KR20210063665A - Recommendation item based on user event information and apparatus performing the same - Google Patents

Abstract

A user device according to the present invention includes: an online market item database management unit in which a user event information-based recommended item provision server extracts, regarding an item purchasable in an online market through an online market application, a label that explains the content of the item on the basis of an image of the item, and indexes the label and stores the label in an online market item database; a style database management unit which analyzes a style image in which a person wears a plurality of fashion items and stores the style image in a style base; a recommended item searching unit which, when receiving per-user behavior signals generated on a plurality of web pages displayed through the online market application, extracts images of the items registered on the web pages in which the behavior signals are generated, then analyzes the images, and extracts recommended items on the basis of the online market item database and the style database according to whether the images are coordinated with another item; and a recommended item provision unit which provides the recommended items according to the types of the behavior signals.

Description

A method for providing a recommended item based on user event information and a device for executing the same {RECOMMENDATION ITEM BASED ON USER EVENT INFORMATION AND APPARATUS PERFORMING THE SAME}

The present invention relates to a method for providing a recommended item based on user event information and an apparatus for executing the same, and more particularly, to a user event information-based recommendation item that provides an item similar to an item registered on a web page when an action signal occurs on a web page It relates to a providing method and an apparatus for implementing the same.

In the background of the recently increased wired and wireless Internet environment, commerce such as public relations and sales using online is being activated. In this regard, when buyers find a product they like while searching for a magazine, blog, or YouTube video on a desktop or mobile terminal connected to the Internet, they search for a product name, etc. and make a purchase.

An example is the case where the name of a bag that a famous actress heard at the airport and the name of a childcare product appearing on an entertainment program rank high in real-time search terms on a portal site. However, in this case, the user has to separately open a web page for the search to search for a product name, manufacturer, vendor, etc., and there is an inconvenience in that it is not easy to search unless clear information about them is already known.

On the other hand, sellers spend a lot of money on media sponsorships and online user recruitment in addition to commercial advertisements to promote their products. This is because word of mouth online recently acts as an important variable in product sales. However, it is often not possible to disclose shopping information, such as product name and vendor, despite the cost of publicity. This is because indirect advertising issues may arise as it is not possible to individually obtain prior approval from media viewers for product name exposure.

As described above, there is a need for both users and sellers to provide shopping information in a more intuitive UI (User Interface) environment for online product images.

According to the present invention, when an action signal is generated on a web page, if an item is coordinated with another item based on an image registered on the web page, another item coordinated with the item is displayed on a part of the web page and recommended to the user. An object of the present invention is to provide a method for providing a recommended item based on user event information that allows easy access to an item that has been made, and an apparatus for executing the same.

In addition, in the present invention, when an action signal is generated on a web page, if an item is not coordinated with other items based on an image registered on the web page, a recommended style suitable for the item among pre-generated styles is determined and then sold in the online market. An object of the present invention is to provide a method for providing a recommended item based on user event information and an apparatus for executing the same by extracting an item corresponding to a recommended style from among products and displaying it on a part of a web page and recommending it so that a user can purchase the product together .

In addition, the present invention analyzes a user's behavioral signal when a behavioral signal is generated on a web page and recommends a product that the user may be interested in when the user accesses the online market server. An object of the present invention is to provide a method for providing a recommended item based on user event information that allows easy access to a desired product by moving, and an apparatus for executing the same.

The objects of the present invention are not limited to the above-mentioned objects, and other objects and advantages of the present invention that are not mentioned can be understood by the following description, and will be more clearly understood by examples of the present invention. In addition, it will be easily understood that the objects and advantages of the present invention can be realized by the means shown in the claims and combinations thereof.

In order to achieve this purpose, the user device provides a user event information-based recommended item providing server for an item that can be purchased in the online market through the online market application, a label describing the content of the item based on the image of the item. An online market item database management unit that extracts, indexes the label and stores it in an online market item database, a style database management unit that analyzes and stores a style image in which a person is wearing a plurality of fashion items, and an online market application. When a user-specific action signal generated on a plurality of displayed web pages is received, an image of an item registered in the web page in which the action signal is generated is extracted, and then the image is analyzed to determine whether it is coordinated with other items. and a recommended item search unit for extracting a recommended item based on an online market item database and the style database, and a recommendation item providing unit for providing the recommended item according to a type of the action signal.

In addition, the method for providing a recommended item based on user event information executed in a user device for achieving this purpose describes the content of the item based on the image of the item for an item that can be purchased in the corresponding online market through the online market application extracting a label, generating an online market item database by indexing the label; generating a style database by analyzing a style image in which a person is wearing a plurality of fashion items; Receiving a behavior signal for each user generated on a web page of , extracting an image for an item registered in a web page where the behavior signal is generated, and analyzing the image to determine whether it is coordinated with other items step, according to the result of the check, searching for other items coordinated with the item in the online market item database and extracting it as a recommended item or selecting a recommended item matching the item based on the online market item database management unit and the style database management unit and extracting and providing the recommended item according to the type of the action signal.

