There has been a great deal of interest in the development of ontology to facilitate knowledge sharing and database integration. In general, ontology is a set of terms or vocabularies of interest in a particular information domain, and shows the relationships among them (Doerr, Hunter, & Lagoze, 2003). It includes machine-interpretable definitions of basic concepts in the domain. Ontology is very popular in the fields of natural language processing (NLP) and Web user interface (Web ontology). To take this advantage into multimedia content analysis, several studies have proposed ontology-based schemes (Hollink & Worring, 2005; Spyropoulos, Paliouras, Karkaletsis, Kosmopoulos, Pratikakis, Perantonis, & Gatos, 2005). Modular structure of the ontology methodology is used in a generic analysis scheme to semantically interpret and annotate multimedia content. This methodology consists of domain ontology, core ontology, and multimedia ontology. Domain ontology captures concepts in a particular type of domain, while core ontology is the key building blocks necessary to enable the scalable assimilation of information from diverse sources. Multimedia ontology is used to model multimedia data, such as audio, image, and video. In the multimedia data analysis the meaningful patterns and hidden knowledge are discovered from the database. There are existing tools for managing and searching the discovered patterns and knowledge. However, almost all of the approaches use low-level feature values instead of high-level perceptions, which make a huge gap between machine interpretation and human understanding. For example, if we have to retrieve anomaly from video surveillance systems, low-level feature values cannot represent such semantic meanings. In order to address the problem, the main focus of research has been on the construction and utilization of ontology for specific data domain in various applications. In this chapter, we first survey the state-of-the-art in multimedia ontology, specifically video ontology, and then investigate the methods of automatic generation of video ontology.
In general, ontology is a set of terms or vocabularies of interest in a particular information domain, and shows the relationships among them. Sharing common understanding of the structure of information is one of the more common goals in developing ontology. Another reason is that it enables reuse of domain knowledge. The other reasons include making domain assumptions explicit, separating domain knowledge from the operational knowledge, and analyzing domain knowledge. For example, once we have ontology for video surveillance camera in one airport, it can be sharable with the other airport surveillance systems, and provide semantic knowledge on the airport system.
Key Terms in this Chapter
Click-Through Rate (CTR) or Cost-per-Click (CPC): CTR/CPC is a count of the number of times a user clicks on an ad in a Web site during a certain time frame (Bhat, Bevans, & Sengupta, 2002). Clicks indicate a behavioral response and click-throughs are an accountable measure of online advertising (Chatterjee, Hoffman, & Novak, 2003; Rosenkrans, 2006). The payment on an online ad is based on the number of times a visitor clicks on it.
Ad Impressions: Ad impressions measure the response of a delivery system to an ad request from a user’s browser (Bhat, Bevans, & Sengupta, 2002; Internet Advertising Bureau, 2004).
Rich Media: Rich media are a range of interactive media that show motion and utilize video, audio, and animation (Chabrow, 2006).
Search Advertising: Search advertising involves fees advertisers paying Internet companies to link or list their company Web site to a specific phrase or word (Internet Advertising Bureau, 2006).
Navigation: Navigation refers to the hypertext connection of the multimedia content and determines ease of information retrieval and a site’s organization (Karayanni & Baltas, 2003).
Stickiness: Site stickiness generally refers to the measurement of Web site attractiveness (Maity & Peters, 2005). It is the Web site’s ability to attract and hold users’ attention (Bhat, Bevans, & Sengupta, 2002).
Bots/Robots: Terms that have become interchangeable with agent to indicate that software can sift through the Internet for information and report back with it (Venditto, 2001; Rosenkrans, 2006).
Page Impression: Page impressions are an estimate of the number of pages served in a period of time and they are a good indicator of a Web page’s exposure (Bhat, Bevans, & Sengupta, 2002).
Conversion, Conversion Rate, Conversion Event: The percentage of users who follow through on an action, such as signing up for a newsletter, completing a credit card transaction, and downloading information (Carrabis, 2005; Dainow, 2004; Rosenkrans, 2005).
Cost-per-Thousand (CPM): The CPM model counts the number of users exposed to an online ad on a particular site (Rosenkrans, 2006).