I wonder how much the upspeak way of typing affects this. People (even the author) often end declarations with question marks. Does this have any influence on the way the LLM parses the prompt?