Business Intelligence and Analytics:

Systems for Decision Support

Global Edition
(10th Edition)

Chapter 7:
Text Analytics, Text Mining,
and Sentiment Analysis
 Learning Objectives
 Describe text mining and understand the need
for text mining
 Differentiate between text mining, Web mining,
and data mining
 Understand the different application areas for
text mining
 Know the process of carrying out a text mining
project
 Understand the different methods to introduce
structure to text-based data
(Continued…)
-2 © Pearson Education Limited 2014
 Learning Objectives
 Describe sentiment analysis
 Develop familiarity with popular applications of
sentiment analysis
 Learn the common methods for sentiment
analysis
 Become familiar with speech analytics as it
relates to sentiment analysis

-3 © Pearson Education Limited 2014

 Opening Vignette…
Machine Versus Men on Jeopardy!:
The Story of Watson

 Situation
 Problem Watch it on YouTube!
 Solution https://www.youtube.com/watch?v=YLR1byL0U8M
 Results
 Answer & discuss the case questions...
-4 © Pearson Education Limited 2014
 Questions for
the Opening Vignette
1. What is Watson? What is special about it?
2. What technologies were used in building
Watson (both hardware and software)?
3. What are the innovative characteristics of
DeepQA architecture that made Watson
superior?
4. Why did IBM spend all that time and money
to build Watson? Where is the ROI?

-5 © Pearson Education Limited 2014

 A High-Level Depiction of IBM
Watson’s DeepQA Architecture

Answer Evidence
sources sources

Candidate Support Deep

Primary
answer evidence evidence
Question search
generation retrieval scoring

Trained
models 3
4
5

2
1

Question Query Hypothesis Soft Hypothesis and Final merging

Synthesis
analysis decomposition generation filtering evidence scoring and ranking

Hypothesis Soft Hypothesis and

generation filtering evidence scoring Answer and
confidence

... ... ...

-6 © Pearson Education Limited 2014

 Text Mining Concepts
 85-90 percent of all corporate data is in some
kind of unstructured form (e.g., text)
 Unstructured corporate data is doubling in size
every 18 months
 Tapping into these information sources is not an
option, but a need to stay competitive
 Answer: text mining
 A semi-automated process of extracting knowledge
from unstructured data sources
 a.k.a. text data mining or knowledge discovery in
textual databases
-7 © Pearson Education Limited 2014
 Text Analytics and Text Mining
TEXT ANALYTICS

Text Mining
Information
Web Mining
Retrieval

Information
Data Mining
Extraction

Natural Language Processing Linguistic Machine Learning

Computer Science Statistics Management Science Artificial Intelligence

-8 © Pearson Education Limited 2014

 Data Mining versus Text Mining
 Both seek for novel and useful patterns
 Both are semi-automated processes
 Difference is the nature of the data:
 Structured versus unstructured data
 Structured data: in databases
 Unstructured data: Word documents, PDF
files, text excerpts, XML files, and so on
 Text mining – first, impose structure to
the data, then mine the structured data.
-9 © Pearson Education Limited 2014
 Text Mining Concepts
 Benefits of text mining are obvious, especially in
text-rich data environments
 e.g., law (court orders), academic research (research
articles), finance (quarterly reports), medicine
(discharge summaries), biology (molecular
interactions), technology (patent files), marketing
(customer comments), etc.
 Electronic communication records (e.g., Email)
 Spam filtering
 Email prioritization and categorization
 Automatic response generation
-10 © Pearson Education Limited 2014
 Text Mining Application Area
 Information extraction
 Topic tracking
 Summarization
 Categorization
 Clustering
 Concept linking
 Question answering

-11 © Pearson Education Limited 2014

 Text Mining Terminology
 Unstructured or semi-structured data
 Corpus (and corpora)
 Terms
 Concepts
 Stemming
 Stop words (and include words)
 Synonyms (and polysemes)
 Tokenizing
-12 © Pearson Education Limited 2014
 Text Mining Terminology
 Term dictionary
 Word frequency
 Part-of-speech tagging
 Morphology
 Term-by-document matrix
 Occurrence matrix
 Singular value decomposition
 Latent semantic indexing

