我通过转换情绪分析脚本来使用它们来教我自己(可能是我的第一个错误)类和方法.

我以为我已经掌握了所有方法,但我一直在努力

未定义全局名称“get_bigram_word_feats”

我确定我也会收到get_word_feats的错误,如果它到那么远的话.

我正在撞击这个伟大的时间.我尝试删除staticmethod并添加self.我究竟做错了什么？

这是我的代码：

def word_feats(words):
    return dict([(word,True) for word in words])


class SentClassifier:

    def __init__(self,name,location):
        self.name = name
        self.location = location
        self.fullpath = location + "/" + name

    def doesexist(self):
        return os.path.isfile(self.fullpath)

    def save_classifier(self):
        rf = open(self.fullpath,'wb')
        pickle.dump(self.fullpath,rf)
        rf.close()

    def load_classifier(self):
        sf = open(self.fullpath,'rb')
        sclassifier = pickle.load(sf)
        sf.close()
        return sclassifier


class Training:

    def __init__(self,neg,pos):
        self.neg = neg
        self.pos = pos
        self.negids = open(self.neg,'rb').read().splitlines(True)
        self.posids = open(self.pos,'rb').read().splitlines(True)
        self.exclude = set(string.punctuation)
        self.exclude = self.exclude,'...'
        self.swords = stopwords.words('english')

    def tokens(self,words):
        words = [w for w in nltk.word_tokenize(words) if w not in self.exclude and len(w) > 1
            and w not in self.swords and wordnet.synsets(w)]
        return words

    def idlist(self,words):
        thisidlist = [self.tokens(tf) for tf in words]
        return thisidlist

    @staticmethod
    def get_word_feats(words):
        return dict([(word,True) for word in words])

    @staticmethod
    def get_bigram_word_feats(twords,score_fn=BigramAssocMeasures.chi_sq,tn=200):
        words = [w for w in twords]
        bigram_finder = BigramCollocationFinder.from_words(words)
        bigrams = bigram_finder.nbest(score_fn,tn)
        return dict([(ngram,True) for ngram in itertools.chain(words,bigrams)])

    @staticmethod
    def label_feats(thelist,label):
        return [(get_word_feats(lf),label) for lf in thelist]

    @staticmethod
    def label_grams(thelist,label):
        return [(get_bigram_word_feats(gf),label) for gf in thelist()]

    @staticmethod
    def combinegrams(grams,feats):
        for g in grams():
            feats.append(g)
        return feats

    def negidlist(self):
        return self.idlist(self.negids)

    def posidlist(self):
        return self.idlist(self.posids)

    def posgrams(self):
        return self.label_grams(self.posidlist,'pos')

    def neggrams(self):
        return self.label_grams(self.negidlist,'neg')

    def negwords(self):
        return self.label_feats(self.negidlist,'neg')

    def poswords(self):
        return self.label_feats(self.posidlist,'pos')

    def negfeats(self):
        return self.combinegrams(self.neggrams,self.negwords)

    def posfeats(self):
        return self.combinegrams(self.posgrams,self.poswords)

starttime = time.time()

myclassifier = SentClassifier("sentanalyzer.pickle","classifiers")

if myclassifier.doesexist() is False:
    print "training new classifier"
    trainset = Training('data/neg.txt','data/pos.txt')
    negfeats = trainset.negfeats()
    posfeats = trainset.posfeats()
    negcutoff = len(negfeats) * 8 / 10
    poscutoff = len(posfeats) * 8 / 10

    trainfeats = negfeats[:negcutoff] + posfeats[:poscutoff]
    testfeats = negfeats[negcutoff:] + posfeats[poscutoff:]
    print 'train on %d instances,test on %d instances' % (len(trainfeats),len(testfeats))

    classifier = NaiveBayesClassifier.train(trainfeats)
    print 'accuracy:',nltk.classify.util.accuracy(classifier,testfeats)
    myclassifier.save_classifier()

else:
    print "using existing classifier"
    classifier = myclassifier.load_classifier()

classifier.show_most_informative_features(20)
mystr = "16 steps to an irresistible sales pitch,via @vladblagi: slidesha.re/1bVV7OS"
myfeat = word_feats(nltk.word_tokenize(mystr))
print classifier.classify(myfeat)

probd = classifier.prob_classify(myfeat)

print probd.prob('neg')
print probd.prob('pos')

donetime = time.time() - starttime

print donetime