Rapid Speaker Adaptation in Low Resource Text to Speech Systems using Synthetic Data and Transfer learning