Citations:case form


 * 2010, Alexander Gelbukh, Computational Linguistics and Intelligent Text Processing, page 202:
 * Note that when the input text is already cased but the case is unreliable, the input capitalization and the estimated defect rate could be used[. ...] To predict the case form of each token, extract the n-grams of length three or less containing it and gather search result snippets [... and] use the case statistics collected to decide which case form to associate with each token[. ... A]ssume that the instances of the target word in a snippet are representative of the case form(s) of the target word in the whole Web document.