WhisBERT: Multimodal Text-Audio Language Modeling on 100M Words