sharda_dss10_ppt_07_GE-211570

download report

Transcript sharda_dss10_ppt_07_GE-211570

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…)
7-2
© Pearson Education Limited 2014
Learning Objectives




7-3
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
© Pearson Education Limited 2014
Opening Vignette…
Machine Versus Men on Jeopardy!:
The Story of Watson





7-4
Situation
Problem Watch it on YouTube!
Solution https://www.youtube.com/watch?v=YLR1byL0U8M
Results
Answer & discuss the case questions...
© Pearson Education Limited 2014
Questions for
the Opening Vignette
1.
2.
3.
4.
7-5
What is Watson? What is special about it?
What technologies were used in building
Watson (both hardware and software)?
What are the innovative characteristics of
DeepQA architecture that made Watson
superior?
Why did IBM spend all that time and money
to build Watson? Where is the ROI?
© Pearson Education Limited 2014
A High-Level Depiction of IBM
Watson’s DeepQA Architecture
Answer
sources
Candidate
answer
generation
Primary
search
Question
Evidence
sources
Support
evidence
retrieval
Deep
evidence
scoring
Trained
models
1
Question
analysis
7-6
Query
decomposition
Hypothesis
generation
Soft
filtering
Hypothesis and
evidence scoring
Hypothesis
generation
Soft
filtering
Hypothesis and
evidence scoring
...
...
...
© Pearson Education Limited 2014
Synthesis
2
3
Final merging
and ranking
Answer and
confidence
4
5
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


7-7
A semi-automated process of extracting knowledge
from unstructured data sources
a.k.a. text data mining or knowledge discovery in
textual databases
© Pearson Education Limited 2014
Text Analytics and Text Mining
TEXT ANALYTICS
Text Mining
Information
Retrieval
Web Mining
Information
Extraction
Natural Language Processing
Computer Science
7-8
Statistics
Data Mining
Linguistic
Management Science
Machine Learning
Artificial Intelligence
© 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:




7-9
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.
© Pearson Education Limited 2014
Text Mining Concepts

Benefits of text mining are obvious, especially in
text-rich data environments


Electronic communication records (e.g., Email)



7-10
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.
Spam filtering
Email prioritization and categorization
Automatic response generation
© Pearson Education Limited 2014
Text Mining Application Area







7-11
Information extraction
Topic tracking
Summarization
Categorization
Clustering
Concept linking
Question answering
© Pearson Education Limited 2014
Text Mining Terminology








7-12
Unstructured or semi-structured data
Corpus (and corpora)
Terms
Concepts
Stemming
Stop words (and include words)
Synonyms (and polysemes)
Tokenizing
© Pearson Education Limited 2014
Text Mining Terminology





Term dictionary
Word frequency
Part-of-speech tagging
Morphology
Term-by-document matrix


Singular value decomposition

7-13
Occurrence matrix
Latent semantic indexing
© Pearson Education Limited 2014
Application Case 7.1
Text Mining for Patent Analysis

What is a patent?



How do we do patent analysis (PA)?
Why do we need to do PA?



7-14
“exclusive rights granted by a country to an
inventor for a limited period of time in
exchange for a disclosure of an invention”
What are the benefits?
What are the challenges?
How does text mining help in PA?
© Pearson Education Limited 2014
Natural Language Processing
(NLP)

Structuring a collection of text



NLP is …




7-15
Old approach: bag-of-words
New approach: natural language processing
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
© Pearson Education Limited 2014
Natural Language Processing
(NLP)

What is “Understanding” ?




7-16
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?
© Pearson Education Limited 2014
Natural Language Processing
(NLP)

Challenges in NLP







Dream of AI community

7-17
Part-of-speech tagging
Text segmentation
Word sense disambiguation
Syntax ambiguity
Imperfect or irregular input
Speech acts
to have algorithms that are capable of automatically
reading and obtaining knowledge from text
© Pearson Education Limited 2014
Natural Language Processing
(NLP)

WordNet




Sentiment Analysis


7-18
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.
A technique used to detect favorable and unfavorable
opinions toward specific products and services
SentiWordNet
© 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.
2.
7-19
How did the Hong Kong government use
text mining to better serve its constituents?
What were the challenges, the proposed
solution, and the obtained results?
© Pearson Education Limited 2014
NLP Task Categories









7-20
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
© Pearson Education Limited 2014
Text Mining Applications

Marketing applications


Security applications



Literature-based gene identification (…)
Academic applications

7-21
ECHELON, OASIS
Deception detection (…)
Medicine and biology


Enables better CRM
Research stream analysis
© Pearson Education Limited 2014
Application Case 7.3
Mining for Lies!

