chunking module attached to NDPredTrans. Might be a nice alternative to robustness. But the current version uses variants too much.

This commit is contained in:
aarne
2014-02-28 16:13:12 +00:00
parent 7f079152bc
commit e20bf3c0a7
5 changed files with 141 additions and 26 deletions

View File

@@ -0,0 +1,48 @@
abstract Chunk = RGLBase - [Pol,Tense], NDPred ** {
cat
Chunks ;
Chunk ;
fun
OneChunk : Chunk -> Chunks ;
PlusChunk : Chunk -> Chunks -> Chunks ;
ChunkPhr : Chunks -> Phr ;
fun
AP_Chunk : AP -> Chunk ;
AdA_Chunk : AdA -> Chunk ;
Adv_Chunk : Adv -> Chunk ;
AdV_Chunk : AdV -> Chunk ;
AdN_Chunk : AdN -> Chunk ;
CN_Pl_Chunk : CN -> Chunk ;
CN_Sg_Chunk : CN -> Chunk ;
CN_Pl_Gen_Chunk : CN -> Chunk ;
CN_Sg_Gen_Chunk : CN -> Chunk ;
Conj_Chunk : Conj -> Chunk ;
IAdv_Chunk : IAdv -> Chunk ;
IP_Chunk : IP -> Chunk ;
NP_Chunk : NP -> Chunk ;
NP_Gen_Chunk : NP -> Chunk ;
Numeral_Chunk : Numeral -> Chunk ;
Ord_Chunk : Ord -> Chunk ;
Predet_Chunk : Predet -> Chunk ;
Prep_Chunk : Prep -> Chunk ;
RP_Chunk : RP -> Chunk ;
Subj_Chunk : Subj -> Chunk ;
VP_Chunk : VP -> Chunk ;
VP_np_Chunk : PrVP_np -> Chunk ;
VP_none_Chunk : PrVP_none -> Chunk ;
VP_s_Chunk : PrVP_s -> Chunk ;
neg_Chunk : Chunk ;
copula_Chunk : Chunk ;
past_copula_Chunk : Chunk ;
future_Chunk : Chunk ;
cond_Chunk : Chunk ;
perfect_Chunk : Chunk ;
past_perfect_Chunk : Chunk ;
}

View File

@@ -0,0 +1,64 @@
concrete ChunkEng of Chunk =
RGLBaseEng - [Pol,Tense],
NDPredEng
** open (PI=PredInstanceEng), ResEng, Prelude in {
lincat
Chunks = {s : Str} ;
Chunk = {s : Str};
lin
OneChunk c = c ;
PlusChunk c cs = cc2 c cs ;
---- ChunkPhr c = c ;
lin
AP_Chunk ap = allAgrSS (\a -> ap.s ! a) ;
AdA_Chunk ada = ada ;
Adv_Chunk adv = adv ;
AdV_Chunk adv = adv ;
AdN_Chunk adn = adn ;
CN_Pl_Chunk cn = ss (cn.s ! Pl ! Nom) ;
CN_Sg_Chunk cn = ss (cn.s ! Sg ! Nom) ;
CN_Pl_Gen_Chunk cn = ss (cn.s ! Pl ! Gen) ;
CN_Sg_Gen_Chunk cn = ss (cn.s ! Sg ! Gen) ;
Conj_Chunk conj = ss conj.s2 ;
IAdv_Chunk iadv = iadv ;
IP_Chunk ip = ss (ip.s ! NCase Nom) ;
NP_Chunk np = ss (np.s ! NCase Nom) | ss (np.s ! NPAcc) ;
NP_Gen_Chunk np = ss (np.s ! NCase Gen) | ss (np.s ! NPNomPoss) ;
Numeral_Chunk num = ss (num.s ! NCard ! Nom) | ss (num.s ! NCard ! Gen) ;
Ord_Chunk ord = ss (ord.s ! Nom) | ss (ord.s ! Gen) ;
Predet_Chunk predet = predet ;
Prep_Chunk prep = prep ;
RP_Chunk rp = ss (rp.s ! RC Neutr (NCase Nom)) | ss (rp.s ! RPrep Masc) ; ----
Subj_Chunk subj = subj ;
VP_none_Chunk, VP_np_Chunk, VP_s_Chunk = \vp ->
let verb = vp.v ! (PI.VASgP1 | PI.VASgP3 | PI.VAPl)
in
allAgrSS (\a ->
verb.p1 ++ verb.p2 ++ vp.adV ++ verb.p3 ++ vp.part ++
vp.adj ! a ++ vp.c1 ++ vp.obj1.p1 ! a ++ vp.c2 ++ vp.obj2.p1 ! a ++ vp.adv ++ vp.ext
) ;
neg_Chunk = ss "not" | ss "doesn't" | ss "don't" ;
copula_Chunk = ss "is" | ss "are" | ss "am" ;
past_copula_Chunk = ss "was" | ss "were" ;
future_Chunk = ss "will" ;
cond_Chunk = ss "would" ;
perfect_Chunk = ss "has" | ss "have" ;
past_perfect_Chunk = ss "had" ;
oper
allAgrSS : (Agr -> Str) -> SS = \f ->
ss (f (AgP3Sg Masc)) ;
---- | ss (f (AgP3Sg Fem)) | ss (f (AgP3Sg Neutr)) |
---- ss (f (AgP1 Sg)) | ss (f (AgP1 Pl)) | ss (f (AgP2 Sg)) | ss (f (AgP2 Pl)) |
---- ss (f (AgP3Pl)) ;
}