According to the present invention as described above, when an action signal is generated on a web page, if an item is coordinated with another item based on an image registered on the web page, the other item coordinated with the item is displayed on a part of the web page. By recommending it, there is an advantage that the user can easily access the already coordinated items.

Also, according to the present invention, when an action signal is generated on a web page, if an item is not coordinated with other items based on an image registered on the web page, a recommended style suitable for the item is determined from among pre-generated styles and then sold in the online market. There is an advantage in that it is possible to induce users to purchase the product together by extracting an item corresponding to the recommended style from among the available products and displaying it on a part of the web page for recommendation.

In addition, according to the present invention, when a behavior signal is generated on a web page, the user's behavior signal is analyzed and when the user accesses the online market server, the product that the user may be interested in is recommended first, so that the user can It has the advantage of being able to easily access the desired product with little movement.

1 is a network configuration diagram illustrating a system for providing a recommended item based on user event information according to an embodiment of the present invention.
2 is a block diagram illustrating an internal structure of a user device according to an embodiment of the present invention.
3 is a flowchart illustrating an embodiment of a method for providing a recommended item based on user event information according to the present invention.

The above-described objects, features, and advantages will be described later in detail with reference to the accompanying drawings, and accordingly, a person of ordinary skill in the art to which the present invention pertains will be able to easily implement the technical idea of the present invention. In describing the present invention, if it is determined that a detailed description of a known technology related to the present invention may unnecessarily obscure the subject matter of the present invention, a detailed description will be omitted. Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. In the drawings, the same reference numerals are used to indicate the same or similar elements.

1 is a network configuration diagram illustrating a system for providing a recommended item based on user event information according to an embodiment of the present invention.

Referring to FIG. 1 , the system for providing recommended items based on user event information includes an online market server 100 and user devices 200_1 to 200_N.

The online market server 100 is a server operated by a seller who sells items through a web page. Such a web page may include a main page, a product detail page, a shopping cart page, a user information page, and the like. Each of these web pages includes images, text, and navigation buttons (eg, add to cart, register items of interest, etc.).

Accordingly, when a behavior signal is generated on the web page, the online market server 100 may provide the user behavior signal generated on the web page to the user event information-based recommendation item providing server 100 .

For example, when the user selects the image of item A on the product detail page a for selling the item A, the online market server 100 sends an action signal indicating that the user clicks the image of the item A on the product detail page a. It may be provided to the information-based recommendation item providing server 100 .

For another example, when the user moves the item A to the shopping cart on the product detail page a for selling the item A, the online market server 100 transmits an action signal indicating that the item A has been moved to the shopping cart based on the user event information. It can be provided to the providing server 100 .

The user devices 200_1 to 200_N are terminals owned by a user who purchases an item through a web page provided by the online market server 100 after accessing the online market server 100 . In an embodiment, the user devices 200_1 to 200_N should be understood as a concept including all types of electronic devices capable of requesting a search and displaying advertisement information, such as a desktop, a smart phone, and a tablet PC.

To this end, an online market application for accessing the online market server 100 is installed in the user devices 200_1 to 200_N. Accordingly, when the online market application is selected and executed, the user devices 200_1 to 200_N may access the online market server 100 through the online market application.

The user devices 200_1 to 200_N may display a web page provided by the online market server 100 through the online market application.

In this case, the web page includes a screen loaded in the electronic device and/or content inside the screen to be immediately displayed on the screen according to the user's scrolling. For example, in the online market application of the user devices 200_1 to 200_N, the entire execution screen of the application extended in the horizontal or vertical direction in a state in which the web page is displayed and displayed according to the user's scroll will be included in the concept of the web page. Also, a screen in camera roll may be included in the concept of the web page.

The user devices 200_1 to 200_N generate action signals generated on the web page when a user input occurs on the web page displayed through the online market application.

In this case, the action signal is generated whenever a user manipulation (eg, touch, click, etc.) occurs on a web page displayed through the online market application. For example, behavior signals include clicking a specific image on a web page, selecting an option for an item registered on the web page and moving it to the shopping cart, and registering an item registered on the web page as a product of interest. can be created

When the action signal is generated, the user devices 200_1 to 200_N may determine another item matching the item registered in the web page as the recommended item, and then provide the recommended item through the online market application according to the type of the action signal.

2 is a block diagram illustrating an internal structure of a user device according to an embodiment of the present invention. An online market application is installed in the user device of FIG. 2 , and a web page may be displayed through the online market application.

Referring to FIG. 2 , the user device 200 includes a user database 110 , a user database manager 120 , an online market item database 130 , an online market item database manager 140 , a style database 150 , and a style database. It includes a management unit 160 , a recommended item search unit 170 , and a recommended item providing unit 180 .

The user database 110 may include user identification information, an action signal for estimating user preference, user preference estimated from the action signal, and user preference information input by the user.