Deception detection



The study


7-22
A difficult problem
If detection is limited to only text, then the
problem is even more difficult
analyzed text-based testimonies of persons of
interest at military bases
used only text-based features (cues)
© Pearson Education Limited 2014
Application Case 7.3
Mining for Lies
Statements
Transcribed for
Processing
Statements Labeled as
Truthful or Deceptive
By Law Enforcement
Cues Extracted &
Selected
Classification Models
Trained and Tested on
Quantified Cues
Text Processing
Software Identified
Cues in Statements
Text Processing
Software Generated
Quantified Cues
7-23
© Pearson Education Limited 2014
Application Case 7.3
Mining for Lies
7-24
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.
© 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)



7-25
Logistic regression
Decision trees
Neural networks
67.28
71.60
73.46
© Pearson Education Limited 2014
Text Mining Applications
Gene/
Protein
(Gene/Protein Interaction Identification)
596 12043 24224 281020
42722 397276
D007962
Ontology
D 016923
D 001773
D019254
D044465
D001769
D002477
D003643
D016158
Word
8 51112
9
23017
27
5874
2791
8952
1623
5632
17
8252
8 2523
NN
IN
NN
IN
VBZ
IN
JJ
JJ
NN
NN
NN
CC
NN
IN NN
NP
PP
NP
NP
PP NP
Shallow
Parse
185
POS
...expression of Bcl-2 is correlated with insufficient white blood cell death and activation of p53.
7-26
NP
PP
NP
© Pearson Education Limited 2014
Application Case 7.4
Text mining and Sentiment Analysis
Help Improve Customer Service
Performance
Questions for Discussion
1.
2.
7-27
How did the financial services firm use text
mining and text analytics to improve its
customer service performance?
What were the challenges, the proposed
solution, and the obtained results?
© Pearson Education Limited 2014
Text Mining Process
Context diagram for
the text mining
process
Unstructured data (text)
Structured data (databases)
Software/hardware limitations
P rivacy issues
Linguistic limitations
Extract
knowledge
from available
data sources
A0
Context-specific knowledge
Domain expertise
Tools and techniques
7-28
© Pearson Education Limited 2014
Text Mining Process
Task 1
Establish the Corpus:
Collect & Organize the
Domain Specific
Unstructured Data
Task 2
Create the TermDocument Matrix:
Introduce Structure
to the Corpus
Feedback
The inputs to the process
includes a variety of relevant
unstructured (and semistructured) data sources such
as text, XML, HTML, etc.
The output of the Task 1 is a
collection of documents in
some digitized format for
computer processing
Task 3
Extract Knowledge:
Discover Novel
Patterns from the
T-D Matrix
Feedback
The output of the Task 2 is a
flat file called term-document
matrix where the cells are
populated with the term
frequencies
The three-step text mining process
7-29
© Pearson Education Limited 2014
The output of Task 3 is a
number of problem specific
classification, association,
clustering models and
visualizations
Text Mining Process

Step 1: Establish the corpus



7-30
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)
© Pearson Education Limited 2014
Text Mining Process

Step 2: Create the Term-by-Document
Matrix (TDM)
Terms
Documents
Document 1
e
em
e
ine
1
1
1
Document 3
3
Document 4
1
1
Document 5
2
1
1
1
...
7-31
g
rin
k
g
ris
nt
ng
na
t
e
e
a
n
m
e
m
are
ct
lop
stm
w
e
e
t
j
P
e
f
v
so
SA
pro
inv
de
Document 2
Document 6
nt
© Pearson Education Limited 2014
...
Text Mining Process

Step 2: Create the Term-by-Document
Matrix (TDM)

Should all terms be included?




What is the best representation of the indices
(values in cells)?


7-32
Stop words, include words
Synonyms, homonyms
Stemming
Row counts; binary frequencies; log frequencies;
Inverse document frequency
© 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?




7-33
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
© Pearson Education Limited 2014
Text Mining Process

Step 3: Extract patterns/knowledge


Classification (text categorization)
Clustering (natural groupings of text)






7-34
Improve search recall
Improve search precision
Scatter/gather
Query-specific clustering
Association
Trend Analysis (…)
© Pearson Education Limited 2014
Application Case 7.5
(Research Literature Survey with Text Mining)

