For each document, a set of key-phrases and keywords is extracted, hereinafter referred to as key-phrases.
First, a part-of-speech (POS) tagger is used to tag the retrieved documents (snippet or full text). Then a
set of tagging rules is applied and a set of candidate noun key-phrases are extracted. Each key-phrase is
stemmed to remove duplicates. Finally, those candidate key-phrases above a specified frequency
threshold (dependent on the document length) are kept and stored in a document feature vector (DFV) of
the corresponding document.