Identifying Dialogue Act Type

Natural Language Processing with Python

Chapter  6.2

import nltk
from nltk.corpus import nps_chat as nchat

def dialogue_act_features(post):
    features={}
    for word in nltk.word_tokenize(post):
        features['contains(%s)' % word.lower()] = True
    return features
    
def test_dialogue_act_types():
    posts=nchat.xml_posts()[:10000]
    featuresets = [(dialogue_act_features(post.text),post.get('class')) 
                    for post in posts]
    size=int(len(featuresets)*0.1)
    train_set, test_set = featuresets[size:],featuresets[:size]
    classifier = nltk.NaiveBayesClassifier.train(train_set)
    print nltk.classify.accuracy(classifier,test_set)
    classifier.show_most_informative_features(5)

运行结果：

0.668
Most Informative Features
contains(hi) = True Greet : System = 408.2 : 1.0
contains(>) = True Other : System = 384.6 : 1.0
contains(empty) = True Other : System = 339.4 : 1.0
contains(part) = True System : Statem = 302.0 : 1.0
contains(no) = True nAnswe : System = 262.3 : 1.0