Keyness and Relational Learning
Lubomír Popelínský, Jiří Materna · 2007
We propose a novel way to automatically extracting keyness concept of a document that exploits relational learning that allows to express not only keyness of a keyword but also the relation between keywords. It result in higher dicriminative power in information retrieval or text categorization. Informally, keyness is then a cloud of words and groups of words (clled keywords in the rest of this text), their attributes (morphological, syntactical, hyperon ymic) and relations between them that characterize a given document from the collection.