usage CRFSUITE - beyondnlp/nlp GitHub Wiki

crfsuite

* http://www.chokkan.org/software/crfsuite/

crfsuite learn [옵션] [학습문서]

-m 모델명
-g N : N개의 그룹으로 학습문서를 분리
-x : cross-validation을 하기 위한 옵션
-p key=val : 내부의 프로퍼티를 설정하기 위한 옵션( ex> mincount설정 )
- crfsuite learn -H 으로 설정 가능한 프로퍼티 리스트를 확인할수 있음

loat feature.minfreq = 0.000000;
The minimum frequency of features.
 
int feature.possible_states = 0;
Force to generate possible state features.
 
int feature.possible_transitions = 0;
Force to generate possible transition features.
 
float c1 = 0.000000;
Coefficient for L1 regularization.
 
float c2 = 1.000000;
Coefficient for L2 regularization.
 
int max_iterations = 2147483647;
The maximum number of iterations for L-BFGS optimization.
 
int num_memories = 6;
The number of limited memories for approximating the inverse hessian matrix.
 
float epsilon = 0.000010;
Epsilon for testing the convergence of the objective.
 
int period = 10;
The duration of iterations to test the stopping criterion.
 
float delta = 0.000010;
The threshold for the stopping criterion; an L-BFGS iteration stops when the
improvement of the log likelihood over the last ${period} iterations is no
greater than this threshold.
 
string linesearch = MoreThuente;
The line search algorithm used in L-BFGS updates:
{   'MoreThuente': More and Thuente's method,
    'Backtracking': Backtracking method with regular Wolfe condition,
    'StrongBacktracking': Backtracking method with strong Wolfe condition
}
 
 
int max_linesearch = 20;
The maximum number of trials for the line search algorithm.

-g : 알고리즘 ( lbfgs, l2sgd, ap, pa, arow )
-e M : M번째 그룹만 테스트에 사용, 나머지는 학습에 사용, 출력결과에 클래스별 precision, recall, f1-score를 표시

cross-validation을 실행중일때는 model파일이 생성되지 않음
- 을 stdin으로 사용할수 있음

crfsuite tag [옵션] [테스트문서]

-m 모델명 : 테스트에 사용할 모델
-t : 모델 평가 결과 리포트
-r : 테스트 문서에 있는 레이블을 같이 출력
-p : 레이블의 확률값 출력
-i : 아이템별 marginal 확률 출력
-q : 테스트 모드에서 태깅 결과 생략

crfsuite dump [모델]

chunking.py input format

Rockwell NNP B-NP
International NNP I-NP
Corp. NNP I-NP
's POS B-NP
Tulsa NNP I-NP
unit NN I-NP
said VBD B-VP
it PRP B-NP
signed VBD B-VP
a DT B-NP
$

