nltk31_twitter情感分析

已经生成4个pickle文件，分别为documents,word_features,originalnaivebayes5k,featurests

其中featurests容量最大，3百多兆，如果扩大5000特征集，容量继续扩大，准确性也提供

https://www.pythonprogramming.net/sentiment-analysis-module-nltk-tutorial/

Creating a module for Sentiment Analysis with NLTK

# -*- coding: utf-8 -*-
"""
Created on Sat Jan 14 09:59:09 2017

@author: daxiong
"""

#File: sentiment_mod.py

import nltk
import random
import pickle
from nltk.tokenize import word_tokenize

documents_f = open("documents.pickle", "rb")
documents = pickle.load(documents_f)
documents_f.close()

word_features5k_f = open("word_features5k.pickle", "rb")
word_features = pickle.load(word_features5k_f)
word_features5k_f.close()

def find_features(document):
    words = word_tokenize(document)
    features = {}
    for w in word_features:
        features[w] = (w in words)

    return features

featuresets_f = open("featuresets.pickle", "rb")
featuresets = pickle.load(featuresets_f)
featuresets_f.close()

random.shuffle(featuresets)
print(len(featuresets))

testing_set = featuresets[10000:]
training_set = featuresets[:10000]

open_file = open("originalnaivebayes5k.pickle", "rb")
classifier = pickle.load(open_file)
open_file.close()

def sentiment(text):
    feats = find_features(text)
    return classifier.classify(feats)

print(sentiment("This movie was awesome! The acting was great, plot was wonderful, and there were pythons...so yea!"))
print(sentiment("This movie was utter junk. There were absolutely 0 pythons. I don‘t see what the point was at all. Horrible movie, 0/10"))

测试效果

还是比较准，the movie is good 测试不准，看来要改进算法，考虑用频率分析和过滤垃圾词来提高准确率