Issues in developing pronunciation lexicon for Marathi

Language and speech resources are crucial for advancement of speech technology. Pronunciation lexicon is one among these. In this paper we describe our methodology of data collection for creating pronunciation lexicon for Marathi, an Indian language from an Indo-Aryan family. We mention in detail the issues which need to be dealt with as there is no one to one correspondence between the written forms and pronunciation. Selecting informants is equally important in creating pronunciation lexicon. We also list the issues which are directly related to informants´ speech as they need special attention while recording the data. The issues are resolved by using different methods of elicitation.