Semisupervised Grammar Induction based on Text Chunking Information
Lourdes Araujo · 2012
naden distintos tipos de constituyentes. Palabras clave: InducciGramatical, Msemisupervisados, Ansuperficial Abstract: Syntactic analysis of texts requires the availability of the grammar underlying the language. The grammar can be obtained from corpora syntactically annotated, but such corpora do not exist for many languages. This has led to a growing interest in unsupervi- sed grammar induction, which does not require such annotations, but provides less accurate results. Aiming at improving the accuracy of this kind of approach, we have resorted to additional information, which can be obtained more easily. Shallow parsing or chunking identifies the sentence constituents, but without specifying their internal structure. In this work we have investigated how the results of a pattern-based unsupervised grammar induc- tion system improve as data on new kind of phrase are added.