Automatic Haiku Generation Using VSM
Tsan Martin Wong, On Kee Sigma Lee, Hon Wai Chun · 2008
In this paper, we present our current research progress in using an AI approach and Web 2.0 techniques to automatically create modern poetry from text harvested from the blogosphere. Our haikus are generated using keyword seeds and the semantics of these keywords to extract lines from blogs to create a well-rounded poem. We believe this approach will yield much more human-understandable haiku poetry with much more realism than other traditional AI approaches as the language used in blogs is current as well as the events and cultural references. Our algorithm is based on formulating relationships between sentences as vectors using Vector Space Model (VSM) and then comparing relationships between sentences pairs using angle measurements based on simple cosine metric between the formed vectors.