CODEWITHSUNDEEP
pythonnltksentiment-analysis
Jitendra
Jitendra

Reputation: 307

How to tweak the NLTK Python code in such a way that I train the classifier only once

I have tried performing Sentiment Analysis on a huge data set which is about 10000 sentences. Now, when I use the NLTK Python code for performing training and testing using Naive Bayes, I will have train the classifier each time when I need to classify a set of new sentences. This is taking a lot of time.Is there a way I can take the output of the training part and then use it for classification which would save a lot of time.This is the NLTK code that I have used.

import nltk
import re
import csv
#Read the tweets one by one and process it



def processTweet(tweet):
    # process the tweets
    #convert to lower case
    tweet = tweet.lower()
    #Convert www.* or https?://* to URL
    tweet = re.sub('((www\.[\s]+)|(https?://[^\s]+))','URL',tweet)
    #Convert @username to AT_USER
    tweet = re.sub('@[^\s]+','AT_USER',tweet)
    #Remove additional white spaces
    tweet = re.sub('[\s]+', ' ', tweet)
    #Replace #word with word
    tweet = re.sub(r'#([^\s]+)', r'\1', tweet)
    #trim
    tweet = tweet.strip('\'"')
    return tweet

def replaceTwoOrMore(s):
    #look for 2 or more repetitions of character and replace with the character itself
    pattern = re.compile(r"(.)\1{1,}", re.DOTALL)
    return pattern.sub(r"\1\1", s)
#end

#start getStopWordList
def getStopWordList(stopWordListFileName):
    #read the stopwords file and build a list
    stopWords = []
    stopWords.append('AT_USER')
    stopWords.append('url')
    stopWords.append('URL')
    stopWords.append('rt')

    fp = open(stopWordListFileName)
    line = fp.readline()
    while line:
        word = line.strip()
        stopWords.append(word)
        line = fp.readline()
    fp.close()
    return stopWords
#end

#start getfeatureVector
def getFeatureVector(tweet):
    featureVector = []
    #split tweet into words
    words = tweet.split()
    for w in words:
        #replace two or more with two occurrences
        w = replaceTwoOrMore(w)
        #strip punctuation
        w = w.strip('\'"?,.')
        #check if the word starts with an alphabet
        val = re.search(r"^[a-zA-Z][a-zA-Z0-9]*$", w)

        #ignore if it is a stop word
        if(w in stopWords or val is None):
            continue
        else:
            featureVector.append(w.lower())
    return featureVector
#end

def extract_features(tweet):
    tweet_words = set(tweet)
    features = {}
    for word in featureList:
        features['contains(%s)' % word] = (word in tweet_words)
    return features

inpTweets = csv.reader(open('sheet3.csv', 'rb'), delimiter=',')
stopWords = getStopWordList('stopwords.txt')
featureList = []



# Get tweet words
tweets = []
for row in inpTweets:
    sentiment = row[0]
    tweet = row[1]
    processedTweet = processTweet(tweet)
    featureVector = getFeatureVector(processedTweet)
    featureList.extend(featureVector)
    tweets.append((featureVector, sentiment));
#end loop

# Remove featureList duplicates
featureList = list(set(featureList))

# Extract feature vector for all tweets in one shote
training_set = nltk.classify.util.apply_features(extract_features, tweets)

NBClassifier = nltk.NaiveBayesClassifier.train(training_set)

ft = open("april2.tsv")
line = ft.readline()

fo = open("dunno.tsv", "w")

fo.seek(0,0)
while line:
    testTweet = line
    processedTestTweet = processTweet(testTweet)
    line1 = fo.write( NBClassifier.classify(extract_features(getFeatureVector(processedTestTweet))) + "\n");
    line = ft.readline()

fo.close()
ft.close()

Upvotes: 1

Views: 789

Answers (2)

alvas
alvas

Reputation: 122330

If you want to stick with NLTK, try pickle, e.g. https://spaghetti-tagger.googlecode.com/svn/spaghetti.py, see https://docs.python.org/2/library/pickle.html :

#-*- coding: utf8 -*-

from nltk import UnigramTagger as ut
from nltk import BigramTagger as bt
from cPickle import dump,load

def loadtagger(taggerfilename):
    infile = open(taggerfilename,'rb')
    tagger = load(infile); infile.close()
    return tagger

def traintag(corpusname, corpus):
    # Function to save tagger.
    def savetagger(tagfilename,tagger):
        outfile = open(tagfilename, 'wb')
        dump(tagger,outfile,-1); outfile.close()
        return
    # Training UnigramTagger.
    uni_tag = ut(corpus)
    savetagger(corpusname+'_unigram.tagger',uni_tag)
    # Training BigramTagger.
    bi_tag = bt(corpus)
    savetagger(corpusname+'_bigram.tagger',bi_tag)
    print "Tagger trained with",corpusname,"using" +\
                "UnigramTagger and BigramTagger."
    return

Otherwise, try other machine learning libraries such as sklearn or shogun

Upvotes: 2

FoxRocks
FoxRocks

Reputation: 68

The Naive Bayes Classifier module in NLTK is breathtakingly slow because it's a pure Python implementation. For this reason, consider using a different Machine Learning (ML) library like sci-kit learn.

YS-L's tip is good for using cPickle is good for your purposes at the moment but, if you ever have to retrain the classifier, it'd probably be best to switch to a different Naive Bayes implementation.

Upvotes: 1

Related Questions