我需要在大型数据集上运行nltk.pos_tag,并且需要使其输出像斯坦福标记器提供的那样。
例如,在运行以下代码时,我有:
import nltk text=nltk.word_tokenize("We are going out.Just you and me.") print nltk.pos_tag(text)
输出为:[[‘We’,’PRP’),(’are’,’VBP’),(’going’,’VBG’),(’out.Just’,’IN’),(’you ‘,’PRP’),(’and’,’CC’),(’me’,’PRP’),(’。’,’。’)]
在这种情况下,我需要像这样:
We/PRP are/VBP going/VBG out.Just/NN you/PRP and/CC me/PRP ./.
我更喜欢不使用字符串函数,并且需要直接输出,因为文本数量如此之多,并且给处理增加了很多时间上的复杂性
简而言之:
' '.join([word + '/' + pos for word, pos in tagged_sent]
总而言之:
我认为您对使用字符串函数来连接字符串的想法过高,这实际上并不那么昂贵。
import time from nltk.corpus import brown tagged_corpus = brown.tagged_sents() start = time.time() with open('output.txt', 'w') as fout: for i, sent in enumerate(tagged_corpus): print(' '.join([word + '/' + pos for word, pos in sent]), end='\n', file=fout) end = time.time() - start print (i, end)
在我的笔记本电脑上,棕色语料库的所有57339个句子花了2.955秒。
[出]:
$ head -n1 output.txt The/AT Fulton/NP-TL County/NN-TL Grand/JJ-TL Jury/NN-TL said/VBD Friday/NR an/AT investigation/NN of/IN Atlanta's/NP$ recent/JJ primary/NN election/NN produced/VBD ``/`` no/AT evidence/NN ''/'' that/CS any/DTI irregularities/NNS took/VBD place/NN ./.
但是使用字符串将单词和POS连接起来会在以后需要读取标记的输出时引起麻烦,例如
>>> from nltk import pos_tag >>> tagged_sent = pos_tag('cat / dog'.split()) >>> tagged_sent_str = ' '.join([word + '/' + pos for word, pos in tagged_sent]) >>> tagged_sent_str 'cat/NN //CD dog/NN' >>> [tuple(wordpos.split('/')) for wordpos in tagged_sent_str.split()] [('cat', 'NN'), ('', '', 'CD'), ('dog', 'NN')]
如果要保存标记的输出然后再阅读,最好使用pickle保存tagd_output的方法,例如
pickle
>>> import pickle >>> tagged_sent = pos_tag('cat / dog'.split()) >>> with open('tagged_sent.pkl', 'wb') as fout: ... pickle.dump(tagged_sent, fout) ... >>> tagged_sent = None >>> tagged_sent >>> with open('tagged_sent.pkl', 'rb') as fin: ... tagged_sent = pickle.load(fin) ... >>> tagged_sent [('cat', 'NN'), ('/', 'CD'), ('dog', 'NN')]