The development of technology has a huge impact on people life. Specifically, it changes the way how things are stored and transmitted. Lots of information are put on the Internet or in local storage instead of printed or written on papers so that they can be easily looked up and processed. The amount of electronic data grows very fast and they exist in plenty of structures and contents, which lead to the need of efficient ways handling such a huge data volume. In this context, document modeling methods become crucial. Those methods give a common representation for different document structures and make it easy to extract information from the collection. This paper proposes an efficient document modeling framework that can be applied in real-world applications, such as search engines and web recommendation systems. The framework adopts vector space models and topic models with some modified factors to transform documents before modeling and improve the efficiency.