Statistical Learning and Text Classification with NLTK and scikit-learn
54 likes18,105 views
The document discusses statistical learning and text classification using NLTK and scikit-learn, highlighting applications such as spam filtering, sentiment analysis, and malware detection. It covers model building using features like binary occurrences, frequencies, and TF-IDF of n-grams, along with implementations of classifiers like naive Bayes and SVMs. Additionally, it explores performance results and online APIs for text processing, providing resources for further exploration.
![Typical features for text documents
● Tokenize document into list of words: uni-grams
['the', 'quick', 'brown', 'fox', 'jumps', 'over',
'the', 'lazy', 'dog']
● Then chose one of:
● Binary occurrences of uni-grams:
{'the': True, 'quick': True, ...}
● Frequencies of uni-grams: nb times word_i / nb
words in document:
{'the': 0.22, 'quick': 0.11, ...}
● TF-IDF of uni-grams (see next slides)](https://image.slidesharecdn.com/nltkscikit-learnpyconfr2010ogrisel-100828080703-phpapp02/85/Statistical-Learning-and-Text-Classification-with-NLTK-and-scikit-learn-5-320.jpg)
![Using a NLTK corpus
>>> from nltk.corpus import movie_reviews
>>> pos_ids = movie_reviews.fileids('pos')
>>> neg_ids = movie_reviews.fileids('neg')
>>> len(pos_ids), len(neg_ids)
1000, 1000
>>> print movie_reviews.raw(pos_ids[0])[:100]
films adapted from comic books have had plenty of success ,
whether they're about superheroes ( batm
>>> movie_reviews.words(pos_ids[0])
['films', 'adapted', 'from', 'comic', 'books', 'have', ...]](https://image.slidesharecdn.com/nltkscikit-learnpyconfr2010ogrisel-100828080703-phpapp02/85/Statistical-Learning-and-Text-Classification-with-NLTK-and-scikit-learn-10-320.jpg)
![The NLTK - Naïve Bayes Classifier
from nltk.classify import NaiveBayesClassifier
mr = movie_reviews
neg_examples = [(features(mr.words(i)), 'neg')
for i in neg_ids]
pos_examples = [(features(mr.words(i)), 'pos')
for i in pos_ids]
train_set = pos_examples + neg_examples
classifier = NaiveBayesClassifier.train(train_set)
# later on a previously unseed document
predicted_label = classifier.classify(new_doc_features)](https://image.slidesharecdn.com/nltkscikit-learnpyconfr2010ogrisel-100828080703-phpapp02/85/Statistical-Learning-and-Text-Classification-with-NLTK-and-scikit-learn-13-320.jpg)
![Features Extraction in scikit-learn
from scikits.learn.features.text import *
text = u"J'ai mangxe9 du kangourou ce midi, c'xe9tait pas
trxeas bon."
print WordNGramAnalyzer(min_n=1, max_n=2).analyze(text)
[u'ai', u'mange', u'du', u'kangourou', u'ce', u'midi',
u'etait', u'pas', u'tres', u'bon', u'ai mange', u'mange du',
u'du kangourou', u'kangourou ce', u'ce midi', u'midi etait',
u'etait pas', u'pas tres', u'tres bon']
char_ngrams = CharNGramAnalyzer(min_n=3, max_n=6)
print char_ngrams[:5] + char_ngrams[5:]
[u"j'a", u"'ai", u'ai ', u'i m', u' ma', u's tres', u' tres
', u'tres b', u'res bo', u'es bon']](https://image.slidesharecdn.com/nltkscikit-learnpyconfr2010ogrisel-100828080703-phpapp02/85/Statistical-Learning-and-Text-Classification-with-NLTK-and-scikit-learn-16-320.jpg)
Statistical Learning and Text Classification with NLTK and scikit-learn
- 1. Statistical Learning and Text Classification with NLTK and scikit-learn Olivier Grisel http://twitter.com/ogrisel PyCON FR – 2010
- 2. Applications of Text Classification Task Predicted outcome Spam filtering Spam, Ham Language guessing English, Spanish, French, ... Sentiment Analysis for Product Positive, Neutral, Negative Reviews News Feed Topic Politics, Business, Technology, Categorization Sports, ... Pay-per-click optimal ads Will yield money, Won't placement Personal twitter filter Will interest me, Won't Malware detection in log files Normal, Malware
- 3. Supervised Learning Overview ● Convert training data to a set of vectors of features (input) & label (output) ● Build a model based on the statistical properties of features in the training set, e.g. ● Naïve Bayesian Classifier ● Logistic Regression / Maxent Classifier ● Support Vector Machines ● For each new text document to classify ● Extract features ● Asked model to predict the most likely outcome
- 4. Supervised Learning Summary features Training vectors Text Documents Machine Learning Algorithm Labels New features Text vector Predictive Expected Document Model Label
- 5. Typical features for text documents ● Tokenize document into list of words: uni-grams ['the', 'quick', 'brown', 'fox', 'jumps', 'over', 'the', 'lazy', 'dog'] ● Then chose one of: ● Binary occurrences of uni-grams: {'the': True, 'quick': True, ...} ● Frequencies of uni-grams: nb times word_i / nb words in document: {'the': 0.22, 'quick': 0.11, ...} ● TF-IDF of uni-grams (see next slides)
- 6. Better than freqs: TF-IDF ● Term Frequency ● Inverse Document Frequency => No real need for stop words any more, non informative words such as “the” are scaled done by IDF term
- 7. More advanced features ● Instead of uni-grams use ● bi-grams of words: “New York”, “very bad”, “not good” ● n-grams of chars: “the”, “ed ”, “ a ” (useful for language guessing) ● And the combine with: ● Binary occurrences ● Frequencies ● TF-IDF
- 8. NLTK ● Code: ASL 2.0 & Book: CC-BY-NC-ND ● Tokenizers, Stemmers, Parsers, Classifiers, Clusterers, Corpus Readers
- 10. Using a NLTK corpus >>> from nltk.corpus import movie_reviews >>> pos_ids = movie_reviews.fileids('pos') >>> neg_ids = movie_reviews.fileids('neg') >>> len(pos_ids), len(neg_ids) 1000, 1000 >>> print movie_reviews.raw(pos_ids[0])[:100] films adapted from comic books have had plenty of success , whether they're about superheroes ( batm >>> movie_reviews.words(pos_ids[0]) ['films', 'adapted', 'from', 'comic', 'books', 'have', ...]
