Investigating representations of verb bias in neural language models