Syntactic Knowledge:
Dependency Grammar and Parsing
Contact Person:
Vincenzo Lombardo
Researchers:
Cristina Bosco
Dependency Grammars describe the structure of a sentence in terms
of binary head-modifier (also called dependency) relations
on the words of the sentence. A dependency relation is an asymmetric relation
between a word called head (governor, parent), and a word called modifier
(dependent, daughter). A word in the sentence can play the role of the
head in several dependency relations, i.e. it can have several modifiers;
but each word can play the role of the modifier exactly once. One special
word does not play the role of the modifier in any relation, and it is
named the root. The set of the dependency relations that can be
defined on a sentence form a tree, called the
dependency tree.
In our approach, the knowledge about syntactic constraints is expressed
by means of Dependency Tables, which state which category can act
as modifier of another category. Dependency tables are a declarative representation
of the syntactic knowledge, but they are rather inefficient when used by
a parser directly. So, they are compiled into Parse Tables, which
are used by the parser to control the analysis. Parse tables fulfill the
role that in standard parsers for Phrase Structure Grammars is covered
by shift/reduce tables, i.e. for each pair [parser status, input category]
they specify which operation has to be carried out.
A recognizer and two parsers have been developed for Projective Dependency
Grammar. All of them use as basic knowledge source the parse tables. The
recognizer is an improved Earley-type recognizer. The first parser is an
all-path parser taking advantage of substructure sharing. The second parser
is a quasi-deterministic parser exploiting flexible lookahead and intelligent
backtracking (recovery) in case of failure.
A wide-coverage grammar for the projective fragment of Italian has been
written. Current effort goes toward a deeper lexicalization of the grammar,
by taking into account subcategorization constraints in order to build
a hierarchy of lexical subcategories.
|