Figure 1 for Predicting Prosodic Prominence from Text with Pre-trained Contextualized Word Representations