Ambiguity exists when something can have more than one interpretation. Ambiguity is completely pervasive in language. There is almost nothing that you will say, hear, write, or read today that will not be ambiguous in some way, but humans are so good at what’s called “resolving” ambiguities that we only rarely notice them. Computers, on the other hand, are not–they have no way of not stumbling on them, and you could think of the entire challenge of people like me whose work involves computers and language as being finding ways to let computers resolve ambiguity.
Ambiguity is often manipulated in humor, and jokes, cartoons, and the like are the best way that I know of to get people to notice it. In that spirit, here’s an example of what’s called ambiguity of anaphoric reference. Anaphora is the phenomenon of something in language having its referent–you could think of that as the thing that gives it its meaning—from something else in language. Consider this cartoon:
Why is it funny? The joke is built on the fact that the word one in which one? gets its meaning from something earlier in the conversation. To a human, it’s obvious that the intended referent is an island in the Galapagos. However, the speaker interprets the referent as being one of his uncles. Why can he do that? Because the one is, in fact, ambiguous—there’s nothing in the linguistic structure of the sentences that indicates one way or the other whether the referent is the uncle, or the island. The humor works by violating our expectations. A computer program, on the other hand, doesn’t have any such expectations—unless we can figure out how to give them to it. Hence my job.
Want to know more about anaphora? Check out my colleague Ruslan Mitkov‘s excellent book Anaphora resolution. #cleaningthebasement
Zipf's Law 
Ok, so which features do we extract from this conversation to yield our natural expectations instead of the joke’s expectations? What’s the logic in the choice?
LikeLiked by 1 person
Logic? Probabilities? Both? Who knows…
LikeLike
Not everyone is equally good at using context to make a reasonable assumption, though.
LikeLiked by 1 person
Empirical question…
LikeLike
As far as the logic goes, I see three interpretations:
Q1: which island? A: San Cristobal
Q2: which uncle? A: Fred
Q3: which birthmark? A: the one under his arm on the left side
So I think this is pragmatics. I bet I could construct sentences
All of these are weird questions, because they require the guy asking the question to be asking for detail that he really doesn’t seem to need. Knowledge of the names and maybe shapes of Galapagos Islands is something that as a reader I don’t expect the guys to share knowledge. Also, Prone Guy saying “I had an uncle …” strongly suggests that this is an uncle who is not known to Top of the Hill guy, and Q2’s “Fred” cancels this suggestion. Q3 requires the guys to have detailed shared knowledge of (somehow) which uncle this is, as well as his birthmarks: Top of the Hill Guy has to somehow know the locations of the birthmarks, but not their shape. This scenario is so far out that no reasonable person would come upon it without having been twisted out of shape by linguists.
LikeLiked by 1 person
didn’t finish. I meant to say: if I change the objects and properties while keeping the structure, I think I can make any of Q1,Q2,Q3 become the top interpretation.
LikeLiked by 1 person