T-Rex: Natural language processing researchers are all, "Oh boo hoo, computationally generating and understanding natural language is hard, we totally need more time to build Data from Star Trek, boo hoo hoo!"
T-Rex: Looks like you dropped the ball, NLP researchers!!
T-Rex: Because MY email autoresponder takes in text with all sorts of imperfect and irregular language, disambiguates word senses, builds a semantic understanding AND generates an appropriate response in the time it would take a real person to type it out. Oh snap! Did a rank amateur, wanting only to ensure that his body wouldn't be discovered for a few years after he died, just brutally advance the state of the art? SEEMS LIKE IT, BABY!
Utahraptor: So how's it work?
T-Rex: Just fine, thank you!
Utahraptor: No, I mean, how'd you do it? How do you figure out the semantics and word senses? Statistical methods? But those have problems with recall and precision, even with - are you using the web as a corpus? How are you handling the noise?
T-Rex: Utahraptor, please! So many questions!
T-Rex: I just wrote "Email Responder TWO THOUSAND" on a giant novelty chef's hat, taped a page from the dictionary to it, and then stuffed the whole deal into my computer's CD tray!
Off panel: W- What?
T-Rex: GRANT MONEY PLEASE