The user database management unit 120 provides a query for the user's age, gender, occupation, fashion field of interest, previously owned items, etc., receives a user input for the query, generates user preference information, and uses the same in the user database 110 . ) can be reflected in

Furthermore, the user database management unit 120 according to an embodiment of the present invention is a time that a user browses an arbitrary style book provided through an application, item information that generates a like tag, a query item, through the application or other applications. By combining user behavior signals for estimating user preferences, such as purchased fashion item information and time information at which the information was created, taste information for a style that the user is interested in at that time is generated and stored in the user database 110 . can reflect

Furthermore, the user database manager 120 may generate body shape information of the user and reflect it in the user database 110 .

For example, when a body image obtained by photographing a user's body from a plurality of angles is generated through an image capturing module (not shown) and transmitted to the user event information-based recommendation item providing server 100 , the user database management unit 120 . can create a user body model from a machine learning framework that learns human body features from a large amount of body images. The user's body type model may include not only size information of each part of the user's body, but also information on the proportion and skin tone of each part of the user's body.

According to an additional embodiment of the present invention, the user database manager 120 may generate the user's preference information for a fashion item and reflect it in the user database 110 . The preference information may include information on the user's preferred price, preferred brand, and preferred use.

For example, when a fashion item is viewed or purchased through the online market application 310, the user database management unit 120 reflects the weight for reading or purchasing differently to provide information on a preferred price, a preferred brand, and a preferred use. It can be created and reflected in the user database 110 .

In particular, the user database manager 120 according to an embodiment of the present invention estimates the user's "taste" corresponding to a human feeling, generates the estimated taste information in a computer-recognizable form, and stores it in the user database 110 . It has a reflective feature.

For example, the user database manager 120 may extract a label for estimating the user's taste from the user's behavior signal. The label may be extracted as meanings of fashion items included in the user behavior signal, such as a style book viewed by the user, an item for which a like tag is generated, a query item, and a purchase item. Furthermore, the label may be generated as information on the look and feel and trend of fashion items included in the user action signal.

The label generated from the user behavior signal is weighted according to the user behavior, and the user database manager 120 may combine them to generate user preference information for estimating user preference and store it in the user database 110 . User preference information, user body type information, and user preference information included in the user database 110 may be used to set a recommended item or an exposure priority of the recommended item.

The online market item database 130 stores detailed item information such as country of origin, size, vendor, and wearing shot of items available for purchase in the online market.

The online market item database management unit 140 extracts a label describing the contents of the item based on the image of the item for an item that can be purchased in the online market, and indexes the label to create the online market item database 130 . create

First, the online market item database manager 140 may collect item information about items sold in the online market. For example, the online market item database management unit 140 may include a crawler, a parser, and an indexer to collect web documents of the online market and access text information such as images, item names, and prices included in the web documents. .

For example, a crawler collects a list of web addresses in an online market, and collects data related to item information by checking websites and tracking links. In this case, the parser interprets the web documents collected during the crawling process and extracts item information such as images, item prices, and item names included in the page, and the indexer can index the location and meaning.

On the other hand, the online market item database management unit 140 according to an embodiment of the present invention may collect and index item information from a website of any online market, but may also receive item information in a preset format from an affiliate market. .

Then, the online market item database manager 140 may process the image. This is to determine a recommended item based on whether an image is similar, without depending on text information such as an item name or a sales category.

According to an embodiment of the present invention, a recommended item may be determined based on whether images are similar, but the present invention is not limited thereto. That is, depending on the implementation, the image as well as the item name or sales category can be used as a single or auxiliary query. For this purpose, the user device 200 structures text information such as the item name and item category in addition to the image of the item to structure the online market item. The database 130 may be created.

In addition, the online market item database manager 140 according to an embodiment of the present invention may extract features of an image and index feature information of the images for efficiency of search.

More specifically, the online market item database manager 140 may detect a feature area of the images (Interest Point Detection). Here, the feature region refers to a main region from which a descriptor for a feature of an image, ie, a feature descriptor, is extracted for determining whether the images are identical or similar.

According to an embodiment of the present invention, such a feature region is more than a contour included in an image, a corner such as a corner among contours, a blob distinguished from a surrounding region, a region that is invariant or covariant according to the deformation of the image, or an ambient brightness. It can be a pole with dark or light features, and can target a patch (piece) of an image or the entire image.

Furthermore, the online market item database manager 140 may extract a feature descriptor from the feature area. The feature descriptor expresses the features of an image as vector values.

According to an embodiment of the present invention, such a feature descriptor may be calculated using the position of the feature region in the corresponding image, or brightness, color, sharpness, gradient, scale, or pattern information of the feature region. For example, the feature descriptor may convert a brightness value of a feature region, a change value of brightness, or a distribution value of the feature region into a vector and calculate it.

Meanwhile, according to an embodiment of the present invention, the feature descriptor for the image includes a global descriptor, a frequency descriptor, and a binary descriptor as well as a local descriptor based on the feature region as described above. Alternatively, it may be expressed as a neural network descriptor.

More specifically, the feature descriptor is a global descriptor ( global descriptor).

