KR20080091738A - Apparatus and method for context aware advertising and computer readable medium processing the method - Google Patents

Abstract

A device and a method for providing context-based AD(ADvertisement), and a computer-readable recording medium for implementing the same are provided to select the proper ADs reflecting visual feature similarity between contents by storing feature extracted from an AD image, extracting the feature from an image included in the received contents, and using the extracted feature for searching the proper contents. An image input unit(10) receives, distributes, and stores the information to a database unit(40), which comprises advertisement database(41), advertisement image database and image feature database, when an advertiser or an AD agency provides AD information. A text manager(20) indexes AD stored in the database by analyzing text information included in the AD information. An image manager(30) extracts image feature from image information included in the AD information and stores the image feature to the database unit. A contents processor includes a contents input unit(50) receiving contents when the contents are selected, a contents analyzer(60) analyzing the text information of the received contents, a category searcher(70) setting an AD search target by searching the database unit with the text information, an image information searcher(80) detecting the AD among the AD search targets, and an AD provider(90) providing the detected AD and the contents.

Description

Context-based advertising device and method thereof, and a computer-readable recording medium capable of implementing the same {APPARATUS AND METHOD FOR CONTEXT AWARE ADVERTISING AND COMPUTER READABLE MEDIUM PROCESSING THE METHOD}

1 is a flow chart showing a general online context-based advertising method.

2 is a flow chart showing a conventional online context-based advertising method.

Figure 3 is a flow chart illustrating a method of registering and providing an advertisement of an embodiment of the present invention.

Figure 4 is a block diagram showing the configuration of an embodiment of the present invention.

5 is a block diagram showing a detailed configuration of an embodiment of the present invention.

6 is a flow chart showing an advertisement registration process of an embodiment of the present invention.

7 is a flow chart showing an advertisement providing process of an embodiment of the present invention.

** Description of symbols for the main parts of the drawing **

10: advertisement input unit 20: text information management unit

30: image management unit 40: database unit

41: advertisement data DB 42: advertisement image DB

43: image characteristics by category DB 50: content input unit

60: content analysis unit 61: keyword extraction unit

62: category determination unit 70: category search unit

80: image information search unit 81: image determination unit

82: image feature extractor 83: keyframe extractor

84: image feature search unit 90: advertising provider

91: advertisement selector 92: advertisement combiner

93: display unit

The present invention relates to a context-based advertisement, and more particularly, to extract information to be used for searching from text information and image or video information of an advertisement target content, and to use this to effectively search and provide an advertisement including image or video information. The present invention relates to an advertisement device, a method thereof, and a computer readable recording medium capable of implementing the same.

Recently, due to the explosive growth of various portal services (blogs, cafes, etc.) with the rapid spread of the Internet and search functions, it is common to create and register various contents and share them. The various portal services not only provide a search service for such content, but also provide a service for providing an advertisement to the content provided through the search as a revenue model.

In general, when content and advertisement are provided together, there may be a method of randomly inserting advertisements, but recently, a so-called context-based advertisement (CONTEXT AWARE ADVERTISING) scheme that selects and provides appropriate advertisements corresponding to the contents is provided. Introduced to increase the user's interest in advertising.

In general, the context-based advertising method that selects and displays advertisements based on the content of the searched or selected content can identify the contents of the content in various ways and select and provide advertisements with the same or relevant themes to induce user interest. Advertisement method, which mainly analyzes text-based content, identifies keywords, and then selects and provides highly matched advertisements. In addition, various techniques such as a method of effectively selecting keywords or limiting search targets through category analysis have been studied to increase the accuracy of advertisement selection.

In a general advertisement providing method including a context-based advertisement, when the advertisement provided for the content is irrelevant or low in relation to the content, it is difficult to expect an effect on the advertisement. For example, if an ad delivered with content is exposed with inappropriate content, it will increase the impression count of the ad, which in turn increases the click-through rate of the ad registered in the online advertising system. ) Is a major factor in lowering. In other words, the advertiser has a problem that the advertisement of the advertiser is displayed on the inappropriate content, and the effect of the advertisement is lowered, and the general user who checks the content may not be able to obtain useful advertisement information and may be uncomfortable by the inappropriate advertisement. . Therefore, the higher the rate of incorrectly matching the content and the advertisement, the lower the CTR, thereby reducing the unit price and sales.

Therefore, the technology that can accurately grasp the contents of the content, select the main themes, search the appropriate advertisements more accurately, and provide them together is the most important topic in the online advertising market.

