Transcript Text Mining: Finding Nuggets in Mountains of Textual Data

Jochen Dijrre, Peter Gerstl, Roland Seiffert
Presented by Shamil Mustafayev
04/16/2013
1
Outline
Definition
 Motivation
 Methodology
 Feature Extraction
 Clustering and Categorizing
 Some Applications
 Comparison with Data Mining
 Conclusion & Exam Questions

2
Definition

Text Mining:
 the discovery by computer of new, previously
unknown information, by automatically extracting
information from different written resources.
 Also referred to as text data mining, roughly
equivalent to text analytics, refers to the process
of deriving high-quality information from text.
3
Outline
Definition
 Motivation
 Methodology
 Feature Extraction
 Clustering and Categorizing
 Some Applications
 Comparison with Data Mining
 Conclusion & Exam Questions

4
Motivation

A large portion of a company’s data is
unstructured or semi-structured
 Letters
 Emails
 Phone transcripts
 Contracts
 Technical documents
 Patents
 Web pages
 Articles
5
Typical Applications








Summarizing documents
Discovering/monitoring relations among
people, places, organizations, etc
Customer profile analysis
Trend analysis
Documents summarization
Spam Identification
Public health early warning
Event tracks
6
Outline
Definition
 Motivation
 Methodology
 Comparison with Data Mining
 Feature Extraction
 Clustering and Categorizing
 Some Applications
 Conclusion & Exam Questions

7
Methodology: Challenges
Information is in unstructured textual
form
 Natural language interpretation is
difficult & complex task! (not fully
possible)

 Google and Watson are a step closer

Text mining deals with huge collections
of documents
 Impossible for human examination
8
Google vs Watson
Google justifies the
answer by returning
the text documents
where it found the
evidence.
 Google finds
documents that are
most suitable to a
given Keyword.




Watson tries to
understand the
semantics behind
a given key phrase
or question.
Then Watson will
use its huge
knowledge base to
find the correct
answer.
Watson uses more AI
9
Methodology: Two Aspects

Knowledge Discovery
 Extraction of codified information
○ Feature Extraction
 Mining proper; determining some structure

Information Distillation
 Analysis of feature distribution
10
Two Text Mining Approaches

Extraction
 Extraction of codified information from single
document

Analysis
 Analysis of the features to detect patterns,
trends, etc, over whole collections of
documents
11
Outline
Definition
 Motivation
 Methodology
 Feature Extraction
 Clustering and Categorizing
 Some Applications
 Comparison with Data Mining
 Conclusion & Exam Questions

12
Feature Extraction

Recognize and classify “significant”
vocabulary items from the text

Categories of vocabulary





Proper names – Mrs. Albright or Dheli[sic], India
Multiword terms – Joint venture, online document
Abbreviations – CPU, CEO
Relations – Jack Smith-age-42
Other useful things: numerical forms of
numbers, percentages, money, etc
13
Canonical Form Examples

Normalize numbers, money
 Four = 4, five-hundred dollar = $500
Conversion of date to normal form
 Morphological variants

 Drive, drove, driven = drive

Proper names and other forms
 Mr. Johnson, Bob Johnson, The author = Bob
Johnson
14
Feature Extraction Approach
Linguistically motivated heuristics
 Pattern matching
 Limited lexical information (part-ofspeech)
 Avoid analyzing with too much depth

 Does not use too much lexical information
 No in-depth syntactic or semantic analysis
15
IBM Intelligent Miner for Text
IBM introduced Intelligent Miner for Text in
1998
 SDK with: Feature extraction, clustering,
categorization, and more
 Traditional components (search engine, etc)
 The rest of the paper describes text mining
methodology of Intelligent Miner.

16
Advantages to IBM’s approach
Processing is very fast (helps when
dealing with huge amounts of data)
 Heuristics work reasonably well
 Generally applicable to any domain

17
Outline
Definition
 Motivation
 Methodology
 Comparison with Data Mining
 Feature Extraction
 Clustering and Categorizing
 Some Applications
 Conclusion & Exam Questions

18
Clustering
Fully automatic process
 Documents are grouped according to
similarity of their feature vectors
 Each cluster is labeled by a listing of the
common terms/keywords
 Good for getting an overview of a
document collection

19
Two Clustering Engines

Hierarchical clustering
 Orders the clusters into a tree reflecting
various levels of similarity

Binary relational clustering
 Flat clustering
 Relationships of different strengths between
clusters, reflecting similarity
20
Clustering Model
21
Categorization





Assigns documents to preexisting
categories
Classes of documents are defined by
providing a set of sample documents.
Training phase produces “categorization
schema”
Documents can be assigned to more than
one category
If confidence is low, document is set aside
for human intervention
22
Categorization Model
23
Outline
Definition
 Motivation
 Methodology
 Feature Extraction
 Clustering and Categorizing
 Some Applications
 Comparison with Data Mining
 Conclusion & Exam Questions

24
Applications

Customer Relationship Management
application provided by IBM Intelligent
Miner for Text called “Customer
Relationship Intelligence”
 “Help companies better understand what
their customers want and what they think
about the company itself”
25
Customer Intelligence Process
Take as input body of communications
with customer
 Cluster the documents to identify issues
 Characterize the clusters to identify the
conditions for problems
 Assign new messages to appropriate
clusters

26
Customer Intelligence Usage

Knowledge Discovery
 Clustering used to create a structure that
can be interpreted

Information Distillation
 Refinement and extension of clustering
results
○ Interpreting the results
○ Tuning of the clustering process
○ Selecting meaningful clusters
27
Outline
Definition
 Motivation
 Methodology
 Feature Extraction
 Clustering and Categorizing
 Some Applications
 Comparison with Data Mining
 Conclusion & Exam Questions

28
Comparison with Data Mining

Data mining
 Discover hidden
models.
 tries to generalize all
of the data into a
single model.
 marketing, medicine,
health care
 Text mining



Discover hidden
facts.
tries to understand
the details, cross
reference between
individual instances
biosciences,
customer profile
analysis
29
Outline
Definition
 Motivation
 Methodology
 Feature Extraction
 Clustering and Categorizing
 Some Applications
 Comparison with Data Mining
 Conclusion & Exam Questions

30
Conclusion
This paper introduced text mining and
how it differs from data mining proper.
 Focused on the tasks of feature
extraction and clustering/categorization
 Presented an overview of the
tools/methods of IBM’s Intelligent Miner
for Text

31
Exam Question #1

What are the two aspects of Text Mining?
 Knowledge Discovery: Discovering a
common customer complaint in a large
collection of documents containing customer
feedback.
 Information Distillation: Filtering future
comments into pre-defined categories
32
Exam Question #2

How does the procedure for text mining
differ from the procedure for data mining?
 Adds feature extraction phase
 Infeasible for humans to select features
manually
 The feature vectors are, in general, highly
dimensional and sparse
33
Exam Question #3

In the Nominator program of IBM’s Intelligent
Miner for Text, an objective of the design is to
enable rapid extraction of names from large
amounts of text. How does this decision affect
the ability of the program to interpret the
semantics of text?
 Does not perform in-depth syntactic or semantic
analysis of the text; the results are fast but only
heuristic with regards to actual semantics of the text.
34
35