Semantic Role Labeling

22.7 Selectional restrictions

1. How do selectional restrictions disambiguate “I want to eat someplace nearby”?
   The THEME of eat should be edible; someplace nearby isn’t, so it must be a locative adjunct of intransitive eat, not the direct object (the “speaker as Godzilla” reading).
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 22, §22.7, p. 510
2. Are selectional restrictions associated with words or senses?
   Senses: serve in “serves green-lipped mussels” restricts its THEME to food; in “Which airlines serve Denver?” to locations.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 22, §22.7, p. 510
3. How specific can selectional restrictions be?
   Very variable: imagine restricts its AGENT to animate beings but its THEME hardly at all; diagonalize requires a matrix; odorless requires something that could have an odor. The needed concept set is open-ended.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 22, §22.7, p. 510
4. How can selectional restrictions be represented in neo-Davidsonian logic?
   Add a type constraint on the role filler: \(\exists e, x, y\; Eating(e) \wedge Agent(e, x) \wedge Theme(e, y) \wedge EdibleThing(y)\). “ate a takeoff” becomes inconsistent.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 22, §22.7, p. 511
5. What are the problems with logical selectional restrictions, and the practical alternative?
   FOL is overkill and requires a huge commonsense knowledge base. Instead, specify a WordNet synset per argument: a filler is acceptable if it is a hyponym of that synset (eat’s THEME = {food, nutrient}; hamburger’s hypernym chain reaches food).

   hamburger, beefburger
    ⇒ sandwich
     ⇒ snack food
      ⇒ dish
       ⇒ nutriment, nourishment
        ⇒ food, nutrient
         ⇒ substance … ⇒ entity

   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 22, §22.7, p. 511
6. How do WordNet synsets handle the THEMEs of imagine, lift and diagonalize?
   Restrict them to {entity}, {physical entity} and {matrix}: this allows imagine/lift a hamburger but rules out diagonalize a hamburger.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 22, §22.7, pp. 511–512
7. Why are selectional restrictions better modeled as preferences than strict constraints?
   Violations occur in well-formed sentences, e.g. under negation (“you can’t eat gold for lunch”) or when restrictions are overstated (“Mr. Kulkarni ate glass on an empty stomach”).
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 22, §22.7, p. 512
8. What is Resnik’s selectional preference strength?
   How much information a predicate gives about the semantic class of its arguments: the KL divergence between the class distribution for that verb’s arguments and the general class distribution, \(S_R(v) = D(P(c \mid v)\,\|\,P(c)) = \sum_c P(c \mid v)\log\frac{P(c \mid v)}{P(c)}\). Eat has high strength, be low.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 22, §22.7, p. 512
9. What is Resnik’s selectional association?
   The relative contribution of a class to a verb’s preference strength: \(A_R(v, c) = \frac{1}{S_R(v)}P(c \mid v)\log\frac{P(c \mid v)}{P(c)}\), estimated from parsed corpora treating each word as a partial observation of all WordNet classes containing it.
   • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 22, §22.7, p. 513
10. What simpler measures of selectional preference exist?
    The conditional probability of an argument noun given the verb and relation, \(P(n \mid v, r)\), from co-occurrence counts in a huge parsed corpus; the inverse \(P(v \mid n, r)\); or log co-occurrence counts (better for subjects).
    • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 22, §22.7, pp. 513–514
11. How are selectional preference models evaluated with pseudowords?
    Replace each verb’s actual object (car) with a pseudoword combining it with a frequency-matched confounder (car/house); the model must choose which is the original object of drive. Alternatively, correlate with human plausibility ratings obtained by magnitude estimation.
    • Jurafsky & Martin, Speech and Language Processing, 3rd ed. draft, ch. 22, §22.7, p. 514