-13 © Pearson Education Limited 2014

 Application Case 7.1
Text Mining for Patent Analysis
 What is a patent?
 “exclusive rights granted by a country to an
inventor for a limited period of time in
exchange for a disclosure of an invention”
 How do we do patent analysis (PA)?
 Why do we need to do PA?
 What are the benefits?
 What are the challenges?
 How does text mining help in PA?
-14 © Pearson Education Limited 2014
 Natural Language Processing
(NLP)
 Structuring a collection of text
 Old approach: bag-of-words
 New approach: natural language processing
 NLP is …
 a very important concept in text mining
 a subfield of artificial intelligence and computational
linguistics
 the studies of "understanding" the natural human
language
 Syntax versus semantics-based text mining

-15 © Pearson Education Limited 2014

 Natural Language Processing
(NLP)
 What is “Understanding” ?
 Human understands, what about
computers?
 Natural language is vague, context driven
 True understanding requires extensive
knowledge of a topic

 Can/will computers ever understand natural

language the same/accurate way we do?
-16 © Pearson Education Limited 2014
 Natural Language Processing
(NLP)
 Challenges in NLP
 Part-of-speech tagging
 Text segmentation
 Word sense disambiguation
 Syntax ambiguity
 Imperfect or irregular input
 Speech acts

 Dream of AI community
 to have algorithms that are capable of automatically
reading and obtaining knowledge from text
-17 © Pearson Education Limited 2014
 Natural Language Processing
(NLP)
 WordNet
 A laboriously hand-coded database of English words,
their definitions, sets of synonyms, and various
semantic relations between synonym sets.
 A major resource for NLP.
 Need automation to be completed.
 Sentiment Analysis
 A technique used to detect favorable and
unfavorable opinions toward specific products and
services
 SentiWordNet
-18 © Pearson Education Limited 2014
 Application Case 7.2
Text Mining Improves Hong Kong
Government’s Ability to Anticipate and
Address Public Complaints

Questions for Discussion

1. How did the Hong Kong government use text
mining to better serve its constituents?
2. What were the challenges, the proposed
solution, and the obtained results?
-19 © Pearson Education Limited 2014
 NLP Task Categories
 Information retrieval, information extraction
 Named-entity recognition
 Question answering
 Automatic summarization
 Natural language generation & understanding
 Machine translation
 Foreign language reading & writing
 Speech recognition
 Text proofing, optical character recognition
-20 © Pearson Education Limited 2014
 Text Mining Applications
 Marketing applications
 Enables better CRM
 Security applications
 ECHELON, OASIS
 Deception detection (…)
 Medicine and biology
 Literature-based gene identification (…)
 Academic applications
 Research stream analysis
-21 © Pearson Education Limited 2014
 Application Case 7.3
Mining for Lies!
 Deception detection
 A difficult problem
 If detection is limited to only text, then the
problem is even more difficult
 The study
 analyzed text-based testimonies of persons of
interest at military bases
 used only text-based features (cues)
-22 © Pearson Education Limited 2014
 Application Case 7.3
Mining for Lies
Statements
Transcribed for
Processing

Statements Labeled as
Cues Extracted &
Truthful or Deceptive
Selected
By Law Enforcement

Classification Models Text Processing

Trained and Tested on Software Identified
Quantified Cues Cues in Statements

Text Processing
Software Generated
Quantified Cues

-23 © Pearson Education Limited 2014

 Application Case 7.3
Mining for Lies
Category Example Cues
Quantity Verb count, noun-phrase count, ...
Complexity Avg. no of clauses, sentence length, …
Uncertainty Modifiers, modal verbs, ...
Nonimmediacy Passive voice, objectification, ...
Expressivity Emotiveness
Diversity Lexical diversity, redundancy, ...
Informality Typographical error ratio
Specificity Spatiotemporal, perceptual information …
Affect Positive affect, negative affect, etc.
-24 © Pearson Education Limited 2014
 Application Case 7.3
Mining for Lies
 371 usable statements are generated
 31 features are used
 Different feature selection methods used
 10-fold cross validation is used
 Results (overall % accuracy)
 Logistic regression 67.28
 Decision trees 71.60
 Neural networks 73.46

-25 © Pearson Education Limited 2014

 Text Mining Applications
(Gene/Protein Interaction Identification)
Protein

596 12043 24224 281020 42722 397276

Gene/

D007962

D 016923
Ontology

D 001773

D019254 D044465 D001769 D002477 D003643 D016158

...expression of Bcl-2 is correlated with insufficient white blood cell death and activation of p53.
Word

