Python is a high-level, object-oriented development tool. Here is a quick, hands-on tutorial on how to use the text analytics function.

Python enables four kinds of analytics:

Text matching
Text classification
Topic modelling
Summarization

Let’s begin by understanding some of the NLP features of Python, how it is set up and how to read the file used for:

Basics of NLP

Reading a text file

Tokenisation
Stemming & Lemmatization
Dispersion Plots
Word frequency

Setting up NLTK

import nltk
from nltk.book import*
nltk.download()

Reading a text file

import os

os.chdir(‘F:/Work/Philip Adams/Course Content/Data’)

f=open(‘Genesis.txt’).read().decode(‘utf8’)

Our programs will often need to deal with different languages, and different character sets. The concept of “plain text” is a fiction.
ASCII
Unicode is used to process non-ASCII charcters
Unicode supports over a million characters. Each character is assigned a number, called a code point.
Translation into unicode is called decoding.

Let’s move a step deeper and understand the four basics of NLP in detail:

Tokenisation

Breaking up the text into words and punctuations
Each distinct word and punctuation

line=’Because he was so small, Stuart was often hard to find around the house. – E.B. White’

tokens=Because, he, was, so, small, Stuart, was, often, hard, to, find, around, the, house,’,’, E, B, White, ‘.‘, ‘-’

tokens=nltk.word_tokenize(f)

len(tokens)

tokens[:10]

Dispersion plots

Shows the position of a word across the document/text corpora

text.dispersion_plot([‘God’,’life’,’earth’,’empty’])

Converting tokens to NLTK text

To apply NLTK processes, tokens need to be converted to NLTK test

text=nltk.Text(tokens)

Collocations

Words frequently occurring together

text.collocations()

one hundred; years old; Paddan Aram; young lady; seven years; little ones; found favor; burnt offering; living creature; every animal; four hundred; every living; thirty years; Yahweh God; n’t know; nine hundred; savory food; taken away; God said; ‘You shall

Word at a particular position

text[225]

Position of a particular word

text.index(‘life’)

Concordance

Finding the context of a particular word in the document

text.concordance(‘life’)

Total number of words in a document

len(tokens)

Total number of distinct words in a document

len(set(tokens))

Diversity of words or percent of distinct words in the document

len(set(tokens))/len(tokens)

Percentage of text occupied by one word

100*text.count(‘life’)/len(tokens)

Frequency distribution of words in a document

from nltk.probability import FreqDist

fdist=FreqDist(tokens)

Function to return the frequency of a particular

def freq_calc(word,tokens):

from nltk.probability import FreqDist

fdist=FreqDist(tokens)

return fdist[word]

Most frequent words

fdist.most_common(50)

Other frequency distribution functions

fdist.max(), fdist.plot(), fdist.tabulate()

Counting the word length for all the words

([len(w) for w in text])

Frequency distribution of word lengths

fdistn=FreqDist([len(w) for w in text])

Fdistn

Returning words longer than 10 letters

[w for w in tokens if len(w)>10]

Stop words

Words which are commonly used as end points of sentences and carry less contextual meaning

from nltk.corpus import stopwords

stop_words=set(stopwords.words(‘english’))

Filtering stop words

filtered=[w for w in tokens if not w in stop_words ]

filtered

Stemming

Keeping only the root/stem of the word and reducing all the derivatives to their root words

For e.g. ‘walker’, ‘walked’, ‘walking’ would return only the root word ‘walk’

from nltk.stem import PorterStemmer

ps=PorterStemmer()

for w in tokens:

print ps.stem(w)

Lemmatization

Similar to stemming but more robust as it can distinguish between words based on Parts of Speech and context

For e.g. ‘walker’, ‘walked’, ‘walking’ would return only the root word ‘walk’

from nltk.stem import WordNetLemmatizer

lm=WordNetLemmatizer()

for w in tokens:

print lm.lemmatize(w)

lm.lemmatize(‘wolves’)

Result: u’wolf

lm.lemmatize(‘are’,pos=’v’)

Result: u’be

POS (Part of Speech) Tagging

Tagging each token/word as a part of speech

nltk.pos_tag(tokens)

Regular

Regular Expressions

Expressions to denote patterns which match words/phrase/sentences in a text/document

re.search

matchObject = re.search(pattern, input_str, flags=0)

Stops after first match

import re

regex=r”(\d+)”

match=re.search(regex,”91,’Alexander’,’Abc123′”)

match.group(0)

Result: 91

re.findall

matchObject = re.findall(pattern, input_str, flags=0)

Stops after first match

import re

regex=r”(\d+)”

match=re.findall(regex,”91,’Alexander’,’Abc123′”)

match.group(0)

Result: 91, 123

re.sub

replacedString = re.sub(pattern, replacement_pattern, input_str, count, flags=0)

import re

regex=r”(\d+)”

re.sub(regex,”,”91,’Alexander’,’Abc123′”)

Result: “,’Alexander’,’Abc'”

Text Cleaning

Removing a list of words from the text

noise_list = [“is”, “a”, “this”, “…”]

def remove_noise(input_text):

words = input_text.split()

noise_free_words = [word for word in words if word not in noise_list]

noise_free_text = ” “.join(noise_free_words)

return noise_free_text

remove_noise(“this is a sample text”)

Replacing a set of words with standard terms

input_text=”This rt is actually an awsm dm which I luv”

words = input_text.split()

new_words = []

for word in words:

if word.lower() in lookup_dict.keys():

word = lookup_dict[word.lower()]

new_words.append(word)

new_words

new_text=” “.join(new_words)

new_text

N-Grams

N-grams is a sequence of words n items long.

def generate_ngrams(text, n):

words = text.split()

output = []

for i in range(len(words)-n+1):

output.append(words[i:i+n])

return output

generate_ngrams(“Virat may break all the records of Sachin”,3)

TF-IDF

Term Frequency – Inverse Document Frequency

convert the text documents into vector models on the basis of occurrence of words in the documents

Term	Definition
Term Frequency (TF)	Frequency of a term in document D
Inverse Document Frequency (IDF)	logarithm of ratio of total documents available in the corpus and number of documents containing the term T
TF-IDF	TF IDF formula gives the relative importance of a term in a corpus (list of documents)

TF-IDF

from sklearn.feature_extraction.text import TfidfVectorizer

obj = TfidfVectorizer()

corpus = [‘Ram ate a mango.’, ‘mango is my favorite fruit.’, ‘Sachin is my favorite’]

X = obj.fit_transform(corpus)

print X

Other tasks

Text Classification

Naïve Bayes Classifier
SVM

Text Matching

Levenheisten distance – minimum number of edits needed to transform one string into the other
Phonetic matching – A Phonetic matching algorithm takes a keyword as input (person’s name, location name etc) and produces a character string that identifies a set of words that are (roughly) phonetically similar

Different ways of reading a text file

f=open(‘genesis.txt’)

words= f.read().split()

f.close()

f=open(‘genesis.txt’)

words=[]

for line in f:

print line.split()

f.close