Supervised Adaptation of Sequence-to-Sequence Speech Recognition Systems using Batch-Weighting