Panel: The limits of automation: optimists vs skeptics

For building resources and performing MT: Do automated techniques deliver? This (possibly controversial) panel will examine where the recently developed techniques of automated (statistical and other) methods for MT and Computational Linguistics are leading. In addition to performing some subtasks of MT, these techniques have proven rather useful for building resources. But, ultimately, the question remains: how much do these new techniques help? Do system builders still have to go in there by hand, to enrich the lexicons and fix the rules? Is it a matter of now being able to come up to speed more quickly with a new system, or is there a true improvement in quality as well? In other words: do automated techniques deliver?