Content Based Search in Web Archives

摘要

The widespread use of Internet as potentially useful and important data repository has led to the proliferation of Internet usage in search of valuable information to be used for important decision making. As the amount of the data stored and made available in the Internet grows, it becomes extremely necessary to create and maintain an Internet archive as a data repository for the purpose of supporting backup and record-keeping. Searching in the archival database is very complex process, however, in that there is simply too much information to be searched for. Furthermore, the ever-changing and dynamic natures of the web pages add more problems to the search process for desired information. In this paper, we propose an efficient approach to the problem of searching for the most relevant data source from the archives. There are two main issues that affect the search process in an archival database. One is the problem of understanding and interpreting the user's search intentions often represented as a form of a sequence of key words or natural language sentences. The other issue is the problem of mapping between the identified user intentions and the most relevant web pages satisfying the search objectives. In this paper, we present a sound solution to the latter problem of mapping pages with search intentions by using web page's content. To achieve the goal a web page indexing by vector features is developed. Active contour models are employed to extract geometric features. The main advantage and implication of this method is the indexing and geometric features extraction procedure, which will lead to the improvement of the accuracy of the search results as well as quicker retrieval of the most desired web pages.