Text Analytics

[email protected]
BX9T5ZHNQF

Machine Learning

Proprietary content. ©Great Learning. All Rights Reserved.

This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
Sharing or publishing the contents in part or full is liable for legal action.
 Agenda

• Introduction to Text Analytics

• Text Analytics and Applications

• Unstructured Vs Structured Data, Cleaning

• Bag of words, Word Frequencies

[email protected]
BX9T5ZHNQF

• Hierarchical Clustering

• Sentiment Analysis

• Word Embeddings

• Ensemble Methods

Proprietary content. ©Great Learning. All Rights Reserved.

This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
2
Sharing or publishing the contents in part or full is liable for legal action.
 Text Analytics

• The process of drawing meaning out of written

communication.

• to understand online reviews, tweets, call center agent

notes, survey results, and other types of written feedback
that capture insight into your customers.
[email protected]
BX9T5ZHNQF • Spam detection

• translation

• search and crawl

• sentimental analysis

• entity modeling to support fact based decision making

• text summarization
Proprietary content. ©Great Learning. All Rights Reserved.
This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
3
Sharing or publishing the contents in part or full is liable for legal action.
 Text: Unstructured Data

• Structured: Data is organized into pre-defined structure like

a table of database - with rows and columns.
[email protected]
BX9T5ZHNQF
• UnStructured Data: Data does not have a pre-defined
structure. Think of a collection of emails, a bunch of satellite
images or the entire text of speeches from the british
parliament since 1803.

Proprietary content. ©Great Learning. All Rights Reserved.

This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
4
Sharing or publishing the contents in part or full is liable for legal action.
 Modeling/representing text

• Bag of words - Documents simply represented by the words

in the document and their frequencies. Disregards grammar
and word order

•
[email protected]
BX9T5ZHNQF Bayesian SPAM filter

• Semantic - mapping natural language rules to get a formal

representation of the meaning of the text

• Name entity identification

Proprietary content. ©Great Learning. All Rights Reserved.

This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
5
Sharing or publishing the contents in part or full is liable for legal action.
 Bag of words
• Corpus:

• A: John likes to play soccer

• B: John is reading a book

John likes soccer play book reading a is to

[email protected]
BX9T5ZHNQF

A 1 1 1 1 1
B 1 1 1 1 1

Proprietary content. ©Great Learning. All Rights Reserved.

This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
6
Sharing or publishing the contents in part or full is liable for legal action.
 n-gram model

• The Bag-of-words model is an orderless document representation. Only

the counts of words matter.

• We could do this also by choosing consecutive pairs (2-gram) and

representing each pair

• A: John likes to play soccer

[email protected] • B: John is reading a book

BX9T5ZHNQF

• 2-gram (bigram):

John likes likes to play soccer to play John is is reading reading a a book

A 1 1 1 1
B 1 1 1 1

Proprietary content. ©Great Learning. All Rights Reserved.

This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
7
Sharing or publishing the contents in part or full is liable for legal action.
 Cleaning text

• Stop words: Common words that are not useful in providing

value or context. Eg: ‘the’, ‘an’, ‘in’ etc.

• Stemming: Returning words to their original stem. Eg:

‘Chopping’, ‘Chopped’ are all replaced with ‘Chop’
[email protected]
BX9T5ZHNQF

• Lower case conversion

• Remove punctuations

• Strip extra white spaces

• Remove numbers

Proprietary content. ©Great Learning. All Rights Reserved.

This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
8
Sharing or publishing the contents in part or full is liable for legal action.
 Example

[email protected]
BX9T5ZHNQF

Proprietary content. ©Great Learning. All Rights Reserved.

This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
9
Sharing or publishing the contents in part or full is liable for legal action.
 Example

[email protected]
BX9T5ZHNQF

Proprietary content. ©Great Learning. All Rights Reserved.

This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
10
Sharing or publishing the contents in part or full is liable for legal action.
 Term-Document Matrix (TDM)
Doc 1 Doc 2 … Doc N
Term 1

Term 2
…

…
[email protected]
BX9T5ZHNQF
Term M

Document-Term Matrix (DTM)

Term 1 Term 2 … Term M
Doc 1
Doc 2
…
Doc N

Proprietary content. ©Great Learning. All Rights Reserved.

This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
11
Sharing or publishing the contents in part or full is liable for legal action.
 • Each document is represented by a vector in the term document
matrix

• This lends itself to a number of ML techniques

• For example, these vectors (documents) can be clustered to

identify similar documents

[email protected]
BX9T5ZHNQF

Proprietary content. ©Great Learning. All Rights Reserved.

This fileUnauthorized use or
is meant for personal distribution
use prohibited. only.
by [email protected]
12
Sharing or publishing the contents in part or full is liable for legal action.