For example, the feature descriptor includes a frequency descriptor that converts and extracts the number of times that specific descriptors classified in advance are included in an image, the number of times that a global feature such as a conventionally defined color table is included, etc. In the layer of the binary descriptor and neural network, which extracts in bit units whether or not the elements are included or whether the size of each element constituting the descriptor is larger or smaller than a specific value, and converts it into an integer type. It may include a neural network descriptor that extracts image information used for learning or classification.

Furthermore, according to an embodiment of the present invention, the online market item database manager 140 may convert the feature information vector extracted from the image to a lower dimension. For example, feature information extracted through an artificial neural network corresponds to 40,000-dimensional high-dimensional vector information, and it is appropriate to convert it into a low-dimensional vector of an appropriate range in consideration of the resources required for the search.

For the transformation of the feature information vector, various dimensional reduction algorithms such as PCA and ZCA may be used, and the feature information converted into a low dimensional vector may be indexed into a corresponding image.

Furthermore, the online market item database manager 140 according to an embodiment of the present invention may extract a label for the meaning of the image by applying a machine learning technique based on the image. The label may be expressed as an abstract value, but may also be expressed in text form by interpreting the abstract value.

According to an embodiment of the present invention, the online market item database management unit 140 defines a label in advance, creates a neural network model that learns the characteristics of the image corresponding to the label, classifies the item in the image, and the item label can be extracted. In this case, the user event information-based recommendation item providing server 100 may assign a corresponding label to an image that matches a specific pattern with an arbitrary probability through a neural network model that has learned a pattern of an image corresponding to each label.

According to another embodiment of the present invention, the online market item database manager 140 forms an initial neural network model by learning the characteristics of the image corresponding to each label, and applies a large number of image items to the neural network model. It can also be expanded with precision. Furthermore, the user event information-based recommendation item providing server 100 may create a new group including the corresponding item if the corresponding item is not included in any group.

According to the above embodiments, the online market item database management unit 140 provides information on items such as women's bottoms, skirts, dresses, short sleeves, long sleeves, pattern shape, material, color, and abstract feeling (innocence, chic, vintage, etc.). A label that can be used as meta-information is defined in advance, a neural network model that learns the characteristics of an image corresponding to the label is generated, and a label for an advertisement target image can be extracted by applying the neural network model to an image of an advertiser. .

According to another embodiment of the present invention, the online market item database manager 140 may apply images to a neural network model formed in a hierarchical structure formed of a plurality of layers without separate learning of labels. Furthermore, a weight may be assigned to the feature information of the image according to the request of the corresponding layer, and the images may be clustered using the processed feature information.

In this case, additional analysis may be required to confirm whether the corresponding images are clustered according to any attribute of the feature value, that is, to connect the clustering result of the images with a concept that can be recognized by a human.

For example, the online market item database management unit 140 classifies the items into three groups through image processing, and provides A for the features of the first group, B for the features of the second group, and the features of the third group. When the label C for Daehan is extracted, it needs to be interpreted ex post that A, B, and C mean, for example, women's tops, blouses, and checkered patterns, respectively.

According to the above embodiment, the online market item database management unit 140 includes women's bottoms, skirts, dresses, short sleeves, long sleeves, pattern shapes, materials, colors, and abstract feelings (innocence, chic, vintage, etc.) in the clustered image group. Labels that can be interpreted ex post, etc. can be given, and labels assigned to image groups to which individual images belong can be extracted as labels of the corresponding image.

Meanwhile, the online market item database management unit 140 according to an embodiment of the present invention may express a label extracted from an image as text, and a text-type label may be used as tag information of an item.

Conventionally, the tag information of the item is subjectively directly provided by the seller, which is inaccurate and has low reliability. Item tags subjectively given by the seller act as noise, which lowers the efficiency of the search.

However, as in the embodiment of the present invention, when a label is extracted based on an image, the extracted label is converted into text and used as tag information of the corresponding item, the tag information of the item is used for human intervention based on the image of the corresponding item. It can be extracted mathematically without the need to increase the reliability of tag information and improve the accuracy of the search.

Furthermore, in the above embodiment, the online market item database manager 140 may generate category information of the corresponding item based on the image content. In this case, the label and the category information may be generated respectively, but the label may be used as the category information or the category information may be used as the label.

For example, when a label for an arbitrary image is extracted as a woman, a top, a blouse, linen, a stripe, a long sleeve, a blue, and an office look, the online market application 310 sets a label for a woman, a top, and a blouse to the corresponding item. Labels for linen, stripe, long sleeves, blue, and office look can be used as labels to explain the characteristics of items outside of the category.

Alternatively, the online market item database management unit 140 may index the corresponding item without distinguishing between the label and the category information. In this case, the category information and/or the label of the item may be used as a parameter for increasing the reliability of the image search.

Meanwhile, the online market item database manager 140 according to an embodiment of the present invention may filter the collected item description images. This is to construct an image database excluding images that may act as noise in image search.

More specifically, the online market item database manager 140 may determine whether to filter the image by comparing the label extracted from the image with the category and/or tag information directly provided by the seller.

