kanjirz50/viet-morphological-analysis-crf

Vietnamese morphological analysis with using CRF

Vietnamese Morphological Anlyzer with CRF

Vietnamese morphological analysis with using CRF

Requirements

• Python 2
• CRF++ 0.58
• model file

Usage

Please download model file to ./path/to/model_file

from crfpp import CRF_PP

viet_morph_analyzer = CRF_PP('/path/to/model_file')

sentence = 'Số điện thoại của trường'
result = viet_morph_analyzer.analyze(sentence)

How to make model file

Get tagged Corpus

• vnPOS

Convert format from vnPOS to IOB2 tag format

Corpus is given below format.

Tấp_nập//JJ sắm//VB đtdđ//NN đầu//NN năm//NC
...

Change format to IOB2 tag format.(Use only I tag and B tag.)

% cat vnPOS.txt | python ./utils/vnPOS_to_iob2.py > vnPOS.iob2
# Output likes below one.
Tấp		B-JJ
nập		I_JJ
sắm		B-VB
đtdđ	B-NN
đầu		B-NN
năm		B-NC

...

Training with CRF++

% crf_learn ./crf_template ./vnPOS.iob2 ./vnPOS.crfpp.model

"crf_template" is a feature template files. You can change features.