Internet has become a very important tool in the present scenario. Website designing and development have become a challenging task with the increase of client needs and expectations. Especially multi-tasking and multi-linguality has become a major concern now. Existing information extraction system mainly concerns with extracting important keywords or key phrases that represent the content of the documents. The semantic aspects of such keywords have been explored extensively. Web pages now-a-days have evolved into media pages with language and text inter mixed with audio and images and even video. Many times the language of the page can be in different languages with words from one language used freely in the other without translation. For example 'computer' is used in any regional language without translation as it reaches the web surfer easily than a translated word! So the object of this study is to look into such words and web pages so that, the content can be extracted fast. Conventional data mining which normally deals with structured documents may not be suitable and instead the basic pixel-based approach could be more general and effective as it does not need definition of language. The paper presents typical pixel-based approach for some selected words in different formats and using statistical approach an interpretation of pattern matching with content word is developed. A typical ‘yahoo news’ page in different languages is an example of this. © 2015 Elsevier B.V., All rights reserved.