1 and 2 is a flow chart briefly illustrating a general context advertising method, Figure 1 is a flow chart for a basic context-based advertising method, Figure 2 is a flow chart for a context-based advertising method using a method for classifying content in a clustering method. .

1 extracts the advertising keywords from the content to be advertised as shown. This method can be used to extract the noun part from the important words, and can also be extracted a variety of similar keywords.

Using the keywords extracted through the above process, the keywords of the received advertisements are searched to select matching advertisements. Keywords of the advertisement may be keywords specified by an advertiser when receiving the advertisement.

The top N advertisements having a high correlation with the advertisement target content are selected from the selected advertisements. At this time, the correlation is mainly used for the frequency of appearance of the keyword.

Thereafter, the selected top N advertisements are provided to the user who requested the content together with the advertisement target content.

Through this, advertisements related to the content targeted for advertisement may be selected and provided.

2 is to improve the efficiency of context-based advertising by using a method of classifying content by subject. As shown in FIG. 2, content is collected in advance to find a topic (topic) of high importance in the corresponding content. Classify the contents.

In addition, the ordered advertisement is also classified and stored by theme. That is, after ordering an advertisement, a classification of the advertisement is designated as advertisement-related information, which may be automatically processed.

When the process is completed, when a general user requests content classified by the topic, the content is extracted and a matching process is performed by searching for an advertisement having the same theme as the topic to which the corresponding content belongs. Thereafter, the matched advertisement is inserted into the content and provided.

In addition to the methods described above, a method of identifying a main subject of content and selecting an appropriate advertisement based on this method has been researched in various ways, but there are limitations in that the contextual advertisement is mainly dependent on text information. There is nothing else.

In other words, for contextual advertising, it is necessary to be able to grasp the contents of the content accurately and to be able to extract information for searching from it, but if the content is mostly information such as images or videos, general contextual advertising that extracts text-based search information In this way, it is difficult to specify the content. In particular, in the new Internet culture brought about by the rapid spread of User Created Contents (UCC), which include images or videos created by users, the contents are composed of only videos or images. It is difficult to select an advertisement suitable for the content by extracting search information by analyzing text information that is only about description or file name.

The purpose of the embodiments of the present invention is to newly extract the appropriate search information from the content including the image or video and to search for a corresponding advertisement. The feature information is extracted from the advertisement image and stored together with the advertisement information. By extracting feature information of an image included in the input content and using the feature information of the image as one of search information for selecting a suitable advertisement, it is possible to select appropriate advertisements reflecting the similarity to the visual characteristics of the content. It is an object of the present invention to provide a context-based advertising device, a method thereof, and a computer-readable recording medium embodying the same.

Another object of the embodiments of the present invention is to extract the key frame from the video to obtain the image information, and to extract the feature information from the image to utilize it as information representing the video, so that the advertisement and video or image containing the video The present invention provides a context-based advertising device, a method thereof, and a computer-readable recording medium capable of implementing the same, so that the similarity with the included content can be quantitatively identified so that multimedia advertisements related to the multimedia content can be appropriately selected.

Still another object of the embodiments of the present invention is to extract a keyframe of a video to obtain an image, and then to extract a feature of the image, or to extract a feature from the image to be regarded as information representing the video or image, the image or video It is to provide a context-based advertising device, a method thereof, and a computer-readable recording medium capable of implementing the same.

Another object of the embodiments of the present invention is to use an image feature obtained from an embedded image or an image feature of a keyframe of a video together with a keyword obtained from text in an advertisement or content in which text, an image, and a video are mixed. In addition, the present invention provides a context-based advertising apparatus, a method thereof, and a computer-readable recording medium capable of implementing the same.

The object of the present invention is not limited to the above-mentioned objects, and other objects that are not mentioned will be clearly understood by those skilled in the art from the following description.

In order to achieve the above object, the context-based advertising device according to an embodiment of the present invention classifies information related to the advertisement into text-based information and image-based information, and extracts the features of the image as one of the image-based information An ad processing unit; A database unit having a plurality of databases for storing the classified information through the advertisement processing unit; And extracting the image-based search information including the text-based search information and the image feature information from the input or searched content, and searching the corresponding advertisement in the database based on the search information and providing the content advertisement with the content. do.

When the information related to the advertisement includes video information, the advertisement processing unit extracts a key frame from the video information and processes it as an image, and the content processing unit includes video information in the input or searched content information. In this case, the keyframe is extracted from the video information and processed as an image.

