Provides routines for "decompressing" further the expanded trees
formed by multiword token splitting.
Multiword token expansion leaves constituent words as siblings in a
"flat" tree structure. This often represents an incorrect parse of
the sentence. For example, the phrase "Ministerio de Finanzas" should
not be parsed as a flat structure like
(grup.nom (np00000 Ministerio) (sp000 de) (np00000 Finanzas))
but rather a "deep" structure like
(grup.nom (sp (prep (sp000 de))
(sn (grup.nom (np0000 Finanzas)))))
This class provides methods for detecting common linguistic patterns
that should be expanded in this way.