Academic journal article Journal of Electronic Commerce Research

An Annotation Approach to Contextual Advertising for Online Ads

Academic journal article Journal of Electronic Commerce Research

An Annotation Approach to Contextual Advertising for Online Ads

Article excerpt

(ProQuest: ... denotes formulae omitted.)

1. Introduction

Web advertising has recently become one of the most commonly used marketing channels. Mu and Galletta reported that web advertising is one of the top three advertising mediums worldwide [2007]. Contextual advertising [Anagnostopoulous et al. 2007; Broder et al., 2007; Ciaramita et al., 2008] is one of the major approaches to textbased web advertising; contextual advertising is a form of targeted advertising in which the content of an advertisement is directly correlated to the content of the web page a user is viewing. An example includes a user visiting a website concerning traveling in Asia and then seeing an advertisement pop-up offering a special price on a flight to Taiwan. Contextual advertising is also called "In-Text" advertising or "In-Context" technology. The essence of this technique is to place an advertisement on a website on the basis of content similarity between advertisements and web pages. The success of contextual advertising depends on the ability of a system to determine which advertisements are most relevant to the content of a web page.

In the past, two main methods have been used for matching advertisements and web pages, and these are the vector space model and keyword-based model. In the vector space model, the advertisements and pages are represented as term vectors in a vector space, and the matching process is typically based on the similarity or correlation between an advertisement vector and a page vector. In the keyword-based model, the advertisements and pages are represented as a set of phrases or keywords, and the matching process is typically based on crossover between an advertisement keyword set and a page keyword set.

Multimedia advertisements have become increasingly popular because advertisers prefer vivid multimedia content with no text-such as a video, animation, image, or sound-rather than advertisements with long text, which do not draw the attention of users. Traditional approaches to contextual advertising have the following drawbacks when addressing multimedia advertisements.

(1) Although the vector space model is suitable for representing web pages, it is not ideal for advertisements; this phenomenon is due to multimedia advertisements having no text. The context of an advertisement cannot be determined without text.

(2) The keyword-based model cannot represent web pages efficiently because only a few keywords are selected to represent the page, rather than a full vector. The potential information contained in a web page is not fully used, thus resulting in a decreased performance.

A research question arising immediately is whether a new model that is suitable for contextual advertising with multimedia advertisements can be developed. Therefore, this study combined the strengths of both the vector space and keyword-based models to develop a new model for multimedia advertisements; in other words, we represented web pages and advertisements by using the vector space and keyword-based models, respectively. Because advertisers understand their advertisements clearly, this study assumed that advertisers assume responsibility for annotating advertisements. Annotating multimedia resources is a commonly used operation in numerous current ecommerce sites, such as Flicker, Twitter, and YouTube; therefore, annotating the advertisements should not be a difficult task for advertisers. The proposed approach comprises the following parts:

(1) Multimedia advertisements, such as images or videos, are distinguished by annotating keywords or phrases as tags.

(2) Web pages are represented as vectors.

(3) A mechanism is proposed to match the advertisement keywords with page vectors.

The proposed approach has the following advantages. (1) The model is simple and easy to use; advertisers can easily annotate their advertisements. (2) The model can determine the context of multimedia advertisements, even without text. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.