The image feature may be at least one or more digital information among unique features of an image for distinguishing a similar image including a color histogram, a color distribution, a boundary histogram, a main color, and a shape of an included object.

In addition, the context-based advertising device according to another embodiment of the present invention includes an advertisement processing unit for obtaining and storing the characteristics of the video or image when registering the advertisement; An advertisement data DB for storing all registered advertisement related information; An image feature DB for separately storing the feature of the image provided by the advertisement processor in correspondence with the advertisement related information stored in the advertisement data DB; When the content is registered or selected, the content processing unit for acquiring the features of the video or image included in the content and searching the image feature DB based on the selected content and selecting an advertisement having an image similar to the video or image included in the content. It can be made, including.

The advertisement processor and the content processor further include means for extracting key frames from a video to obtain image information and to obtain feature information from the image information.

According to another aspect of the present invention, there is provided a context-based advertising method comprising: receiving information related to an advertisement, classifying the information into text-based information and image-based information, and storing the advertisement-related information in a database unit by using the information; Extracting feature information of an image from the image-based information and storing the extracted feature information in correspondence with the stored advertisement information; Classifying text-based information and image-based information included in the corresponding content during content registration or search; Extracting text-based search information from the text-based information of the content, extracting feature information of the image as search information from the image-based information, and then searching the database unit using the information; And providing the advertisement related information obtained as a result of the search together with the contents.

The present invention as described above will be described in detail with reference to the accompanying drawings and embodiments.

3 is a flow chart for conceptually describing an operation method of an embodiment of the present invention. When registering (or selecting) content including a video or an image, an appropriate advertisement is selected from among advertisements including a video or an image. The process shown is provided with.

First, when a user registers a content including non-text information such as an image or a video in a specific portal server, or selects it by specifying or searching, the user understands the information of the content and based on the text and the image or video. The information is divided into information based on the image, and the keyword string is extracted if the information based on the text exists. The method of extracting the keyword string may be applied to the keyword string extraction method used in various contextual advertisements, and most of the keyword string extraction methods may be applied together with the present embodiment.

In this way, it is possible to identify the subject of the contents through keyword extraction and to classify the contents. If the text-based information (document contents, file names of various configuration files, address names where the contents are registered, and registration positions of the contents), In case of insufficient category name, text included in linked document, etc., it is difficult to identify appropriate keywords. For example, when the content is composed only of an image or a video, the keyword must be extracted from limited text-based information such as a file name or registered category information, so that the accuracy is low. Alternatively, if a keyword obtained by analyzing text-based information is popular or general content, searching for a corresponding advertisement based on this may result in too many search results.

Therefore, as shown in the figure, the binary information of the image is extracted by detecting the existence of the image or video, and the feature of the image is extracted, so that even the image or video information that could not be searched can be utilized as information for searching. In particular, in the case of a video, when a method of extracting a key frame is applied, it may be treated like an image.

The image feature extraction method basically means a method of acquiring the core qualities of the image and collecting them into digital information, and basically a color histogram for obtaining statistical color information and a color distribution on the image ( Color Distribution for measuring color layout, Edge Histogram for distinguishing texture by measuring boundaries, Dominant Color, Color Structure, Type of object included At least one or more of feature information for distinguishing similar images including (Object Shapes) may be used.

In addition to the above method, various methods for distinguishing similar images may be applied. The standardization of such images is being applied to a limited object such as a video format or digital broadcasting, and is used for pattern recognition, biometric recognition, or various visions. It is also being studied in various fields such as systems (factory automation, robotics, reconnaissance, surveillance, etc.). That is, the image feature of the present embodiment is made of standardized digital information, and is to quantitatively grasp the similarity of the images, so any method that can standardize the visual feature of the image is applied to the image feature extraction method of the present embodiment. Can be. Therefore, the image features that are to be searched are also extracted from the same or similar conditions and environments, so that accurate search is possible.

When the image or the video is included in the content as described above, the feature information is extracted and then keyword search is performed according to the existence of the keyword. The keyword search can encompass all the keyword search methods used in general contextual advertising, and can be referred to as a primary level search using text-based search information. In some cases, the search level can be relaxed so that the search results are not too limited.

Then, the image similarity is searched using the image feature extraction value with respect to the results obtained through the keyword search to select a result having a similar image and provide the result.

The above method sequentially searches using text-based search information, and searches again using image-based search information (image characteristics) on the search results, and finally, image information (image, video, etc.) having visual similarity. ) To provide highly relevant search results.

