I tried to prepare a model using g2p tool and existing hindi lexicon. What I am observing is that the trained model is not accepting some hindi characters while applying the model for text. For example : अकीर is the hindi word which already exists in the hindi lexicon used for training. When I tried to generate phonemes for अकीर using the training model the g2p tool is skipping  अ and giving phonemes for कीर. It is the same for any new word starting with अ. How can I rectify this problem?

--- (Edited on 11/12/2014 4:24 am [GMT-0600] by srinidhikrs) ---