-
Notifications
You must be signed in to change notification settings - Fork 5
/
ambiguous-rules.tdl
91 lines (78 loc) · 3.89 KB
/
ambiguous-rules.tdl
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
;;; -*- Mode: TDL; Package: LKB; Coding: utf-8 -*-
;;;
;;; Based on:
;;; JACY Grammar:
;;; Developed by Melanie Siegel, Francis Bond
; (MS 2004-04-02) Empty case particles. Try to restrict: to case particle function and lexical nouns.
; I'm sure this will add loads of ambiguity. Only for parsing spoken Japanese!
; (MS 2004-12-14) RELS and HCONS and HOOK should come from the argument.
; (MS 2004-12-16) Maybe this adds too much ambiguity. Taking it out again.
; (FCB 2006-19-31) can we restrict it to adjacent arguments ?
;pp_np_rule-case := empty-preposition-type &
; [SYNSEM [LOCAL [CAT [HEAD empty-case-p_head & [CASE ga-or-wo],
; VAL.COMPS < >],
; CONT [HOOK #hook,
; RELS #rels,
; HCONS #hcons]],
; NON-LOCAL.ROBUST +],
; C-CONT.RELS <! !>,
; ARGS.FIRST.SYNSEM [LEX +,
; LOCAL [CONT [HOOK #hook,
; RELS #rels,
; HCONS #hcons],
; CAT.VAL saturated]]].
; (ERB 2001-12-13) For cases where the splitter can't break apart
; two sentences for lack of punctuation. This rule allows two sentences
; be coordinated with no overt marking of it. In pet/japanese.set
; it is marked as a spanning-only-rule so that it can only apply at
; the very top of the tree. I guess that won't be true in the lkb??
; (ERB 2001-12-13) This is causing some memory problems with cheap.
; Perhaps the problem is that nothing is requiring the second daughter
; to be a sentence. Adding that.
; (ERB 2001-12-13) Still causing memory problems, so commenting this
; out (as well as its ranking in japanese.set) until I can consult
; with Uli on Monday.
; (ERB 2001-12-17) Uli fixed the problem: the spanning-only-rules
; condition wasn't really being checked. But this is still producing
; way too many parses. Part of the problem is that te forms are
; [H-TENSE tensed] instead of [H-TENSE te] (in order to get the pure
; aspect analysis of the aspect auxiliaries). So, try blocking the
; te forms in the left daughter here by requiring [FIN +].
; (ERB 2002-01-02) Don't want this to be [NUCL +] on the
; second daughter, if hadj rules are going to be [NUCL -].
; (MS 2002-1-7) I'll restrict the rule a little bit further to non-adjectives.
; Also, both verbs are MAIN-PRD +. This prevents NA to be the first verb.
; MODUS nonoperative prevents NAI to be the first verb. It has to be seen if that is too restrictive...
; (MS 2002-1-7) It is: "表示できませんでした申し訳ございません.
; (MS 2002-1-10) Another restriction is EMPTY - in the head of the first argument.
; This prevents an empty-cop-id-rule or an empty-cop-rule to be applied to this.
; (ERB 2002-01-22) NUCL conj now on supertype.
; (ERB 2002-01-29) If we want runon_s with SAPs on the end, we need the
; head-dtr here to be utt-valid instead of verb_head. (Remembering that
; in cheap at least this rule is restricted to spanning-only-edges.)
; (MS 2002-05-28) If we want the first sentence to have an SAP, we need utt-valid there, too.
; (MS 2002-06-05) Maybe that's not a good idea, leads to spurious ambiguity in, e.g.,
; ご飯を食べるかよい
; (MS 2002-06-19) But we seem to need this anyway: "口座振替依頼書送ってください,お願いします"
; (MS 2002-06-20) Trying to restrict this with NUCL nucl_plus.
; (FCB 2005-06-28) take this one out for the time being
; runon_s := sentence-coord-type &
; [SYNSEM [LOCAL [BAR +,
; CAT.VAL saturated],
; NON-LOCAL.ROBUST +],
; ARGS <[SYNSEM.LOCAL [CAT [HEAD utt-valid & [H-TENSE present-past-or-future,
; FIN +,
; EMPTY -,
; ADJ -,
; MAIN-PRD +],
; VAL saturated],
; BAR +,
; NUCL nucl_plus],
; J-NEEDS-AFFIX -],
; [SYNSEM synsem & [LOCAL [BAR +,
; CAT [ HEAD utt-valid &
; [ H-TENSE present-past-or-future,
; MAIN-PRD +,
; EMPTY -],
; VAL saturated]]],
; J-NEEDS-AFFIX -]>].