Sign

Japanese version

The type of signs is defined in the file "grammar/types.lil"．Signs in Enju are designed based on [Pollar and Sag, 1994] and < href="http://www.delph-in.net/erg/">LinGO English Resource Grammar (ERG). But we make things simpler by omitting some complex constraints and picking a semantic representation different from Minimal Recursion Semantics (MRS)．

PHON: the list of words covered by the parsed phrase

SYNSEM

LOCAL

CAT

HEAD		constraint inherited from the head of the parsed phrase
		MOD: constraint on the modifiee of the parsed phrase(local list)
		POSTHEAD: allow post-modification or not(hpsg_binary)

VAL		subcategorization frame
		SUBJ: constraint on the subject that appears to the left of the parsed phrase(synsem list)
		COMPS: constraint on the constituent to the right of the parsed phrase(synsem list)
		SPR: constraint on the specifier to the left of the parsed phrase(synsem list)
		SPEC: constraint placed by a specifier(the parsed phrase) on the head(synsem list)
		?CONJ: constraint on a parallel structure conjoined to the parsed phrase(synsem list)

CONT: predicate-argument structure

CONX: not used

NONLOCAL

constraint on long distance dependencies

INHER		constraint on long distance dependencies inherited from daughter(s) of the parsed phrase
		QUE: not used
		REL: constraint on the relative pronoun used(local list)
		?SLASH: constraint on the moved constituent(local list)
		F_REL: constraint on free relatives

TO_BIND		constraint on long distance dpendencies cancelled by the parsed phrase
		QUE: not used
		REL: constraint on the relative pronoun used(local list)
		SLASH: constraint on the moved constituent(local list)
		F_REL: constraint on free relatives

PROB_FEATURE: features used by the statistical model

LEX_SPEC: name of the lexical head (lex_entry)

Types accepted by the HEAD feature are defined correponding to the following part of speeches．

type	part of speech	feature(s)
hpsg_verb	verb	VFORM: verb form(hpsg_vform), AUX: auxiliary verb (hpsg_aux), INV: inverted (hpsg_binary), TENSE: tense (hpsg_tense)
hpsg_noun	noun	CASE: case(hpsg_case)
hpsg_prep	preposition	PFORM: preposition form(hpsg_pform)
hpsg_adj	adjective
hpsg_adv	adverb
hpsg_pl	particle
hpsg_comma	comma
hpsg_punct	punctuation
hpsg_conj	conjunction
hpsg_subc	subordinated clause
hpsg_comp	complementizer
hpsg_det	determiner

The CONT feature is assigned a predicate-argument structure as the semantic representation of the parsed phrase．For further information, please look at predicate-argument structure．

The PROB_FEATURE feature is used by the disambiguation model. For further information, please look at statistical model．

Enju Developers' Manual Enju Home Page Tsujii Laboratory

MIYAO Yusuke (yusuke@is.s.u-tokyo.ac.jp)