Comment by blauwbilgorgel | Hacker News Reader

Comment by blauwbilgorgel | Hacker News Reader

  Can an alligator run the hundred-metre hurdles?

http://www.youtube.com/watch?v=pjslsKZYXQ8 (Can alligators jump and climb? Yes, they can!)

http://www.animalquestions.org/reptiles/alligators/can-allig... (many have thought was a myth, it is true that alligators are in fact able to climb fences.)

These anaphora questions are better solved through natural language parsing, for example turning them into predicate logic. The article is right when it says that big data algo's have a problem with these sentences. But that is not to say they can't help. In the article:

  The town councillors refused to give the angry 
  demonstrators a permit because they feared violence. Who
  feared violence?

You can use big data to calculate the semantic closeness between "town council" and "fear violence". If that is closer than between "demonstrators" and "fear violence" you can make a good guess.

From the Wikipedia article on anaphora:

  We gave the bananas to the monkeys because they were hungry.
  We gave the bananas to the monkeys because they were ripe.
  We gave the bananas to the monkeys because they were here.

A quick Google search for "bananas were ripe", "monkeys were ripe", "monkeys were hungry" and "bananas were hungry" and counting the results will solve this.

Using semantic closeness can also work against you in the case of ambiguity:

  The robber sits on the bank.

Is "bank" a furniture here? A money bank? A river bank? Semantic closeness might picture the robber sitting on a money bank. A second or third pass is necessary: calculate the chance that a person will sit on a building vs. the chance that a person will sit on furniture.

This is a decades-old hard problem that is close to philosophy. Is the robot inside Searle's Chinese room intelligent? Is it really understanding what it is saying? Or is it "faking" it?

> You can calculate semantic closeness between "town council" and "fear violence". If that is closer than between "demonstrators" and "fear violence" you can make a good guess.

Only for that particular example. It's easy to come up with examples where that heuristic fails:

"The town council denied the protestors a permit because they were in a bad mood."

"The factory failed to produce the car on time because it was undergoing maintenance."

You're sort of validating the technique: in those examples the meaning of the sentence is absolutely ambiguous. The fact that a google search wouldn't return a conclusive result is expected.

Not for the second one; "car undergoing maintenance" will probably return more results than "factory undergoing maintenance", but one parsing of the sentence (the one where the factory is being maintained) makes far more sense than the other.

But at the same time, people use sentences like this very frequently and there seems to be some sort of heuristic we are using to disambiguate it.

It's not the case that every time an ambiguous sentence is used in real life people will stop and consider the possible cases. Granted, sometimes we will make the wrong assumption on how to disambiguate sentences that are absolutely ambiguous.

Another example where semantic closeness won't really help you (though context/situation would help):

"I looked at the man with the telescope."

Case 1: I looked at the man using the telescope.

Case 2: I looked at the man who was holding the telescope.

Actually I don't really get why anaphora are considered so hard:

the difference between We gave the bananas to the monkeys because they were hungry.

and

  We gave the bananas to the monkeys because they were ripe.

is simply to check in what you know about the world. Monkeys are animals, which can be hungry. Bananas are fruits which can be ripe. The third example is ambiguous because both animals and fruits can be 'here'.

Similarly, out of all the meanings of 'bank', there's only one that can be sat on.

Consider me naive, but it shouldn't be too hard to build a program backed by nothing more than an SQL database that can handle these.

And consider me SUPER naive, but that's exactly what I am doing at the moment in my spare time.

Excellent point made here.

The idea that pops into my mind is that whenever we have speech output from computers to humans, the computer should structure the sentences to minimize ambiguity (while remaining efficient). It would be unfortunate indeed if a human misunderstood a computer's instructions or warnings due to ambiguity on the part of the machine.

This is a very simplistic way of looking at the problem. To say that anaphora, ellipsis, parallelism, and the dozen or so other coherence relations are simply "solved through natural language parsing" is just incorrect (and probably meaningless).

I didn't say they were simply solved through natural language parsing. I said they were better solved (and that combining it with other approaches might help).

Reply on news.ycombinator.com