Text summarization is a well-known and developed research topic, which aims to extract the informative sentences to form a small story instead of a full document. Traditional text summarization techniques focus on summarizing news articles. However, these techniques could not directly apply to video films. Video summarization is a new research topic that is able to handle different media types and using text summarization techniques to produce a video summary. This paper proposes a video summarization system that extracts important summaries from a given video via expanding lexical information from web. The experiments show that the web information can help us to generate human-readable summaries even the OCR recognition rate is limited.