Daniel Jurafsky

My lab studies speech and language understanding and production in both humans and machines. On the human side, we use computational and probabilistic models to understand how humans comprehend and produce language (How do we understand sentences? How do we produce words?). For example we have shown that humans are subconscious probabilistic reasoners in dealing with language, and that computation of these probabilities is sensitive to contextual factors (like who you're speaking to!). On the engineeering side, we leverage this knowledge about human processing to build better computational tools for tasks like understanding meaning or inducing language structures from raw text, automatic recognition and synthesis of speech, or automatic question answering. We also apply natural language processing to answer social science questions, such as exploring the history of science by extracting ideas from on-line papers.