Note that there are some explanatory texts on larger screens.

plurals
  1. POAI: Partial Unification in Open-World Reference Resolution
    primarykey
    data
    text
    <p>When performing reference resolution on predicates describing the semantics of dialogue expressions, I need to be able to allow for partial unification due to working in an open world.</p> <p>For example, consider the following scenario:</p> <p>There is a blue box in front of you. We refer to this blue box using the id <code>3</code>.</p> <p>A set of predicates <code>box(x)^blue(x)</code> can easily resolve to the blue box you know about. Making this query will return <code>3</code></p> <p>A set of predicates <code>ball(x)^yellow(x)</code> will not resolve to anything. This is fine.</p> <p>But now consider <code>ball(x)^yellow(x)^box(y)^blue(y)^behind(x,y)</code> <em>that is, the yellow ball behind the blue box.</em></p> <p>We don't know about a yellow ball, but we do know about a blue box! Of course it's possible that there's no ball behind the known box and that another box was being spoken of. But we're pretty sure we know what box is being talked about. </p> <p>I am working within a probabilistic framework in which I calculate the probability of each set of bindings satisfying the set of propositions; the reference resolution process then returns the most likely unifier/set of bindings. Unfortunately, when considering <code>behind(x,y)</code>, my system wipes out the probability of <code>3</code> being bound to <code>y</code> because it does not know of any yellow balls behind the box with id <code>3</code>. </p> <p>Is there a way to do <em>partial</em> unification of predicates, so that the system determines that the most likely resolution of the statement is <code>y/3 x/?</code> i.e. y is bound to 3 and the identity of x is unknown?</p>
    singulars
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    plurals
    1. This table or related slice is empty.
    1. This table or related slice is empty.
    1. CONote: I've considered adding a wildcard identifier `?` to represent an unknown place. In this situation, instead of considering just, for example, the bindings (1,1),(1,2),(2,1) and (2,2), the system would also consider (1,?), (2,?),(?,1),(?,2) and (?,?). Unfortunately I don't think this is a viable solution, as I can't picture how to calculate those probabilities; It is easy to calculate the probabilities of known entities having the properties in question, but I do not see how I would calculate the probability that an unknown entity exists with those properties.
      singulars
    2. COWhat exactly do you mean by resolution here? In automated theorem proving, [resolution](http://en.wikipedia.org/wiki/Resolution_(logic)) usually refers to a rule that takes two _disjunctive_ clauses and produces new _disjunctive_ clause. You have conjunctions, though. If you had a sentence like "every yellow ball is behind every blue box", the clause would be 1:{-yellow(x),-ball(x),-blue(y),-box(y),behind(x,y)}. Then, your observation of a blue box is two clauses: 2:{box(3)} and 3:{blue(3)}. Resolving 1 and 2 produces 4:{-yellow(x),-ball(x),-blue(3),behind(x,3)}. Then, resolving 3 and 4…
      singulars
    3. CO…produces 5:{-yellow(x),-ball(x),behind(x,3)}. If you later find out individual is a yellow ball, you'll have 6:{yellow(y)} and 7:{ball(7)}. After two more resolutions, you'll have {behind(3,7)}. Is this the kind of thing you're trying to do?
      singulars
 

Querying!

 
Guidance

SQuiL has stopped working due to an internal error.

If you are curious you may find further information in the browser console, which is accessible through the devtools (F12).

Reload