Yeah, I guess you should just split your source data into words.