The Txala dependency parser [ACM05] gets parsed sentences -that is, sentence objects which have been enriched with a parse_tree by the chart_parser (or by any other means).
class dep_txala : public dependency_parser { public: /// constructor dep_txala(const std::string &, const std::string &); /// Enrich all sentences in given list with a depenceny tree. void analyze(std::list<sentence> &); };
The constructor receives two strings: the name of the file containging the dependency rules to be used, and the start symbol of the grammar used by the chart_parser to parse the sentence.
The dependency parser works in three stages:
<GRPAR>
rules are used to complete the shallow parsing
produced by the chart into a complete parsing tree. The rules are
applied to a pair of adjacent chunks. At each step, the selected
pair is fused in a single chunk. The process stops when only one chunk remains
<GRLAB>
rules
The syntax and semantics of <GRPAR>
and <GRLAB>
rules are described in
section 3.18.1.