According to an embodiment of the present invention, the online market item database management unit 140 has a plurality of images for a specific item, and if a label extracted from one image and a category given by the seller to the item are different from each other, the corresponding image Alternatively, specific items in the image can be filtered from the database.

For example, there are first to third images for item A, in the first image (female top, jacket), in the second image (female top, jacket) and (sunglasses, round), and in the third image, A case in which the label (sunglasses, round) is extracted can be considered. At this time, if the online market item database management unit 140 has the sales category of item A “sunglasses”, the user event information-based recommendation item providing server 100 excludes the first image and uses only the second and third images in the image database. can be configured.

This filtering is to reduce the noise of image search. In the above example, item A is actually about sunglasses. If the online market item database 130 is configured including all of the first to third item description images, even if the search target item extracted from the event information is a jacket, the first image and the By determining that they are similar, item A for sunglasses may be determined as an advertisement item. Therefore, the image that can reduce the accuracy of the search is filtered and the online market item database 130 is built.

The style database manager 160 creates a style database 150 for a style image in which a person is wearing a plurality of fashion items.

The style database manager 160 may collect style images online. For example, the style database management unit 160 collects a list of web addresses of fashion magazines, fashion brands, drama production companies, celebrity agencies, SNS, online markets, etc., and checks the websites to track the links, It is possible to collect image information contained in

On the other hand, the style database management unit 160 according to an embodiment of the present invention can collect and index images from websites such as fashion magazines, fashion brands, drama production companies, celebrity agencies, SNS, online markets, etc., but from affiliated companies. Image information may be separately provided along with index information.

Thereafter, the style database manager 160 may filter out images inappropriate for style recommendation among the collected images. For example, the style database manager 160 may filter the remaining images while leaving only the images including the human-shaped item and the plurality of fashion items among the collected images.

In this case, since the style image is used to determine other items that can be coordinated with the query item, it is appropriate to filter the image for a single fashion item. Furthermore, constructing a database with images of a person directly wearing a plurality of fashion items may be more useful than an image of the fashion item itself.

Accordingly, the style database management unit 160 according to an embodiment of the present invention may determine a style image included in the style database by filtering the remaining images while leaving only the image including the human-shaped item and the plurality of fashion items.

Thereafter, the style database manager 160 may process features of the fashion item item image included in the style image.

More specifically, the style database management unit 160 extracts the image features of the fashion item items included in the style image, and expresses the feature information as a vector value to generate a feature value of the fashion item item and to structure the feature information of the images. can

Furthermore, the style database manager 160 according to an embodiment of the present invention may extract a label from the style image and cluster the style images based on the label.

It is appropriate that the above label is extracted as the look and feel of the fashion item, such as the look and feel, and the trend. According to a preferred embodiment of the present invention, a label for a feeling that a person can feel in the appearance of a single fashion item included in a style image, a combination of a plurality of items, etc. can be extracted and used as a label. For example, the label may include a celebrity look, a magazine look, a summer look, a feminine look, a sexy look, an office look, a drama look, and a Chanel look.

According to an embodiment of the present invention, the style database management unit 160 defines a label in advance, creates a neural network model that learns the characteristics of the image corresponding to the label, classifies the item in the style image, and provides information about the item. You can extract the label. In this case, the user device 200 may assign a corresponding label to an image matching a specific pattern with an arbitrary probability through a neural network model that has learned the pattern of the image corresponding to each label.

According to another embodiment of the present invention, the style database manager 160 forms an initial neural network model by learning characteristics of images corresponding to each label, and applies a large number of style image items to the neural network model more precisely. It can also be expanded.

Meanwhile, according to another embodiment of the present invention, the style database manager 160 may apply the style images to the neural network model formed in a hierarchical structure formed of a plurality of layers without separate learning of the label. Furthermore, weight is given to the characteristic information of the style image according to the request of the corresponding layer, the images are clustered using the processed characteristic information, and celebrity look, magazine look, summer look, feminine look, sexy look, Labels that are interpreted ex-post as office look, drama look, Chanel look, etc. can be given.

In this case, the style database manager 160 may cluster style images using the label and generate a plurality of style books. This is intended to be provided as a reference to users. The user may browse a specific stylebook among a plurality of stylebooks provided by the user event information-based recommended item providing server 100 to find a favorite item, and may request item information search for the corresponding item.

Meanwhile, the style database management unit 160 may pre-classify items having a very high appearance ratio, such as a white shirt, jeans, and a black skirt.

For example, since jeans are a basic item in fashion, their appearance rate in style images is very high. Therefore, even if the user inquires about any item, the probability that jeans will be matched as a coordination item will be significantly higher than that of other items.

Therefore, the style database management unit 160 according to the embodiment of the present invention pre-classifies items having a very high appearance rate in the style image as buzz items, and sets the style book versions into those that include the buzz item and those that do not include the buzz item. can be created differently.

