NLP | Classifier-based tagging

ClassifierBasedPOSTagger class:

• It is a subclass of ClassifierBasedTagger that uses classification technique to do part-of-speech tagging.
• From the words, features are extracted and then passed to an internal classifier.
• It classifies the features and returns a label i.e. a part-of-speech tag.
• The feature detector finds multiple length suffixes, does some regular expression matching, and looks at the unigram, bigram, and trigram history to produce a fairly complete set of features for each word

Code #1 : Using ClassifierBasedPOSTagger

from nltk.tag.sequential import ClassifierBasedPOSTagger
from nltk.corpus import treebank

# initializing training and testing set
train_data = treebank.tagged_sents()[:3000]
test_data = treebank.tagged_sents()[3000:]

tagging = ClassifierBasedPOSTagger(train = train_data)

a = tagging.evaluate(test_data)

print ("Accuracy : ", a)

Output :

Accuracy : 0.9309734513274336

ClassifierBasedPOSTagger class inherits from ClassifierBasedTagger and only implements a feature_detector() method. All the training and tagging is done in ClassifierBasedTagger.

Code #2 : Using MaxentClassifier

from nltk.classify import MaxentClassifier
from nltk.corpus import treebank

# initializing training and testing set
train_data = treebank.tagged_sents()[:3000]
test_data = treebank.tagged_sents()[3000:]


tagger = ClassifierBasedPOSTagger(
        train = train_sents, classifier_builder = MaxentClassifier.train)

a = tagger.evaluate(test_data)

print ("Accuracy : ", a)

Output :

Accuracy : 0.9258363911072739

custom feature detector detecting features
There are two ways to do it:

1. Subclass ClassifierBasedTagger and implement a feature_detector() method.
2. Pass a function as the feature_detector keyword argument into ClassifierBasedTagger at initialization.

Code #3 : Custom Feature Detector

from nltk.tag.sequential import ClassifierBasedTagger
from tag_util import unigram_feature_detector
from nltk.corpus import treebank

# initializing training and testing set
train_data = treebank.tagged_sents()[:3000]
test_data = treebank.tagged_sents()[3000:]

tag = ClassifierBasedTagger(
        train = train_data,
        feature_detector = unigram_feature_detector)

a = tagger.evaluate(test_data)

print ("Accuracy : ", a)

Output :

Accuracy : 0.8733865745737104