NLTK 2.0 classifier batch classifier method

Question

When I run this code it throws an error which I believe is due to the batch_classify method not being present in NLTK 3.0. I curious how to resolve this type of issue where something from an older version disappears in the newer version.

def accuracy(classifier, gold):
    results = classifier.batch_classify([fs for (fs,l) in gold])
    correct = [l==r for ((fs,l), r) in zip(gold, results)]
    if correct:
        return float(sum(correct))/len(correct)
    else:
        return 0

def apr (classifier, gold):
    results = classifier.batch_classify([fs for (fs,l) in gold])
    #gold_class_dict = defaultdict(list)
    #classifier_class_dict = {}
    sys_correct_dict = Counter()
    num_guessed = Counter()
    gold_num = Counter()
    num_right = 0
    total = 0

Why are you using an old version of nltk? Try pip install -U nltk on the terminal / powershell. — alvas
@alvas I think the question is about porting code from NLTK 2.0 to NLTK 3.0. — lenz
Hopefully this would help: github.com/nltk/nltk/wiki/Porting-your-code-to-NLTK-3.0 — alvas

lenz lenz · Accepted Answer · 2016-10-22T20:36:34

The method was renamed to classify_many() (I couldn't find documentation of NLTK 2.0 to check it, but I'm pretty sure that's what happened).

You have to replace all occurrences of batch_classify(...) with classify_many(...) in your code. When moving from one major version of a library to another, you have to expect this kind of backwards-incompatible changes; they should ideally be documented in the changelog. However, I have to admit that in the past NLTK introduced backwards-incompatible changes even between minor versions, which I think is bad practice.

NLTK 2.0 classifier batch classifier method

1 Answers