Introduction to NLP in Python @ London Python Meetup Sept 2016

A very gentle introduction to the field of Natural Language Processing (NLP) using Python tools.

The tutorial at https://github.com/bonzanini/nlp-tutorial shows more detailed examples and could be used as a companion for these slides.

Transcript

1. (a very gentle) Introduction to! Natural Language Processing! in Python Marco Bonzanini! London Python Meet-up 
 Sept 2016
2. Nice to Meet You
3. Objectives for Today • Getting one Python person to try some NLP • Getting one NLP person to try some Python
4. http://speakerdeck.com/marcobonzanini http://github.com/bonzanini/nlp-tutorial
5. Natural Language
6. Language Instrument of communication
7. Natural Not planned, not artificial
8. SELECT name, address
 FROM businesses
 WHERE business_type = ‘pub’
 AND postcode_area = ‘EC2M’ vs Where is the nearest pub?
9. Natural = Easy ?
10. Language is huge
11. • OED: 171,000 words in current use (47k obsolete) • Average person: 10,000-40,000 words
 (according to: The Guardian 1986, BBC 2009, and several random people on the Web) • Average person? (Lies, damn lies and statistics) • Active vs Passive vocabulary
12. Language is confusing (sometimes)
13. That that is is that that is not is not is that it it is (That’s proper English)
14. That that is, is. That that is not, is not. Is that it? It is. More fun at:
 https://en.wikipedia.org/wiki/List_of_linguistic_example_sentences Pics:
 https://en.wikipedia.org/wiki/Socrates and https://en.wikipedia.org/wiki/Parmenides
15. Language is ambiguous
16. Word ambiguity
17. “They ate pizza with anchovies” Syntactic ambiguity
18. Common sense
 is implied (but computers don’t really have it)
19. Natural Language Processing Computational
 Linguistics Computer
 Science NLP NLP vs Text Mining vs Text Analytics
20. NLP Goals Text Data Useful Information Actionable Insights
21. NLP Applications • Text Classification • Text Clustering • Text Summarisation • Machine Translation
 • Semantic Search • Sentiment Analysis • Question Answering • Information Extraction
24. I was told there would be Python
26. NLP Pipeline • pip install nltk • Sentence Boundary Detection • Word Tokenisation • Word Normalisation • Stop-word removal • Bigrams / trigrams / n-grams
27. Sentence Boundary Detection
28. How about str.split(‘.’) ??? How about Mr., Dr. or U.S.A.?
29. from nltk.tokenize
 import sent_tokenize
30. Word Tokenisation
31. How about str.split(‘ ’) ??? How about punctuation?
32. from nltk.tokenize
 import word_tokenize
33. >>> s = "Talking about #NLProc in #Python at @python_london meetup" >>> word_tokenize(s) ['Talking', 'about', '#', 'NLProc', 'in', '#', 'Python', 'at', '@', 'python_london', 'meetup']
34. from nltk.tokenize
 import TweetTokenizer
35. Text Normalisation
36. >>> "python" == "Python" False >>> "python" == "Python".lower() True
37. Stemming • Map a token into its stem • Fish, Fishes, Fishing ‛ Fish
38. from nltk.stem
 import PorterStemmer
39. from nltk.stem
 import SnowballStemmer
40. Lemmatisation • Map a token into its lemma • Go, goes, going, went ‛ Go
41. from nltk.stem
 import WordNetLemmatizer
42. Stop-word Removal
43. from nltk.corpus import stopwords
 
 stopwords.words(‘english’)
44. >>> stop_list = [ … ] # custom
 >>> s = "a piece of butter" >>> [tok for tok in word_tokenize(s) 
 if tok not in stop_list] ['piece', 'butter']
45. n-grams
46. from nltk import bigrams from nltk import trigrams from nltk import ngrams
47. Good for capturing phrases:
 “bad movie”, “good olive oil”, … ! How about stop-words?
48. … Now what?
49. Exploring Text Data
50. from collections import Counter ! frequencies = Counter(all_tokens)
51. Visualising Text Data
52. pip install wordcloud @MarcoBonzanini
53. Can we do something smarter?
54. –J.R. Firth 1957 “You shall know a word 
 by the company it keeps.”
55. • From Bag-of-Words to Word Embeddings
 (e.g. word2vec) • Similar context = close vectors • Semantic relationships: vector arithmetic! • pip install gensim
56. from gensim.models import Word2Vec model = Word2Vec(sentences) ! model.most_similar(positive=['king', 'woman'], negative=['man']) ! [('queen', 0.50882536), ...] Tutorial: https://rare-technologies.com/word2vec-tutorial/
57. More NLP Libraries • spaCy — “industrial-strength NLP”
 Designed for speed and accuracy • scikit-learn — Machine Learning
 Good support for text
 (e.g. TfidfVectorizer)
58. Summary • 80/20 rule: preprocessing is 80%? • Counting words vs Neural Networks
 (in less than 1h!) • Rich Python ecosystem
59. Thank You https://github.com/bonzanini/nlp-tutorial
 @MarcoBonzanini