Enhancing Neural Spoken Language Recognition: An Exploration with Multilingual Datasets