Text this: Global-Local Self-Attention-Based Long Short-Term Memory with Optimization Algorithm for Speaker Identification