A Deep Learning Approach Toward Analyzing the Cross-Lingual Acoustic-Phonetic Similarities in Multilingual Speech Emotion Recognition