Mining the published IS literature







7-35
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
© Pearson Education Limited 2014
Application Case 7.5
(Research Literature Survey with Text Mining)
Journal Year
Author(s)
MISQ
2005
A. Malhotra,
S. Gosain and
O. A. El Sawy
ISR
1999
JMIS
2001
R. Aron and
E. K. Clemons
…
…
…
7-36
Title
Vol/No Pages
Absorptive capacity
configurations in
supply chains:
Gearing for partnerenabled market
knowledge creation
D. Robey and
Accounting for the
M. C. Boudreau contradictory
organizational
consequences of
information
technology:
Theoretical directions
and methodological
implications
Keywords
Abstract
145-187 knowledge management
supply chain
absorptive capacity
interorganizational
information systems
configuration approaches
2-Oct 167-185 organizational
transformation
impacts of technology
organization theory
research methodology
intraorganizational power
electronic communication
mis implementation
culture
systems
Achieving the optimal 18/2 65-88
information products
balance between
internet advertising
investment in quality
product positioning
and investment in selfsignaling
promotion for
signaling games
information products
…
29/1
…
…
…
© Pearson Education Limited 2014
The need for continual value
innovation is driving supply
chains to evolve from a pure
transactional focus to
leveraging interorganizational
partner ships for sharing
Although much contemporary
thought considers advanced
information technologies as
either determinants or enablers
of radical organizational
change, empirical studies have
revealed inconsistent findings to
support the deterministic logic
implicit in such arguments. This
paper reviews the contradictory
When producers of goods (or
services) are confronted by a
situation in which their offerings
no longer perfectly match
consumer preferences, they
must determine the extent to
which the advertised features of
…
7-37
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
CLUSTER: 4
CLUSTER: 5
CLUSTER: 6
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
35
30
25
20
15
10
5
0
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
No of Articles
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
Application Case 7.5
(Research Literature Survey with Text Mining)
35
30
25
20
15
10
5
0
CLUSTER: 1
CLUSTER: 2
CLUSTER: 3
35
30
25
20
15
10
5
0
CLUSTER: 7
CLUSTER: 8
CLUSTER: 9
YEAR
© Pearson Education Limited 2014
Application Case 7.5
(Research Literature Survey with Text Mining)
100
90
80
70
60
50
40
30
20
10
0
ISR
JMIS
MISQ
No of Articles
CLUSTER: 1
ISR
JMIS
MISQ
CLUSTER: 2
ISR
JMIS
MISQ
CLUSTER: 3
100
90
80
70
60
50
40
30
20
10
0
ISR
JMIS
MISQ
CLUSTER: 4
ISR
JMIS
MISQ
CLUSTER: 5
ISR
JMIS
MISQ
CLUSTER: 6
100
90
80
70
60
50
40
30
20
10
0
ISR
JMIS
MISQ
CLUSTER: 7
ISR
JMIS
MISQ
CLUSTER: 8
JOURNAL
7-38
© Pearson Education Limited 2014
ISR
JMIS
MISQ
CLUSTER: 9
Text Mining Tools

Commercial Software Tools





Free Software Tools



7-39
IBM SPSS Modler - Text Miner
SAS Enterprise Miner – Text Miner
Statistical Data Miner – Text Miner
ClearForest, …
RapidMiner
GATE
Spy-EM, …
© Pearson Education Limited 2014
Application Case 7.6
A Potpourri of Text Mining Case Synopses
1.
2.
3.
4.
5.
7-40
Alberta’s Parks Division gains insight from
unstructured data
American Honda Saves Millions by Using Text and
Data Mining
MaspexWadowice Group Analyzes Online Brand
Image with Text Mining
Viseca Card Services Reduces Fraud Loss with
Text Analytics
Improving Quality with Text Mining and Advanced
Analytics
© 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


7-41
Positive versus Negative
… versus Neutral?
© Pearson Education Limited 2014
Example –
Real-Time Social Signal
7-42
© Pearson Education Limited 2014
(by Attensity)
Application Case 7.7
Whirlpool Achieves Customer Loyalty and
Product Success with Text Analytics
Questions for Discussion
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?
1.
7-43
© Pearson Education Limited 2014
Sentiment Analysis Applications








7-44
Voice of the customer (VOC)
Voice of the Market (VOM)
Voice of the Employee (VOE)
Brand Management
Financial Markets
Politics
Government Intelligence
… others
© Pearson Education Limited 2014
Sentiment
Analysis
Process
Textual Data
A statement
Step 1
Calculate the
O-S Polarity
Lexicon
No
Is there a
sentiment?
Yes
O-S
polarity
measure
Yes
Step 2
Calculate the NP
N-P Polarity
polarity of the
sentiment
Record the Polarity,
Strength, and the
Target of the
sentiment.
Lexicon
Step 3
Identify the target
for the sentiment
Target
Step 4
Tabulate & aggregate
the sentiment
analysis results
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


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

7-46
Fact [= objectivity] versus Opinion [= subjectivity]
N [= negative] versus P [= positive]
© 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.)


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

7-47
Level of difficulty  the application domain
Word  Statement  Paragraph  Document
© 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


2.
Using pre-classified training documents

7-48
WordNet [wordnet.princeton.edu]
SentiWordNet [sentiwordnet.isti.cnr.it]
Data mining / machine learning
© Pearson Education Limited 2014
P-N Polarity and S-O Polarity
Subjective (S)
P-N Polarity
S-O Polarity
Positive (P)
(+)
Objective (O)
7-49
© Pearson Education Limited 2014
Negative (N)
(-)
Sentiment Analysis and
Speech Analytics

Speech analytics – analysis of voice


Content versus other Voice Features
Two Approaches

The Acoustic Approach


The Linguistic Approach




7-50
Intensity, Pitch, Jitter, Shimmer, etc.
Lexical: words, phrases, etc.
Disfluencies: filled pauses, hesitation, restarts, etc.
Higher semantics: taxonomy/ontology, pragmatics
Many uses and use cases exist
© 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


7-51
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?
© Pearson Education Limited 2014
End of the Chapter

7-52
Questions, comments
© 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.
7-53
© Pearson Education Limited 2014