185 8 51112 9 23017 27 5874 2791 8952 1623 5632 17 8252 8 2523
POS

NN IN NN IN VBZ IN JJ JJ NN NN NN CC NN IN NN
Shallow
Parse

NP PP NP NP PP NP NP PP NP

-26 © Pearson Education Limited 2014

 Application Case 7.4
Text mining and Sentiment Analysis Help
Improve Customer Service Performance

Questions for Discussion

1. How did the financial services firm use text
mining and text analytics to improve its
customer service performance?
2. What were the challenges, the proposed
solution, and the obtained results?
-27 © Pearson Education Limited 2014
 Text Mining Process
Context diagram for Software/hardware limitations
Privacy issues
the text mining Linguistic limitations
process
Unstructured data (text) Extract Context-specific knowledge
knowledge
from available
Structured data (databases) data sources
A0

Domain expertise
Tools and techniques

-28 © Pearson Education Limited 2014

 Text Mining Process
Task 1 Task 2 Task 3

Establish the Corpus: Create the Term- Extract Knowledge:

Collect & Organize the Document Matrix: Discover Novel
Domain Specific Introduce Structure Patterns from the
Unstructured Data to the Corpus T-D Matrix

Feedback Feedback

The inputs to the process The output of the Task 1 is a The output of the Task 2 is a The output of Task 3 is a
includes a variety of relevant collection of documents in flat file called term-document number of problem specific
unstructured (and semi- some digitized format for matrix where the cells are classification, association,
structured) data sources such computer processing populated with the term clustering models and
as text, XML, HTML, etc. frequencies visualizations

The three-step text mining process

-29 © Pearson Education Limited 2014

 Text Mining Process
 Step 1: Establish the corpus
 Collect all relevant unstructured data
(e.g., textual documents, XML files, emails,
Web pages, short notes, voice recordings…)
 Digitize, standardize the collection
(e.g., all in ASCII text files)
 Place the collection in a common place
(e.g., in a flat file, or in a directory as
separate files)

-30 © Pearson Education Limited 2014

 Text Mining Process
 Step 2: Create the Term-by-Document
Matrix (TDM)
t g
m en e rin
Terms k e e
ris ag gin t
en t an
e e n
m en
tm t m r p
es ojec ft wa v elo P
Documents in v pr s o de SA ...
Document 1 1 1

Document 2 1

Document 3 3 1

Document 4 1

Document 5 2 1

Document 6 1 1
...

-31 © Pearson Education Limited 2014

 Text Mining Process
 Step 2: Create the Term-by-Document
Matrix (TDM)
 Should all terms be included?
 Stop words, include words
 Synonyms, homonyms
 Stemming
 What is the best representation of the indices
(values in cells)?
 Row counts; binary frequencies; log frequencies;
 Inverse document frequency
-32 © Pearson Education Limited 2014
 Text Mining Process
 Step 2: Create the Term–by–Document
Matrix (TDM)
 TDM is a sparse matrix. How can we reduce
the dimensionality of the TDM?
 Manual - a domain expert goes through it
 Eliminate terms with very few occurrences in very
few documents (?)
 Transform the matrix using singular value
decomposition (SVD)
 SVD is similar to principle component analysis

-33 © Pearson Education Limited 2014

 Text Mining Process
 Step 3: Extract patterns/knowledge
 Classification (text categorization)
 Clustering (natural groupings of text)
 Improve search recall
 Improve search precision
 Scatter/gather
 Query-specific clustering
 Association
 Trend Analysis (…)

-34 © Pearson Education Limited 2014

 Application Case 7.5
(Research Literature Survey with Text Mining)

 Mining the published IS literature

 MIS Quarterly (MISQ)
 Journal of MIS (JMIS)
 Information Systems Research (ISR)

 Covers 12-year period (1994-2005)

