding content of the user terminal A1 are provided to the user terminal B3, the possibility of them being consumed in the user terminal B3 is very low.

Similarly, when user interest information and corresponding content of a user terminal (A1) belonging to the category (P1) are provided to the category (P1') or category (P1"), the user terminal belonging to the category (P1', P1") The probability of the content being consumed may decrease. Differences such as gender and age make it difficult to obtain the same response to the same topic.

Figure 4 shows a conceptual diagram of a method for renewing new content in a content server according to an embodiment of the present invention.

Referring to FIG. 4, the content server 100 according to the embodiment extracts user interest information generated by the Maslow categories explained through FIGS. 1 to 3 and subcategories for each category, and creates new interests matching the Maslow categories. Matching content is created by combining content and existing content of interest of the user terminal 10a, and the matching content is provided to the user terminal 10a.

Here, only a portion of the matching content (S0) is assigned to the new interest content (S1), and most of it is assigned as classified according to Maslow's categories. The new interest content (S1) is the first hashtag and the first hashtag of the user terminal (10a). 2 It is not extracted by hashtags, but by user interest information of other users included in the same category (or subcategory). In other words, part (S1) of the matching content (S0) is not content provided through the hashtag of the user terminal (10a).

When the user terminal (10a) views and participates in the matching content (S0) and consumes the new content of interest (S1), the content server 100 determines the number and frequency of the new content of interest (S1) consumed by the user terminal (10a). can be counted.

For example, when user interest information of other user terminals belonging to the same Maslow category and a number of new interest contents corresponding thereto are provided to the user terminal 10a, and the user terminal 10a selects and consumes the content, the content server ( 100) may cumulatively count the consumption number and consumption frequency of user interest information corresponding to new interest content (S1). If the preset number and consumption frequency are satisfied, the content server 100 converts the new content of interest (S1) into the existing content of interest, and then searches the user interest information of another user terminal to create a new content of new interest (S2). is created and new matching content (S0) including it is created and provided to the user terminal (10a). In other words, the matching content (S0) is not fixed once and the content server 100 continuously searches for new content and user interest information that has not been consumed in the user terminal (10a) and continuously updates new matching content. There is.

Meanwhile, an individual's interests may continuously change from one field to another. In other words, it was information of personal interest in the past, but may no longer be information of interest in the present, and a new field may be information of interest.

Therefore, in one embodiment, the content server 100 determines the number and frequency of new content of interest (S1) included in the matching content (S0) and provided to the user terminal (10a) selected and consumed by the user terminal (10a). If it exceeds the preset first threshold, the new content of interest (S1) is not only converted to content of existing interest, but also the number and frequency of being selected and consumed by the user terminal 10a among the existing content of interest is set to a second threshold. If it does not meet the requirements, delete the relevant content from the existing content of interest, search the user interest information of other user terminals belonging to the same Maslow category, create new new content of interest (S2), and create new matching content (S0) including it. This can be provided to the user terminal (10a).

Alternatively, depending on the embodiment, in order to recommend various contents to the user by deleting past interest information and focusing on current interests, the content server 100 may provide information generated for the user terminal 10a before a predetermined period of time. Deleting the first hashtag and the second hashtag and remapping the user terminal 10a to the Maslow category using the first hashtag and the second hashtag generated for the user terminal 10a after the predetermined period of time. The operation can also be performed periodically.

Therefore, the content recommendation method according to embodiments of the present invention can effectively expand the user's experience and interests by continuously exposing the user to content that the user is interested in or is likely to be interested in at the current time.

By applying the present invention to a content provision server, the present invention can be usefully used to enable users to access a wider range of real-life interests and to continuously expand the user's experience and interests.

As described above, the present invention has been described with reference to preferred embodiments, but those of ordinary skill in the art may vary the present invention without departing from the spirit and scope of the present invention as set forth in the claims below. You will understand that it can be modified and changed.

100: content server 110: image extraction unit
120: Hashtag extraction unit 130: Filter unit
140: Category classification unit

Claims (9)

It is performed on a content server connected to the user terminal and network.
a) extracting a first hashtag assigned to content created or consumed in the user terminal and a second hashtag determined through object analysis of an image included in the content;
b) mapping the user terminal to a Maslow category using the first hashtag and the second hashtag;
c) extracting user interest information about the user terminal by referring to at least one of a first hashtag and a second hashtag of another user terminal classified into the same Maslow category as the user terminal; and
d) extracting new interest content corresponding to the user interest information, generating matching content including the new interest content and existing interest content according to the Maslow category of the user terminal, and providing the matching content to the user terminal; A content recommendation method comprising: According to paragraph 1,
A content recommendation method further comprising reclassifying the new interest content into the existing interest content when the new interest content is repeatedly created or consumed in the user terminal more than a preset number of times. According to paragraph 2,
After the step of reclassifying the new content of interest to the existing content of interest,
A content recommendation method further comprising extracting new content corresponding to the user interest information through steps a) to c). According to paragraph 2,
Content further comprising the step of deleting the part of the existing content of interest from the existing content of interest when the part of the content of existing interest is generated or consumed less than a minimum number of times during a predetermined period of time in the user terminal. Recommended method. According to paragraph 1,
The first hashtag and the second hashtag extracted for the user terminal before a predetermined period of time are deleted, and the first hashtag and the second hashtag extracted for the user terminal after the predetermined period of time are deleted. A content recommendation method further comprising: periodically remapping the user terminal to Maslow categories using tags. According to paragraph 1,
The user interest information is,
A content recommendation method, characterized in that the content is extracted with reference to a first hashtag and a second hashtag commonly included in at least one other user terminal that matches the Maslow category of the user terminal among the Maslow categories. According to clause 6,
The user interest information is,
A content recommendation method characterized in that synonym comparison is performed on the first and second hashtags extracted from the other user terminals, and the hashtags corresponding to the synonyms are treated as equivalent to each other and then extracted. According to paragraph 1,
The Maslow categories are:
A content recommendation method characterized by being subdivided into at least one sub-category with reference to user information of the user terminal. According to clause 8,
The user information is,
A content recommendation method comprising at least one of age, gender, and activity region.