The above process may be modified such that text-based search information and image-based search information are searched in different orders or simultaneously used. Also, the text-based search information and the image-based search information may be composed of a plurality of pieces of information, for example, text-based search information consisting of keyword information (plural numbers), category information, and the like, and individual images for the plurality of images. Image-based retrieval information consisting of feature information (applicable to multiple types of feature extraction techniques) may be applied to each other to apply different weights to change the selection mode for a desired result. These different weights may be automatically determined according to the search or registration situation, may be specified by the user before the search or registration, and may be fixed.

4 is a block diagram showing the configuration of an embodiment of the present invention. As shown in FIG. 4, the left side is an advertisement processing unit for appropriately storing advertisement related information in the database unit 40, and the right side thereof. Is a content processing unit for selecting advertisements corresponding to the content by appropriately searching for advertisement related information stored in the database unit 40.

That is, the flowchart of FIG. 3 illustrates an example of an operation process that can be performed through the content processing unit. When the above-described operation method of FIG. 3 is applied to the context advertisement, the configuration as shown in FIG. 4 may be considered.

First, the database unit 40 for storing the information related to the advertisement and the information for the search may be configured as a single database, but it is preferable that the database unit 40 is configured separately according to the type for the convenience of operation. The illustrated configuration includes an advertisement data DB 41 for storing various information related to an advertisement, an advertisement image DB 42 for storing key images extracted from various images included in the advertisement or a video included in the advertisement, and the advertisement image. This is a case where a feature of the various images stored in the DB 42 is divided into category-specific image features DB 43 which are classified and stored for each category. In this case, the advertisement image DB 42 may be omitted. However, when the advertisement image DB 42 provides the advertisement images including the video as a still image or only the images are provided in the slide show method, the corresponding information is required. A case in which the advertisement image DB 42 is further configured will be described as an example.

Looking at the configuration of the advertisement processing unit for storing the various information related to the advertisement in the database unit 40, when the advertiser or advertisement-related businesses provide a variety of information related to the advertisement to receive the distribution to the database unit 40 An advertisement input unit 10 to store the text information, a text information management unit 20 to analyze the text information included in the advertisement related information, and add information for indexing the advertisement stored in the database unit 40, and to the advertisement related information. The image management unit 30 extracts an image feature from the included image information and stores it in the database unit 40. Here, all original information related to the advertisement provided by the advertisement input unit 10 is stored in an advertisement data DB 41. The text information management unit 20 identifies keyword and category information set by an advertiser when the advertisement is input. Alternatively, after extracting keyword and category information by analyzing text information of the advertisement, the advertisement-related information stored in the advertisement data DB 41 is indexed in correspondence with the corresponding keyword and category. This method first analyzes the keyword and category information, and then records corresponding advertisement related information in a predetermined storage area according to the keyword and category obtained according to the result. A method of storing keyword and category information about as a index information in a separate area may be applied.

The image manager 30 stores an image included in advertisement related information in the advertisement image DB 42 and classifies corresponding image feature information into a corresponding category by referring to indexing information of the text information manager 20. Save it. That is, the advertisement data DB 41, the advertisement image DB 42, and the image feature DB 43 for each category are indexed to each other so as to refer to corresponding information, and are classified according to the same category or keyword in each DB. Stored. Therefore, even if only any of these DBs are searched, relevant information (all advertisement-related information, advertisement images, image characteristics) can be obtained through the results.

In addition, the image manager 30 extracts key frames of the video when the video is included in the advertisement-related information, generates it as an image file, stores the image in the advertisement image DB 42, and stores the corresponding images. It is also configured to include a function to extract the image feature and store in the image feature DB (43) for each category. Since it is only necessary to apply a known key frame extraction method and registration, the burden on the advertisement processing system is slight.

Now, a configuration of a content processor for utilizing the stored advertisements as a contextual advertisement method will be described. In order to search for an advertisement corresponding to the content input from the database unit 40 that stores a plurality of advertisement specific information such as keywords, categories, and image features in the above-described manner, appropriate search information must be extracted from the corresponding content.

Therefore, the content processing unit registers the content on a specific site or performs a search to select the specific content, the content input unit 50 to obtain the corresponding content, and text-based information of the content received through the content input unit 50. A content search unit 60 for analyzing the category, a category search unit 70 for searching the database unit 40 through text-based information provided by the content analysis unit 60, and setting a search target of an advertisement; Image information retrieval unit for acquiring image characteristics when there is image information in the content, and the database unit 40 detects a corresponding advertisement among the search targets of the advertisement set through the category retrieval unit 70 based on the image feature. And an advertisement providing unit 90 for providing the advertisement detected by the image information retrieval unit 80 together with the contents.

