That's pretty cool. I wonder if the recent work on caption generation mixed with video understanding can get some sense of likelihood of disambiguated sentences

That's pretty cool. I wonder if the recent work on caption generation mixed with video understanding can get some sense of likelihood of disambiguated sentences to evaluate possible hypotheses.