Recommendation by composition style

Composition style is often an important factor in readers' selection of reading materials. For example, a reader may seek out articles written in similar style as his or her favorite writer. We present a new method for providing recommendations based on the composition style. Our algorithm analyzes and encodes the readability index and syntactical structure of a model document, and then searches for articles with similar readability index and structure. The text readability and syntactical structures are visualized to help readers compare the documents and make the selection. Our method adds a "search by style" component to the traditional keyword based search, and provides recommendation that fits the user's personal preferences better. We demonstrate our method by applying it to product review recommendation based on user preferred composition style.