The document describes a system for query word labeling and back transliteration for Indian languages. It uses character n-gram features and context switch probability to label words in code-mixed languages. For back transliteration, it uses a hash-based mapping approach between languages. The system achieved high accuracy scores for language labeling and moderate scores for transliteration on Hindi, Gujarati, and Bangla test sets. Areas for future work include handling spelling variations and working on more under-resourced languages.