chunking.py output

B-NP    w[0]=Rockwell   w[1]=International  w[2]=Corp.  w[0]|w[1]=Rockwell|International    pos[0]=NNP  pos[1]=NNP  pos[2]=NNP  pos[0]|pos[1]=NNP|NNP   pos[1]|pos[2]=NNP|NNP   pos[0]|pos[1]|pos[2]=NNP|NNP|NNP    __BOS__
I-NP    w[-1]=Rockwell  w[0]=International  w[1]=Corp.  w[2]='s w[-1]|w[0]=Rockwell|International   w[0]|w[1]=International|Corp.   pos[-1]=NNP pos[0]=NNP  pos[1]=NNP  pos[2]=POS  pos[-1]|pos[0]=NNP|NNP  pos[0]|pos[1]=NNP|NNP   pos[1]|pos[2]=NNP|POS   pos[-1]|pos[0]|pos[1]=NNP|NNP|NNP   pos[0]|pos[1]|pos[2]=NNP|NNP|POS
I-NP    w[-2]=Rockwell  w[-1]=International w[0]=Corp.  w[1]='s w[2]=Tulsa  w[-1]|w[0]=International|Corp.  w[0]|w[1]=Corp.|'spos[-2]=NNP   pos[-1]=NNP pos[0]=NNP  pos[1]=POS  pos[2]=NNP  pos[-2]|pos[-1]=NNP|NNP pos[-1]|pos[0]=NNP|NNP  pos[0]|pos[1]=NNP|POS   pos[1]|pos[2]=POS|NNP   pos[-2]|pos[-1]|pos[0]=NNP|NNP|NNP  pos[-1]|pos[0]|pos[1]=NNP|NNP|POS   pos[0]|pos[1]|pos[2]=NNP|POS|NNP
B-NP    w[-2]=International w[-1]=Corp. w[0]='s w[1]=Tulsa  w[2]=unit   w[-1]|w[0]=Corp.|'s w[0]|w[1]='s|Tulsa  pos[-2]=NNP pos[-1]=NNP pos[0]=POS  pos[1]=NNP  pos[2]=NN   pos[-2]|pos[-1]=NNP|NNP pos[-1]|pos[0]=NNP|POS  pos[0]|pos[1]=POS|NNP   pos[1]|pos[2]=NNP|NN    pos[-2]|pos[-1]|pos[0]=NNP|NNP|POS  pos[-1]|pos[0]|pos[1]=NNP|POS|NNP   pos[0]|pos[1]|pos[2]=POS|NNP|NN
I-NP    w[-2]=Corp. w[-1]='s    w[0]=Tulsa  w[1]=unit   w[2]=said   w[-1]|w[0]='s|Tulsa w[0]|w[1]=Tulsa|unit    pos[-2]=NNP pos[-1]=POS pos[0]=NNP  pos[1]=NN   pos[2]=VBD  pos[-2]|pos[-1]=NNP|POS pos[-1]|pos[0]=POS|NNP  pos[0]|pos[1]=NNP|NN    pos[1]|pos[2]=NN|VBD    pos[-2]|pos[-1]|pos[0]=NNP|POS|NNP  pos[-1]|pos[0]|pos[1]=POS|NNP|NN    pos[0]|pos[1]|pos[2]=NNP|NN|VBD
I-NP    w[-2]='s    w[-1]=Tulsa w[0]=unit   w[1]=said   w[2]=it w[-1]|w[0]=Tulsa|unit   w[0]|w[1]=unit|said pos[-2]=POS pos[-1]=NNP pos[0]=NN   pos[1]=VBD  pos[2]=PRP  pos[-2]|pos[-1]=POS|NNP pos[-1]|pos[0]=NNP|NN   pos[0]|pos[1]=NN|VBD    pos[1]|pos[2]=VBD|PRP   pos[-2]|pos[-1]|pos[0]=POS|NNP|NN   pos[-1]|pos[0]|pos[1]=NNP|NN|VBD    pos[0]|pos[1]|pos[2]=NN|VBD|PRP
B-VP    w[-2]=Tulsa w[-1]=unit  w[0]=said   w[1]=it w[2]=signed w[-1]|w[0]=unit|said    w[0]|w[1]=said|it   pos[-2]=NNP pos[-1]=NN  pos[0]=VBD  pos[1]=PRP  pos[2]=VBD  pos[-2]|pos[-1]=NNP|NN  pos[-1]|pos[0]=NN|VBD   pos[0]|pos[1]=VBD|PRP   pos[1]|pos[2]=PRP|VBD   pos[-2]|pos[-1]|pos[0]=NNP|NN|VBD   pos[-1]|pos[0]|pos[1]=NN|VBD|PRP    pos[0]|pos[1]|pos[2]=VBD|PRP|VBD
B-NP    w[-2]=unit  w[-1]=said  w[0]=it w[1]=signed w[2]=a  w[-1]|w[0]=said|it  w[0]|w[1]=it|signed pos[-2]=NN  pos[-1]=VBD pos[0]=PRP  pos[1]=VBD  pos[2]=DT   pos[-2]|pos[-1]=NN|VBD  pos[-1]|pos[0]=VBD|PRP  pos[0]|pos[1]=PRP|VBD   pos[1]|pos[2]=VBD|DT    pos[-2]|pos[-1]|pos[0]=NN|VBD|PRP   pos[-1]|pos[0]|pos[1]=VBD|PRP|VBD   pos[0]|pos[1]|pos[2]=PRP|VBD|DT
B-VP    w[-2]=said  w[-1]=it    w[0]=signed w[1]=a  w[2]=tentative  w[-1]|w[0]=it|signed    w[0]|w[1]=signed|a  pos[-2]=VBD pos[-1]=PRP pos[0]=VBD  pos[1]=DT   pos[2]=JJ   pos[-2]|pos[-1]=VBD|PRP pos[-1]|pos[0]=PRP|VBD  pos[0]|pos[1]=VBD|DT    pos[1]|pos[2]=DT|JJ pos[-2]|pos[-1]|pos[0]=VBD|PRP|VBD  pos[-1]|pos[0]|pos[1]=PRP|VBD|DT    pos[0]|pos[1]|pos[2]=VBD|DT|JJ
B-NP    w[-2]=it    w[-1]=signed    w[0]=a  w[1]=tentative  w[2]=agreement  w[-1]|w[0]=signed|a w[0]|w[1]=a|tentative   pos[-2]=PRP pos[-1]=VBD pos[0]=DT   pos[1]=JJ   pos[2]=NN   pos[-2]|pos[-1]=PRP|VBD pos[-1]|pos[0]=VBD|DT   pos[0]|pos[1]=DT|JJ pos[1]|pos[2]=JJ|NN pos[-2]|pos[-1]|pos[0]=PRP|VBD|DT   pos[-1]|pos[0]|pos[1]=VBD|DT|JJ pos[0]|pos[1]|pos[2]=DT|JJ|NN

초기에 CRFsuite source를 컴파일해서 사용했는데
- L-BFGS terminated with error code (-1001)
- L-BFGS terminated with error code (-998)
문제를 만나 고생을 했다. 설치상에 문제가 있는 것으로 판단되고 해결책은 찾지 못했다.
이런 경우 binary형식을 다운받아 사용하면 된다.