Web portals offer many services and wealth of content to Web users. However, most users do not find interest in all the content present in these sites. Most of them visit some specific sites and browse in specific thematic areas of them. In this paper, a software technique is presented that allows the viewers of web sites to build their own personalized portals, using specific thematic areas of their preferred sites. This transcoding technique is based on an algorithm, which fragments a web page in discrete fragments using the page's internal structure. A training and update procedure is used for identifying the different instances of thematic areas in different time points.