 901 papers are included in the study
 Only the paper abstracts are used
 9 clusters are generated for further analysis
-35 © Pearson Education Limited 2014
 Application Case 7.5
(Research Literature Survey with Text Mining)
Journal Year Author(s) Title Vol/No Pages Keywords Abstract
MISQ 2005 A. Malhotra, Absorptive capacity 29/1 145-187 knowledge management The need for continual value
S. Gosain and configurations in supply chain innovation is driving supply
O. A. El Sawy supply chains: absorptive capacity chains to evolve from a pure
Gearing for partner- interorganizational transactional focus to
enabled market information systems leveraging interorganizational
knowledge creation configuration approaches partner ships for sharing
ISR 1999 D. Robey and Accounting for the 2-Oct 167-185 organizational Although much contemporary
M. C. Boudreau contradictory transformation thought considers advanced
organizational impacts of technology information technologies as
consequences of organization theory either determinants or enablers
information research methodology of radical organizational
technology: intraorganizational power change, empirical studies have
Theoretical directions electronic communication revealed inconsistent findings to
and methodological mis implementation support the deterministic logic
implications culture implicit in such arguments. This
systems paper reviews the contradictory
JMIS 2001 R. Aron and Achieving the optimal 18/2 65-88 information products When producers of goods (or
E. K. Clemons balance between internet advertising services) are confronted by a
investment in quality product positioning situation in which their offerings
and investment in self- signaling no longer perfectly match
promotion for signaling games consumer preferences, they
information products must determine the extent to
which the advertised features of

… … … … … … … …

-36 © Pearson Education Limited 2014

-37
No of Articles

1
1
2
2
3
3
1
1
2
2
3
3
1
1
2
2
3
3

0
5
0
5
0
5
0
5
0
5
0
5
0
5
0
5
0
5

0
5
0
5
0
5
1994 1994 1994
1995 1995 1995
1996 1996 1996
1997 1997 1997
1998 1998 1998
1999 1999 1999
2000 2000 2000
2001 2001 2001

C LU STER : 7
C LU STER : 4
C LU STER : 1
2002 2002 2002
2003 2003 2003
2004 2004 2004
2005 2005 2005

1994 1994 1994

1995 1995 1995
1996 1996 1996
1997 1997 1997
1998 1998 1998
1999 1999 1999

Y EAR
2000 2000 2000
2001 2001 2001

C LU STER : 8
C LU STER : 5
C LU STER : 2

2002 2002 2002

2003 2003 2003
2004 2004 2004
2005 2005 2005
Application Case 7.5

© Pearson Education Limited 2014

1994 1994 1994
1995 1995 1995
1996 1996 1996
1997 1997 1997
1998 1998 1998
1999 1999 1999
2000 2000 2000
2001 2001 2001
C LU S TER : 9
C LU S TER : 6
C LU S TER : 3

2002 2002 2002

2003 2003 2003
2004 2004 2004
2005 2005 2005
(Research Literature Survey with Text Mining)
 Application Case 7.5
(Research Literature Survey with Text Mining)
100
90
80
70
60
50
40
30
20
10
0
IS R J M IS M IS Q IS R J M IS M IS Q IS R J M IS M IS Q

C LU S T ER : 1 C LU S TE R : 2 C LU S TE R : 3
100
90
80
70
No of Articles

60
50
40
30
20
10
0
IS R J M IS M IS Q IS R J M IS M IS Q IS R J M IS M IS Q

C LU S T ER : 4 C LU S TE R : 5 C LU S TE R : 6
100
90
80
70
60
50
40
30
20
10
0
IS R J M IS M IS Q IS R J M IS M IS Q IS R J M IS M IS Q

C LU S T ER : 7 C LU S TE R : 8 C LU S TE R : 9

JO U R N AL

-38 © Pearson Education Limited 2014

 Text Mining Tools
 Commercial Software Tools
 IBM SPSS Modler - Text Miner
 SAS Enterprise Miner – Text Miner
 Statistical Data Miner – Text Miner
 ClearForest, …
 Free Software Tools
 RapidMiner
 GATE
 Spy-EM, …
-39 © Pearson Education Limited 2014
 Application Case 7.6
A Potpourri of Text Mining Case Synopses
1. Alberta’s Parks Division gains insight from
unstructured data
2. American Honda Saves Millions by Using Text and
Data Mining
3. MaspexWadowice Group Analyzes Online Brand
Image with Text Mining
4. Viseca Card Services Reduces Fraud Loss with
Text Analytics
5. Improving Quality with Text Mining and Advanced
Analytics
-40 © Pearson Education Limited 2014
 Sentiment Analysis Overview
 Sentiment  belief, view, opinion, conviction
 Sentiment analysis  opinion mining,
subjectivity analysis, and appraisal extraction
 The goal is to answer the question:
“What do people feel about a certain topic?”
 Explicit versus Implicit sentiment
 Sentiment polarity
 Positive versus Negative
 … versus Neutral?
-41 © Pearson Education Limited 2014
 Example –
Real-Time Social Signal (by Attensity)

