Popular web search engines, such as Google, rely on traditional search result ranking methods such as the vector space model or probabilistic models in combination with the famous PageRank algorithm. In the last couple of years personal document relevance got considered when ranking search results, as it improves the result quality significantly. This thesis extends related work in the area of result personalization, by providing the concept and implementation of a keyword based personal search engine. Compared to other work, the personal relevance is measured by the user's activity with keywords of one document, such as clicking or hovering. A ranking algorithm is introduced, which considers keyword and document frequencies in the vector space model, combined with the interaction of those keywords to compute the document score. The concepts are implemented as an HTML5 browser extension for Google Chrome, which actively measures the user's interaction with the visited content, without interfering with the normal surfing behavior. Querying for visited content retrieves stored documents and orders them according to their personal relevance. An evaluation is conducted to test whether the behavioral ranking factors are significant enough for personal relevance. It is shown, that the interaction of the user with the document's content correlates with its relevance. Furthermore, a benchmark of WebSQL and IndexedDB as HTML5 data storage structures for insert, update and search operations reveals that the latter technology outperforms the former in almost every configuration.