Adaptive parsing for time-constrained tasks
Roberto Basili, Maria Teresa Pazienza, Michele Vindigni, Fabio Massimo Zanzotto · Cineca Institutional Research Information System (Tor Vergata University) · 1999
Real Natural Language Processing (NLP) applications often involve cooperation among different processing modules, involving various degrees of linguistic skill. Advanced NL parsers are expected to recognize grammatical phenomena with a throughput suitable to satisfy ”time constraints” in real applications. We present a robust and efficient syntactic recognizer, Chaos (Chunk analysis oriented system), able to capture at least the grammatical information assumed to be crucial for several linguistic and non linguistic inferences as required by an application system. The parser inherits both the computational efficiency of a shallow parser and the accurate syntactic information typically produced by a lexicalized approach. The potentials of the technology are investigated through different corpora. The parsing architecture proposed is open to the integration of domain specific lexical information, thus realizing an explicit level of adaptativity.