In order to apply a method of sequentially reducing search objects, the configuration may be configured by first applying a text-based category search (or keyword search) that is currently being used to define a primary search target, and the secondary search for the limited area may be performed. This is done using the feature, which shows how to reduce the search time.

However, this simple approach can be modified in various ways and vice versa (keyword / category search after image feature search), or concurrent (by performing image search on an object detected by keyword / category search to calculate match), and different weights. Note that an application method, a method of varying weights according to the amount of image information and text information, and the like can be changed to more search configurations through division or combination thereof. All of these changes are also to be considered as belonging to the present invention because they are encompassed by the technical configuration of the contextual advertisement using non-textual information to search for appropriate advertisements by integrating text-based search information and image-based search information.

FIG. 5 is a block diagram illustrating in detail the configuration of the content processing unit in the configuration of FIG. 4. Note that the configuration is one of many embodiments.

The content analyzer 60 may include a keyword extractor 61 and a category determiner 62 as shown in the drawing, and the content provided to the content analyzer 60 through the content input unit 50. There is not only information on the content itself but also environment or setting information on the content. For example, additional information such as location information of the content, category information in which the content is registered, and topic information such as a blog / cafe / bulletin in which the content is registered are also provided. Information can be used to obtain keywords or basic category information. However, in order to avoid the case in which such information is unclear, the keyword extractor 61 may be additionally or exclusively configured, and the keyword extractor 61 may generally apply a keyword and category extraction method (repeated string extraction) that is generally applied. It extracts the appropriate keywords or categories from the contents by utilizing weights, linking, or extended text information analysis such as previous search contents, etc. on the string registered in the main dictionary. The category determiner 62 generally performs a function of matching various categories set differently for each website or webpage with categories applied to classify advertisement-related information in the database 40.

The keyword and category information provided through the content analyzer 60 may be text-based search information, and appropriate advertisements may be selected by searching the advertisement data DB 41 using the same. If only a text-based search is required without image related information, the search results may be limited by performing keyword search once again on the results obtained after performing category search as shown first. Alternatively, if only a few results are desired, the category search unit 70 may reduce a target for comparing image features by performing category search and keyword search even if there is the image related information.

After the category search 70 primarily sets a search target and sets the content, the content is provided to the image information search unit 80 to search for image features. The image information retrieval unit 80 may include an image determination unit 81 for dividing the types of images included in the content into an image and a video, and a key frame when the included image determined by the image determination unit 81 is a video. A key frame extractor 83 for extracting an image and extracting an image feature from a corresponding image when the included image identified by the image determiner 81 is an image, or an image feature from a key frame obtained from the key frame extractor An image feature extracting unit 82 for extracting an image, and an image feature searching unit 84 for selecting an advertisement having a degree of match or higher than a preset level by searching for features of an advertisement-related image previously stored in the database based on the extracted image feature. It can be made of).

The image feature search unit 84 searches the image feature DB 43 for each category of the database unit 40, and selectively searches for image features of the advertisement selected through the category search unit 70, or the category search. Using the category information determined through the unit 70, only image features of a desired category may be searched among image features classified by category.

Although the category retrieval unit 70 and the image feature retrieval unit 84 have a configuration in which they are connected in series, they may be configured in a parallel manner, respectively, and finally, after obtaining all the retrieval information, the database unit may be a combination thereof. 40 may be configured to search.

The advertisement providing unit 90 may include an advertisement selecting unit 91 for selecting desired advertisements among advertisement information finally obtained, an advertisement combining unit 92 for combining the advertisement with content, and the combined advertisement to the user. The display unit 93 may be provided. The advertisement selecting unit 91 selects the advertisement stored in the advertisement data DB 41 or the advertisement stored in the advertisement image DB 42 according to the type and the situation of the advertisement to be currently provided using the information for selecting the advertisement. Can be selected. For example, if the amount of content to be provided is high or important, or if the exposure is frequent, all the advertisement related information stored in the advertisement data DB 41 is used, and if the amount of content to be provided is small or low in importance, Alternatively, if the exposure is rare, only the image stored in the advertisement image DB 42 may be provided.

