Phrases including several words (that is, tokens) in English would be mapped onto a single word in Telugu. Thus 'vaccaaDu' ((he) came), 'vastaaDaa' (will (he) come?).
Search for "POS tags for morphologically rich languages". http://www.ontotext.com/sites/default/files/publications/eacl2012-pos.pdf is an example for Bulgarian (suffixing inflectional morphology).
In general, you will need complex POS tags specifying not only part-of-speech, but also all the inflectional categories (your POS tags will look like morpheme-by-morpheme glosses, see the examples for rule 4 here: http://www.eva.mpg.de/lingua/resources/glossing-rules.php, except that they just contain "V" for the verbal root, and not a translation of the root), and a morphological analyzer to parse the words.