In this case, the buzz item may be classified by reflecting time information. For example, considering the fashion cycle of fashion items, items that are fashionable and disappear for a short time for a month or two, fashionable items that return every season, and items that are continuously popular for a certain period of time may be considered.

Accordingly, if time information is reflected in the classification of the buzz item and the appearance rate of a specific fashion item is very high for a certain period, the item may be classified as a buzz item together with information about the period. When the buzz items are classified in this way, in the subsequent item recommendation step, there is an effect that the recommendation target item can be recommended in consideration of whether the item to be recommended is in vogue or unrelated to the vogue.

When the recommended item search unit 170 receives a behavior signal for each user generated on a plurality of web pages from the online market server, it searches for a recommended item for an item registered in the web page where the behavior signal is generated.

To this end, when the recommended item search unit 170 receives an action signal for each user generated on a plurality of web pages from the online market server, extracts an image for an item registered in the web page in which the action signal is generated, and then the image Analyze to check whether it is coordinated with other items.

That is, the recommended item search unit 170 analyzes the image to extract the item, then extracts characteristic information of the item, calculates the completion rate of the item based on the characteristic information of the item, and determines that the completion rate is specific. Determines the number of items depending on whether they correspond to a ratio.

In an embodiment, the recommended item search unit 170 extracts a feature descriptor from the feature area of the item and then generates outline information of the item by using the feature descriptor. Then, the recommended item search unit 170 compares the outline information of the item with the predefined item outline information to calculate a completion rate of the item. For example, the predefined item outline information means outline information for each item, such as a long T-shirt, a short T-shirt, long pants, and short pants.

As described above, the reason why the contour information for each item is stored in advance and compared with the contour information of the item extracted from the image is that the image of the item registered in the online market is generally an image in which the model wears the item. However, depending on the photographer photographing the model wearing the item, the model's whole body may be photographed, but only the item worn by the model may be enlarged and photographed.

As described above, in the case of photographing the model's whole body, the image includes all of the items worn by the model, but in the case of photographing only a part of the items worn by the model, the image includes all of the items worn by the model and Some will be included.

For example, if a photographer is photographing a model wearing a white blouse and jeans, the image created by photographing the model's full body may include the white blouse and jeans worn by the model, and the white blouse worn by the model. An image created by photographing the bay may include all of the white blouse worn by the model and part of the jeans.

For this reason, when the image is generated by photographing the whole body of the model, the contour information of each item extracted from the image and the predefined item contour information will match. In addition, when the image is generated by photographing only a part of the model, the contour information of a specific item and the predefined item contour information among the items extracted from the image match, and the contour information of the specific item and the predefined item contour information do not match. won't

Accordingly, the present invention analyzes an image of an item registered in a web page to extract the item, then extracts characteristic information of the item, and calculates a completion rate of the item based on the characteristic information of the item.

Then, the recommended item search unit 170 may extract a label from each item by checking whether the completion rate of the item is equal to or greater than a specific rate.

In an embodiment, the recommended item search unit 170 may extract a label from each of the items when the completion rate of the items is greater than or equal to a specific rate.

In another embodiment, when the completion ratio of the item is less than or equal to a specific ratio, the recommended item search unit 170 extracts feature information of the item having a completion ratio of less than or equal to a specific ratio and uses it as a label. For example, the characteristic information of an item whose completion ratio is less than or equal to a specific ratio may be a floral pattern, a top, a bottom, red, etc., and such characteristic information is used as a label.

That is, since the completion ratio of the item is below a certain ratio, it is not possible to accurately determine which item the item is among the items sold in the online market, but depending on which part of the model the item is worn on the , bottoms, etc.) can be checked, and colors, patterns, patterns, etc. can be checked according to the characteristic information.

As described above, the number of labels may be changed according to the number of items having a completion ratio equal to or greater than a specific ratio. For example, the recommended item search unit 170 extracts one label if the number of items with a completion rate equal to or greater than a specific rate is one, and if the number of items with a completion rate equal to or greater than a specific rate is two, two labels are extracted. will be

Accordingly, the recommended item search unit 170 may determine whether an item is coordinated with another item according to the number of labels and the types of labels.

In an embodiment, when the number of labels is two or more and the types of labels are different, the recommended item search unit 170 may determine that the item is coordinated with another item.

In the case of the above embodiment, the recommended item search unit 170 may extract an item matching a label corresponding to another item from the market item database as a recommended item.

In another embodiment, when the number of labels is one, the recommended item search unit 170 may determine that the item is not coordinated with another item. In this case, after setting the item as a search target item, an item similar to the search target item is searched in the style database 150 based on the image similarity, and an item of a different category from the similar item in the style image in which the similar item is searched is determined as a coordination item.

Thereafter, the recommended item search unit 170 may search the online market item database 130 for the coordination item based on the image similarity, and extract an item similar to the coordination item as a recommended item.

More specifically, the recommended item search unit 170 according to an embodiment of the present invention may structure the characteristic information after extracting the characteristic information of the item. In this case, the recommended item search unit 170 may extract label and/or category information about the meaning of the item by applying the machine learning technique used to build the online market item database 130 . The label may be expressed as an abstract value, but may also be expressed in text form by interpreting the abstract value.

