Skip to main content

Natural Language Processing with Python NLTK part 5 - Chunking and Chinking

Natural Language Processing


Using regular expression modifiers we can chunk out the PoS tagged words from the earlier example. The chunking is done with regular expressions defining a chunk rule. The Chinking defines what we need to exclude from the selection.

Here are list of modifiers for Python:

  • {1,3} = for digits, u expect 1-3 counts of digits, or "places"
  • + = match 1 or more
  • ? = match 0 or 1 repetitions.
  • * = match 0 or MORE repetitions
  • $ = matches at the end of string
  • ^ = matches start of a string
  • | = matches either/or. Example x|y = will match either x or y
  • [] = range, or "variance"
  • {x} = expect to see this amount of the preceding code.
  • {x,y} = expect to see this x-y amounts of the preceding code
source: https://pythonprogramming.net/regular-expressions-regex-tutorial-python-3/

Chunking


import nltk
from nltk.tokenize import word_tokenize

# POS tagging
sent = "This will be chunked. This is for Test. World is awesome. Hello world."

print(nltk.pos_tag(word_tokenize(sent)))

# creating a regular expression for chunking verbs and nouns
chunkRule = r"""chunk: {<NN.?>*<NNS.?>*<NNP.?>*<NNPS.?>*<VB.?>*<VBD.?>*<VBG.?>*<VBN.?>*<VBP.?>*<VBZ.?>*}"""

My_parser = nltk.RegexpParser(chunkRule)
chunked = My_parser.parse(nltk.pos_tag(word_tokenize(sent)))

print(chunked)


This will give the output as follows:


If you have matplotlib installed you can use a simple code to graphically view the tree which is easier to understand.

My_parser = nltk.RegexpParser(chunkRule)
chunked = My_parser.parse(nltk.pos_tag(word_tokenize(sent)))

# draw the tree
chunked.draw()
print(chunked)

This will draw the tree as follows.





Chinking

chinking is the process of excluding. In this example we will be chunking all the tags and excluding nouns.

import nltk
from nltk.tokenize import word_tokenize

sent = "This will be the day that I will chink all the nouns. Everything will be there.Except nouns"

print(nltk.pos_tag(word_tokenize(sent)))

# first chunk everything and chink only nouns
chunkRule = r"""Chunk: {<.*>+}
                        }<NN.?|NNS|NNP|NNPS>+{"""

MyParser = nltk.RegexpParser(chunkRule)
chunked = MyParser.parse(nltk.pos_tag(word_tokenize(sent)))

chunked.draw()
print(chunked)

Outputs:







Popular posts from this blog

Natural Language Processing with Python NLTK part 1 - Tokenizer

Natural Language Processing Starting with the NLP articles first we will try the  tokenizer  in the NLTK package. Tokenizer breaks a paragraph into the relevant sub strings or sentences based on the tokenizer you used. In this I will use the Sent tokenizer, word_tokenizer and TweetTokenizer which has its specific work to do. import nltk from nltk.tokenize import sent_tokenize, word_tokenize, TweetTokenizer para = "Hello there this is the blog about NLP. In this blog I have made some posts. " \ "I can come up with new content." tweet = "#Fun night. :) Feeling crazy #TGIF" # tokenizing the paragraph into sentences and words sent = sent_tokenize(para) word = word_tokenize(para) # printing the output print ( "this paragraph has " + str(len(sent)) + " sentences and " + str(len(word)) + " words" ) # print each sentence k = 1 for i in sent: print ( "sentence ...

Natural Language Processing with Python NLTK part 6 - Named Entity Recognition

Natural Language Processing - NER Named entities are specific reference to something. As a part of recognizing text NLTK has allowed us to used the named entity recognition and recognize certain types of entities. Those types are as follows NE Type Examples ORGANIZATION Georgia-Pacific Corp. ,  WHO PERSON Eddy Bonte ,  President Obama LOCATION Murray River ,  Mount Everest DATE June ,  2008-06-29 TIME two fifty a m ,  1:30 p.m. MONEY 175 million Canadian Dollars ,  GBP 10.40 PERCENT twenty pct ,  18.75 % FACILITY Washington Monument ,  Stonehenge GPE South East Asia ,  Midlothian Source:  http://www.nltk.org/book/ch07.html Simple example on NER: import nltk from nltk.tokenize import word_tokenize, sent_tokenize para = " America is a country. John is a name. " sent = sent_tokenize(para) for s in sent: word = word_tokenize(s) tag = nltk . pos_tag(word) n...