Head-corner Parsing and Robustness

Certain approaches towards robust parsing use the partial results of the parser. In order for such approaches to work it is often assumed that a bottom-up parser is essential: parsers that use top-down information (such as the head-corner parser) may fail to recognize relevant sub-parsers in the context of an ungrammaticality. We show how we use underspecification to obtain all relevant partial analyses for robustness.

Normally the head-corner parser will be called as e.g. :

?- parse(s(Sem),0,12).

(12)

indicating that we want to parse a sentence from position 0 to 12 with category s(Sem) (a sentence with a semantic representation that is yet to be discovered). Suppose however that a specific robustness module is interested in all `maximal projections' anywhere in the sentence. Such a maximal projection may be represented by a term xp(Sem). Furthermore there may be unary grammar rules rewriting such an xp into appropriate categories, e.g.:

xp(Sem) --> np(Sem,_). xp(Sem) --> s(Sem). xp(Sem) --> pp(Sem,_). xp(Sem) --> advp(Sem).

(13)

If we want to recognize all maximal projections at all positions in the input, then we can simply give the following parse goal:

?- parse(xp(Sem),_,_).

(14)

Now it may seem that such an underspecified goal will dramatically slow down the head-corner parser, but this turns out to be a false expectation. In actual fact we have experienced an increase of efficiency using underspecification. This can only be understood in the light of the use of memo-ization [8]. Even though we now have a much more general goal, the number of different goals that we need to solve is much smaller.

Note that the call to the parse predicate has variable extreme positions. This does not imply that all power of top-down prediction is lost; recursive calls to the parse predicate may still have instantiated left and/or right extreme positions. The same applies with even more force for top-down information on categories.

An example clarifies how the extreme position information can still be used even if the parse predicate is called with un-instantiated position arguments. For example, if at some point the parser is trying to find a vp and it predicts that its head is a verb between position p₂ and p₃ then, if a rule is selected which requires a pp to the left of that head, then it will be clear that for this pp the extreme right position will be p₂. Moreover, suppose that the head of this pp is found between position p₀ and p₁ then the parser can conclude that the np-complement of that preposition should be found between p₁ and p₂.