Label-Synchronous Speech-to-Text Alignment for ASR Using Forward and Backward Transformers | Read Paper on Bytez