For example, the recommended item search unit 170 according to an embodiment of the present invention may extract labels for women, dresses, sleeveless, linen, white, and casual looks from the items. In this case, the recommended item search unit 170 uses the labels for women and dresses as category information of the items to be searched, and the labels for sleeveless, linen, white, and casual look are labels that describe the characteristics of the image of items other than the category. can be used as

Thereafter, the recommended item search unit 170 may search the style database 150 based on the similarity of the search target item. This is to search the style database 150 for an item similar to the search target item, and to check another item matching the similar item in the style image.

For example, the recommended item search unit 170 may calculate a similarity between the search target item and the feature values of the fashion item item image included in the style image, and check the item whose similarity is within a preset range.

Furthermore, the recommended item search unit 170 processes the feature value of the search target item by reflecting the weights required by the plurality of layers of the artificial neural network model for machine learning configured for the online market item database 130 , and the search target At least one fashion item group included in the style book having a distance value within a predetermined range from the item may be selected, and items belonging to the group may be determined as similar items.

Meanwhile, according to a preferred embodiment of the present invention, the recommended item search unit 170 searches the style database 150 based on the similarity of the search target item to determine the similar item, and in this case, in order to increase the accuracy of the image search, You can use the label and category information extracted from the image.

For example, the recommended item search unit 170 calculates a similarity between a search target item and a feature value of a style image stored in the style database 150, and labels and/or category information among items with a similarity greater than or equal to a preset range. Similar items may be determined by excluding items that do not match the label and/or category information of the search target item.

According to another embodiment, the recommended item search unit 170 may calculate the item similarity only in the style book having the label and/or category information matching the label and/or category information of the search target item.

In the above embodiment, the recommended item search unit 170 may extract a label from the search target item and specify a similar item based on the similarity of the search target item in a style book matching the label.

Of course, the recommended item search unit 170 may specify a similar item in the style database 150 based on the image similarity to the search target item without extracting a separate label from the search target item.

For example, when the search target item includes a leaf pattern dress, the recommended item search unit 170 may extract a label called tropical from the search target item. Thereafter, the recommended item search unit 170 may specify a similar item having a similarity within a preset range to the leaf pattern dress in the style book clustered with the label tropical.

Thereafter, the recommended item search unit 170 determines an item of another category matched with the similar item found in the style book as a recommended item, and selects an item similar to the recommended item in order to provide item information on the recommended item. The online market item database 130 is searched.

More specifically, the recommended item search unit 170 may search the online market item database 130 based on the image similarity for the recommended item determined in the style database 150 .

Then, the recommended item search unit 170 may determine the priority of exposure by reflecting user preference information. For example, when the user's taste information is focused on office look, the recommended item search unit 170 calculates a priority by weighting the office look label and provides recommended item information according to the calculated priority. have.

Unlike the above-described embodiment, the recommended item search unit 170 searches for an item corresponding to a user's taste stored in the user database 110 based on the online market item database 130 and the style database 150 and searches for the corresponding item. is determined as a user preference item.

More specifically, the recommended item search unit 170 searches the style database 150 to search for an item corresponding to a user's taste, and then searches for an item similar to the item found in the style database 150 in the online market item database 130 . You can search in , and determine the corresponding item as a user preference item.

When the recommended item providing unit 180 receives a user-specific action signal generated on a plurality of web pages from the online market server, the recommended item providing unit 180 provides a user preference item searched for by the recommended item search unit 170 according to the type of the action signal.

That is, when the web page indicates a product detail page and the action signal corresponds to a predetermined event signal, the recommended item providing unit 180 may provide the recommended item searched for by the recommended item search unit 170 on the web page. .

In one embodiment, the recommended item providing unit 180, when the web page indicates the product detail page and the action signal indicates that the scroll of the web page goes down to the bottom and then goes up again, in the recommended item search unit 170 The searched recommendation item may be provided on the web page.

In another embodiment, when the web page indicates the product detail page and the action signal instructs to put the item registered on the product detail page in the shopping cart and move to the shopping cart page, the recommended item providing unit 180 is a recommended item The recommended item searched for by the search unit 170 may be provided on the web page.

In another embodiment, the recommended item providing unit 180, when the web page indicates the product detail page and the action signal indicates to click the image of the item registered in the product detail page, the recommended item search unit 170 ), the searched recommendation item may be provided on the web page.

In another embodiment, when the web page indicates the main page, the recommended item providing unit 180 may provide the user preference item searched for by the recommended item search unit 170 on the web page.

3 is a flowchart illustrating an embodiment of a method for providing a recommended item based on user event information according to the present invention.

Referring to FIG. 3 , the online market application 310 extracts a label describing the contents of the item based on the image of the item for an item that can be purchased in the online market, and indexes the label to create the online market item database. generated (step S310).

The online market application 310 generates a style database by analyzing a style image in which a person is wearing a plurality of fashion items (step S320).