FIG. 6 is a flowchart illustrating a process of inputting advertisement information using the above-described configurations. As illustrated in FIG. 6, an advertiser or an advertisement related person stores advertisement information in a database unit. At this time, the advertisement is composed of text information and image (including video) information, each of the strategic keywords or categories are set according to the target advertising targets. Of course, a prohibition situation or time zone may be additionally set, but a description thereof will be omitted.

As described above, advertisement-related information and setting information (category information and keyword information) are received and stored according to the setting.

If video information is included in the advertisement information, keyframes are extracted from the video information to generate image information, and stored as an advertisement image. If the image information is included, the corresponding image is stored as an advertisement image. In the illustrated case, only one of the video information and the image information is included. If the video information and the image information are included together, keyframes can be extracted from the video and stored as an advertisement image. have.

The image feature is extracted from the stored advertisement image and the advertisement image feature is stored. The feature information of the stored advertisement image corresponds to the previously stored advertisement information so that the corresponding advertisement image and advertisement information can be obtained through the feature information of the advertisement image. In other words, proper indexing should be made to maintain the association between detailed information about the same advertisement.

FIG. 7 illustrates a process of selecting an advertisement suitable for content by searching for advertisement information, advertisement image, and advertisement image characteristic information stored in the same manner as in FIG. 6, and includes video or image information and text information as shown. When the content is newly created and registered in a specific place, or when the corresponding content is searched or selected, the appropriate search information is obtained by checking the information of the content.

If the content information is included in the content information, the search information based on the text is extracted. This may be a keyword or category information of the corresponding content.

The image feature is extracted when the image information is included in the content, and the image feature is extracted from the key frames by extracting the key frame when the video is included in the content.

Searching the database using the image feature obtained from the content and the text-based information may select an advertisement having high similarity and visual similarity between the text-based subject of the content.

In this case, if different weights are applied to the text-based search information and the image-based search information (image characteristics), more appropriate search results can be obtained. That is, by applying different weights to the text-based search information and the image-based search information, it may be determined whether to focus on visual consistency or text-based information. For example, the weight is automatically applied according to the amount of text information and the size of the image information included in the content, or the type of the content (content registered in the image board, content registered in the space where the picture is mainly registered, video It is possible to set the weight, such as automatically variable according to the contents registered in the space to register mainly).

The above-described registration method of advertisement information and a method of selecting advertisements having high theme correspondence by acquiring search information from contents are not merely using text information but also considering similarities to images or videos that provide visual information. This method allows you to better understand the characteristics of content that is mainly composed of visual content such as images and videos, and to increase the advertising hit for the content. By providing a high advertisement, it is possible to provide a contextual advertisement that both advertisers and users can be satisfied with.

As described above, the context-based advertising apparatus and method thereof according to an embodiment of the present invention, and a computer-readable recording medium capable of implementing the same, extract feature information from an advertisement image, store the feature information, and include the content in the input content. By extracting feature information of the image, and using the feature information of the image as one of the search information for selecting a suitable advertisement, it is possible to select appropriate advertisements reflecting the similarity to the visual characteristics of the content.

In accordance with an embodiment of the present invention, a context-based advertising device, a method thereof, and a computer-readable recording medium embodying the same may extract keyframes from a moving image to obtain image information, and extract feature information from the corresponding image to obtain the corresponding moving image. By using the information as representative, it is possible to quantitatively grasp the similarity between the advertisement containing the video and the content containing the video or image, there is an effect that can appropriately select the multimedia advertisement related to the multimedia content.

In accordance with an embodiment of the present invention, a context-based advertising apparatus, a method thereof, and a computer-readable recording medium embodying the same may extract a keyframe of a video to obtain an image, and then extract a feature of the image or extract a feature from the image. By extracting and considering this as information representing the video or image, the image or video itself can be a target of search or relevance judgment, and thus the scope and area of the search can be dramatically expanded.

The context-based advertising apparatus and method thereof, and a computer-readable recording medium embodying the same according to an embodiment of the present invention include an image feature or a key frame of a video in an advertisement or content in which text, an image, and a video are mixed. By using the image features obtained from the extracted images together with the keywords obtained from the text, it is effective to select and provide more relevant advertisements by appropriately adjusting the weight of each, thereby greatly improving the accuracy of the contextual advertisement. It is effective to let.

Claims (1)

An advertisement processor classifying information related to an advertisement into text-based information and image-based information, and extracting a feature of an image as one of the image-based information; And a database unit having a plurality of databases for storing the classified information through the advertisement processing unit.

Images