Part-of-Speech Tagging for Code-Mixed English-Hindi Twitter and Facebook Chat Messages

The paper reports work on collecting and
\nannotating code-mixed English-Hindi so-
\ncial media text (Twitter and Facebook
\nmessages), and experiments on automatic
\ntagging of these corpora, using both a
\ncoarse-grained and a fine-grained part-of-
\nspeech tag set. We compare the perfor-
\nmance of a combination of language spe-
\ncific taggers to that of applying four ma-
\nchine learning algorithms to the task (Con-
\nditional Random Fields, Sequential Mini-
\nmal Optimization, Naïve Bayes and Ran-
\ndom Forests), using a range of different
\nfeatures based on word context and word-
\ninternal information