- 11. Common data cleanup operations ● Switch to lower case: s.lower() ● Remove accentuated chars: import unicodedata s = ''.join(c for c in unicodedata.normalize('NFD', s) if unicodedata.category(c) != 'Mn') ● Extract only word tokens of at least 2 chars ● Using NLTK tokenizers & stemmers ● Using a simple regexp: re.compile(r"bww+b", re.U).findall(s)
- 12. Feature Extraction with NLTK ● Simple word binary occurrence features: def word_features(words): return dict((word, True) for word in words) ● Word Bigrams occurrence features: from nltk.collocations import BigramCollocationFinder from nltk.metrics import BigramAssocMeasures as BAM from itertools import chain def bigram_word_features(words, score_fn=BAM.chi_sq, n=200): bigram_finder = BigramCollocationFinder.from_words(words) bigrams = bigram_finder.nbest(score_fn, n) return dict((bg, True) for bg in chain(words, bigrams))
- 13. The NLTK - Naïve Bayes Classifier from nltk.classify import NaiveBayesClassifier mr = movie_reviews neg_examples = [(features(mr.words(i)), 'neg') for i in neg_ids] pos_examples = [(features(mr.words(i)), 'pos') for i in pos_ids] train_set = pos_examples + neg_examples classifier = NaiveBayesClassifier.train(train_set) # later on a previously unseed document predicted_label = classifier.classify(new_doc_features)
- 15. scikit-learn
- 16. Features Extraction in scikit-learn from scikits.learn.features.text import * text = u"J'ai mangxe9 du kangourou ce midi, c'xe9tait pas trxeas bon." print WordNGramAnalyzer(min_n=1, max_n=2).analyze(text) [u'ai', u'mange', u'du', u'kangourou', u'ce', u'midi', u'etait', u'pas', u'tres', u'bon', u'ai mange', u'mange du', u'du kangourou', u'kangourou ce', u'ce midi', u'midi etait', u'etait pas', u'pas tres', u'tres bon'] char_ngrams = CharNGramAnalyzer(min_n=3, max_n=6) print char_ngrams[:5] + char_ngrams[5:] [u"j'a", u"'ai", u'ai ', u'i m', u' ma', u's tres', u' tres ', u'tres b', u'res bo', u'es bon']
- 17. TF-IDF features & SVMs from scikits.learn.features.text import * from scikits.learn.sparse.svm import LinearSVC hv = SparseHashingVectorizer(dim=1000000, analyzer=) hv.vectorize(list_of_documents) features = hv.get_tfidf() clf = SparseLinearSVC(C=10, dual=false) clf.fit(features, labels) # later with the same clf instance predicted_labels = clf.predict(features_of_new_docs)
- 18. Typical performance results ● Naïve Bayesian Classifier with unigram occurrences on movie reviews: ~ 70% ● Same as above selecting the top 10000 most informative features only: ~ 93% ● TF-IDF unigram features + Linear SVC on 20 newsgroups ~93% (with 20 target categories) ● Language guessing with character ngram frequencies features + Linear SVC: almost perfect if document is long enough
- 19. Confusion Matrix (20 newsgroups) 00 alt.atheism 01 comp.graphics 02 comp.os.ms-windows.misc 03 comp.sys.ibm.pc.hardware 04 comp.sys.mac.hardware 05 comp.windows.x 06 misc.forsale 07 rec.autos 08 rec.motorcycles 09 rec.sport.baseball 10 rec.sport.hockey 11 sci.crypt 12 sci.electronics 13 sci.med 14 sci.space 15 soc.religion.christian 16 talk.politics.guns 17 talk.politics.mideast 18 talk.politics.misc 19 talk.religion.misc
- 20. Handling many possible outcomes ● Example: possible outcomes are all the categories of Wikipedia (565,108) ● Document Categorization becomes Information Retrieval ● Instead of building one linear model for each outcome build a fulltext index and perform TF- IDF similarity queries ● Smart way to find the top 10 search keywords ● Use Apache Lucene / Solr MoreLikeThisQuery
- 21. NLTK – Online demos
- 22. NLTK – REST APIs % curl -d "text=Inception is the best movie ever" http://text-processing.com/api/sentiment/ { "probability": { "neg": 0.36647424288117808, "pos": 0.63352575711882186 }, "label": "pos" }
- 24. Some pointers ● http://www.nltk.org (Code & Doc & PDF Book) ● http://scikit-learn.sf.net (Doc & Examples) http://github.com/scikit-learn (Code) ● http://www.slideshare.net/ogrisel (These slides) ● http://streamhacker.com/(Blog on NLTK & APIs) ● http://github.com/hmason/tc (Twitter classifier – work in progress)