View File

@@ -53360,7 +53360,6 @@ zealand_PN 1.8658456945610598e-5
zimbabwe_PN 1.8658456945610598e-5
zoe_PN 1.8658456945610598e-5
zurich_PN 2.9853531112976957e-4
PhrUtt 1.0
PNeg 2.5657769160446448e-2
PPos 0.9743422308395535
AgentPastPartAP_none 1.4729709824716454e-4
@@ -65214,28 +65213,32 @@ work_VV 1.8203883495145632e-3
write_VV 3.6407766990291263e-4
NoVoc 0.99
VocNP 0.01
AP 3.030247037335075e-3
AdA 9.109495956621799e-4
AdV 4.7404486565341235e-3
Adv 5.328685581035249e-3
CN 0.12420511332822734
Conj 1.153607720713571e-2
Det 6.831013306700838e-2
IAdv 2.5406809209644978e-5
IP 7.206294975826576e-5
N 0.12385449936113424
NP 0.16918717612092532
Num 7.879021089499412e-2
Ord 7.825297236570654e-4
Predet 2.7716519137794524e-4
Prep 4.62935160899013e-2
Pron 1.4304957469001383e-2
Quant 7.624860609005836e-2
RP 3.649803628461909e-3
RS 4.061393937658158e-3
PrS 5.144324534570353e-2
Subj 5.751177721092363e-4
AP_Chunk 3.030247037335075e-3
AdA_Chunk 9.109495956621799e-4
AdV_Chunk 4.7404486565341235e-3
Adv_Chunk 5.328685581035249e-3
CN_Chunk 0.12420511332822734
Conj_Chunk 1.153607720713571e-2
IAdv_Chunk 2.5406809209644978e-5
IP_Chunk 7.206294975826576e-5
NP_Chunk 0.16918717612092532
Num_Chunk 7.879021089499412e-2
Ord_Chunk 7.825297236570654e-4
Predet_Chunk 2.7716519137794524e-4
Prep_Chunk 4.62935160899013e-2
RP_Chunk 3.649803628461909e-3
Subj_Chunk 5.751177721092363e-4
VP_Chunk 0.1
PhrUtt 0.99
PhrChunk 0.01
OneChunk 0.8
PlusChunk 0.2
V 2.0368869914365196e-2
V2 2.869445032137304e-2
neg_Chunk 5.751177721092363e-4
copula_Chunk 5.751177721092363e-4
past_copula_Chunk 5.751177721092363e-4
future_Chunk 5.751177721092363e-4
cond_Chunk 5.751177721092363e-4
perfect_Chunk 5.751177721092363e-4
past_perfect_Chunk 5.751177721092363e-4

View File

@@ -4,7 +4,7 @@ abstract NDTrans =
NDLift
,Extensions [CN,NP,AdA,AdV,CompoundCN,AdAdV,UttAdV,ApposNP]
,Dictionary - [Pol,Tense]
,Chunk
** {
flags
startcat=Phr;

View File

@@ -4,7 +4,7 @@ concrete NDTransEng of NDTrans =
NDLiftEng
,ExtensionsEng [CN,NP,AdA,AdV,CompoundCN,AdAdV,UttAdV,ApposNP]
,DictionaryEng - [Pol,Tense]
,ChunkEng
** open ResEng, PredInstanceEng, Prelude, (Pr = PredEng) in {
flags