Anusha Balakrishnan

I’m currently a researcher at Semantic Machines, where I focus on making natural language generation systems more expressive, truthful, and fluent. The questions I’m most broadly interested in are:

  1. How much complexity and expressivity should input representations to generation systems capture?
  2. How can we better control the behavior of generation systems to make them more reliable?

Prior to joining Microsoft, I worked at Facebook on the Conversational AI team, where I worked on language generation, dialogue state tracking, and policy learning. Before that, I was a Master’s student at Stanford, where I worked with Percy Liang and He He on building dialogue agents for settings like collaborative games and negotiations.