-42 © Pearson Education Limited 2014

 Application Case 7.7
Whirlpool Achieves Customer Loyalty and
Product Success with Text Analytics

Questions for Discussion

1. How did Whirlpool use capabilities of text
analytics to better understand their customers
and improve product offerings?
2. What were the challenges, the proposed
solution, and the obtained results?

-43 © Pearson Education Limited 2014

 Sentiment Analysis Applications
 Voice of the customer (VOC)
 Voice of the Market (VOM)
 Voice of the Employee (VOE)
 Brand Management
 Financial Markets
 Politics
 Government Intelligence
 … others
-44 © Pearson Education Limited 2014
 Textual Data

Sentiment
Analysis
A statement

Step 1

Process
Calculate the
O-S Polarity
Lexicon

Is there a
No sentiment? Yes
O-S
polarity
Yes measure

Step 2
Calculate the NP N-P Polarity
polarity of the
sentiment
Lexicon Record the Polarity,
Strength, and the
Target of the
Step 3 sentiment.

Identify the target

for the sentiment Target

Step 4
Tabulate & aggregate
the sentiment
analysis results
-45 © Pearson Education Limited 2014
 Sentiment Analysis Process
 Step 1 – Sentiment Detection
 Comes right after the retrieval and
preparation of the text documents
 It is also called detection of objectivity
 Fact [= objectivity] versus Opinion [= subjectivity]
 Step 2 – N-P Polarity Classification
 Given an opinionated piece of text, the goal is
to classify the opinion as falling under one of
two opposing sentiment polarities
 N [= negative] versus P [= positive]
-46 © Pearson Education Limited 2014
 Sentiment Analysis Process
 Step 3 – Target Identification
 The goal of this step is to accurately identify
the target of the expressed sentiment (e.g., a
person, a product, an event, etc.)
 Level of difficulty  the application domain
 Step 4 – Collection and Aggregation
 Once the sentiments of all text data points in
the document are identified and calculated,
they are to be aggregated
 Word  Statement  Paragraph  Document
-47 © Pearson Education Limited 2014
 Sentiment Analysis
Methods for Polarity Identification
 Polarity Identification – P vs. N
 Can be made at the level of word, term,
sentence, paragraph, document
 Two competing methods
1. Using a lexicon
 WordNet [wordnet.princeton.edu]
 SentiWordNet [sentiwordnet.isti.cnr.it]
2. Using pre-classified training documents
 Data mining / machine learning

-48 © Pearson Education Limited 2014

 P-N Polarity and S-O Polarity
Subjective (S)

Positive (P) Negative (N)

P-N Polarity
(+) (-)

S-O Polarity

Objective (O)
-49 © Pearson Education Limited 2014
 Sentiment Analysis and
Speech Analytics
 Speech analytics – analysis of voice
 Content versus other Voice Features
 Two Approaches
 The Acoustic Approach
 Intensity, Pitch, Jitter, Shimmer, etc.
 The Linguistic Approach
 Lexical: words, phrases, etc.
 Disfluencies: filled pauses, hesitation, restarts, etc.
 Higher semantics: taxonomy/ontology, pragmatics
 Many uses and use cases exist
-50 © Pearson Education Limited 2014
 Application Case 7.8
Cutting Through the Confusion: Blue Cross
Blue Shield of North Carolina Uses Nexidia’s
Speech Analytics to Ease Member Experience
in Healthcare

Questions for Discussion

 For a large company like BCBSNC with a lot of
customers, what does “listening to customer” mean?
 What were the challenges, the proposed solution,
and the obtained results for BCBSNC?
-51 © Pearson Education Limited 2014
 End of the Chapter

 Questions, comments

-52 © Pearson Education Limited 2014

 All rights reserved. No part of this publication may be reproduced,
stored in a retrieval system, or transmitted, in any form or by any
means, electronic, mechanical, photocopying, recording, or otherwise,
without the prior written permission of the publisher. Printed in the
United States of America.

-53 © Pearson Education Limited 2014