The online market application 310 receives each user-specific action signal generated on a plurality of web pages from the online market server (step S330).

The online market application 310 extracts an image for an item registered in the web page in which the action signal is generated, and then analyzes the image to determine whether it is coordinated with other items (step S340).

The online market application 310 searches for another item coordinated with the item in the online market item database according to the confirmation result and extracts it as a recommended item, or a recommendation matching the item based on the online market item database and the style database The item is extracted (step S350).

The online market application 310 provides the recommended item according to the type of the action signal (step S360).

As described above, although the present invention has been described by the limited embodiments and drawings, the present invention is not limited to the above embodiments, which is, if one of ordinary skill in the field to which the present invention belongs, various modifications and Transformation is possible. Therefore, the idea of the present invention should be grasped only by the claims set forth below, and all equivalent or equivalent modifications thereof will be said to belong to the scope of the idea of the present invention.

100: online market server
200: user device
110: user database
120: user database management unit
130: Online Market Item Database
140: online market item database management unit
150: style database
160: style database management unit
170: recommended item search unit
180: recommended item providing unit

Claims (10)

An online market item database for extracting a label describing the content of an item based on an image of the item for an item that can be purchased in the online market through the online market application, indexing the label, and storing it in the online market item database Management;
a style database manager that analyzes a style image in which a person wears a plurality of fashion items and stores it in a style database;
When a user-specific behavior signal generated on a plurality of web pages displayed through the online market application is received, an image of an item registered in the web page where the behavior signal is generated is extracted, and then the image is analyzed and coordinated with other items a recommended item search unit for extracting a recommended item based on the online market item database and the style database according to whether or not the item is selected; and
and a recommendation item providing unit providing the recommended item according to the type of the action signal.
user device.
The method of claim 1,
The recommended item search unit
After extracting the item by analyzing the image, the characteristic information of the item is extracted, the completion ratio of the item is calculated based on the characteristic information of the item, and when the completion ratio corresponds to a specific ratio, a label from each of the items After extracting , it is characterized in that it is determined whether the item is coordinated with another item according to the type and number of the label.
user device.
The method of claim 2,
The recommended item search unit
When the number of labels is two or more, if the types of labels are different, it is determined that the item is coordinated with another item, and an item matching the label corresponding to the other item is extracted from the market item database as a recommended item characterized by
user device.
The method of claim 3,
The recommended item search unit
If the number of the labels is one, it is determined that the item is not coordinated with other items, and an item similar to the item is searched in the style database based on the image similarity, and the similar item is matched with the similar item in the searched style image. An item of a different category is determined as a coordination item, the item database is searched for the coordination item based on image similarity, and an item similar to the coordination item is extracted as a recommended item.
user device.
5. The method according to any one of claims 2 to 4,
The recommended item providing unit
When the web page indicates a product detail page and the action signal corresponds to a predetermined event signal, the recommended item is provided on the web page
user device.
A method for providing a recommended item based on user event information executed on a user device, the method comprising:
extracting a label describing the contents of the item based on the image of the item for an item that can be purchased in the online market through the online market application, and indexing the label to generate an online market item database;
generating a style database by analyzing a style image in which a person is wearing a plurality of fashion items;
receiving an action signal for each user, each generated on a plurality of web pages displayed through an online market application;
extracting an image of an item registered in a web page in which the action signal is generated, and then analyzing the image to determine whether it is coordinated with another item;
According to the confirmation result, another item coordinated with the item is retrieved from the online market item database and extracted as a recommended item, or a recommended item matching the item is extracted based on the online market item database management unit and the style database management unit step; and
and providing the recommended item according to the type of the action signal.
A method of providing recommended items based on user event information.
The method of claim 6,
The step of analyzing the image to determine whether it is coordinated with other items is
extracting an item by analyzing the image, extracting characteristic information of the item, and calculating a completion rate of the item based on the characteristic information of the item; and
and extracting a label from each of the items when the completion ratio corresponds to a specific ratio, and determining whether the item is coordinated with other items according to the type and number of the labels.
A method of providing recommended items based on user event information.
The method of claim 7,
The step of extracting a recommended item matching the item is
When the number of labels is two or more, if the types of labels are different, it is determined that the item is coordinated with another item, and an item matching the label corresponding to the other item is extracted from the market item database as a recommended item comprising the steps
A method of providing recommended items based on user event information.
The method of claim 8,
The step of extracting a recommended item matching the item is
searching for an item similar to the item in the style database based on image similarity by determining that the item is not coordinated with another item if the number of the labels is one;
determining, as a coordination item, an item of a category different from the similar item in the style image in which the similar item is found;
searching the item database based on the image similarity for the coordination item; and
and extracting an item similar to the coordination item as a recommended item.
A method of providing recommended items based on user event information.
10. The method according to any one of claims 7 to 9,
The step of providing the recommended item according to the type of the action signal includes:
and providing the recommended item on the web page when the web page indicates a product detail page and the action signal corresponds to a predetermined event signal
A method of providing recommended items based on user event information.

Images