Enhancing Audio Retrieval with Attention-based Encoder for Audio Feature Representation