# -*- coding:utf-8 -*-
# Author: hankcs
# Date: 2019-12-26 23:20
from hanlp.pretrained.glove import GLOVE_6B_100D
from hanlp.components.parsers.biaffine_parser import BiaffineDependencyParser, BiaffineSemanticDependencyParser
from tests import cdroot
cdroot()
save_dir = 'data/model/semeval15_biaffine_psd'
parser = BiaffineSemanticDependencyParser()
# parser.fit('data/semeval15/en.psd.train.conll', 'data/semeval15/en.psd.dev.conll', save_dir,
# pretrained_embed={'class_name': 'HanLP>Word2VecEmbedding',
# 'config': {
# 'trainable': False,
# 'embeddings_initializer': 'zero',
# 'filepath': GLOVE_6B_100D,
# 'expand_vocab': True,
# 'lowercase': False,
# 'unk': 'unk',
# 'normalize': True,
# 'name': 'glove.6B.100d'
# }},
# # lstm_dropout=0,
# # mlp_dropout=0,
# # embed_dropout=0,
# epochs=1
# )
parser.load(save_dir)
parser.save_meta(save_dir)
parser.transform.summarize_vocabs()
sentence = [('Is', 'VBZ'), ('this', 'DT'), ('the', 'DT'), ('future', 'NN'),
# -*- coding:utf-8 -*-
# Author: hankcs
# Date: 2019-12-26 23:20
from hanlp.datasets.parsing.semeval2016 import SEMEVAL2016_TEXT_TRAIN, SEMEVAL2016_TEXT_VALID, SEMEVAL2016_TEXT_TEST
from hanlp.pretrained.word2vec import SEMEVAL16_EMBEDDINGS_300_TEXT_CN
from hanlp.utils.tf_util import nice
nice()
from hanlp.components.parsers.biaffine_parser import BiaffineSemanticDependencyParser
from tests import cdroot
cdroot()
save_dir = 'data/model/sdp/semeval16-text'
parser = BiaffineSemanticDependencyParser()
parser.fit(
SEMEVAL2016_TEXT_TRAIN,
SEMEVAL2016_TEXT_VALID,
save_dir,
pretrained_embed={
'class_name': 'HanLP>Word2VecEmbedding',
'config': {
'trainable': False,
'embeddings_initializer': 'zero',
'filepath': SEMEVAL16_EMBEDDINGS_300_TEXT_CN,
'expand_vocab': True,
'lowercase': True,
'normalize': True,
}
},
)
parser.load(save_dir)
# -*- coding:utf-8 -*-
# Author: hankcs
# Date: 2020-01-01 18:26
from hanlp.components.parsers.biaffine_parser import BiaffineSemanticDependencyParser
from hanlp.pretrained.glove import GLOVE_6B_100D
from tests import cdroot
cdroot()
save_dir = 'data/model/sdp/semeval15_biaffine_dm'
parser = BiaffineSemanticDependencyParser()
parser.fit(
'data/semeval15/en.dm.train.conll',
'data/semeval15/en.dm.dev.conll',
save_dir,
pretrained_embed={
'class_name': 'HanLP>Word2VecEmbedding',
'config': {
'trainable': False,
'embeddings_initializer': 'zero',
'filepath': GLOVE_6B_100D,
'expand_vocab': True,
'lowercase': True,
'normalize': True,
}
},
)
sentence = [('Is', 'VBZ'), ('this', 'DT'), ('the', 'DT'), ('future', 'NN'),
('of', 'IN'), ('chamber', 'NN'), ('music', 'NN'), ('?', '.')]
print(parser.predict(sentence))
parser.evaluate('data/semeval15/en.id.dm.auto.conllu', save_dir)
parser.evaluate('data/semeval15/en.ood.dm.auto.conllu', save_dir)
# -*- coding:utf-8 -*-
# Author: hankcs
# Date: 2020-01-01 18:26
from hanlp.components.parsers.biaffine_parser import BiaffineSemanticDependencyParser
from hanlp.pretrained.glove import GLOVE_6B_100D
from tests import cdroot
cdroot()
save_dir = 'data/model/sdp/semeval15_biaffine_pas'
parser = BiaffineSemanticDependencyParser()
parser.fit(
'data/semeval15/en.pas.train.conll',
'data/semeval15/en.pas.dev.conll',
save_dir,
pretrained_embed={
'class_name': 'HanLP>Word2VecEmbedding',
'config': {
'trainable': False,
'embeddings_initializer': 'zero',
'filepath': GLOVE_6B_100D,
'expand_vocab': True,
'lowercase': True,
'normalize': True,
}
},
)
parser.load(
save_dir
) # disable variational dropout during evaluation so as to use CudaLSTM
sentence = [('Is', 'VBZ'), ('this', 'DT'), ('the', 'DT'), ('future', 'NN'),
('of', 'IN'), ('chamber', 'NN'), ('music', 'NN'), ('?', '.')]