Voice Prescription using Natural Language Understanding

Misinterpretation of hand-written drug prescriptions may lead to improper medication given by the chemist, unavailability of proper records of drug prescription results in numerous issues. Research is hindered in the concerned domain due to missing data. Most of the issues can be countered if a systematic and standard prescription is available. In this paper, we intend to provide an interface for the practitioners to prescribe orally and prepare drug prescriptions in the form of a well-arranged table in compliance to the eprescription norms using prescribers dictation as input to remove the need for hand-written drug prescriptions and minimize any error related to misinterpretation of the drug prescription. This paper is focused on the task of extraction of meaningful information from the prescriber’s utterance through the slot filling approach of Natural Language Understanding. We use data augmentation to deal with class imbalance and inadequacy of data and experiment on both real data as well as the data with artificially generated prescriptions. We use the prescriber’s utterance as input, perform speech-to-text conversion and then apply state-of-the-art Natural Language Understanding techniques on it to extract meaningful information from the sentence to generate the drug prescription.