Download Text Mining: Finding Nuggets in Mountains of Textual Data

Document related concepts

Cluster analysis wikipedia , lookup

Transcript
Text Mining:
Finding Nuggets in Mountains of Textual Data
Authors:Jochen Dijrre, Peter Gerstl, Roland Seiffert
Adapted from slides by: Trevor Crum
Presenter: Nicholas Romano
1
Outline
●
●
●
●
●
●
●
●
Definition and Paper Overview
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 unstructured textual
documents.
Also referred to as text data mining, roughly
equivalent to text analytics which refers more
specifically to problems based in a business settings.
3
Paper Overview
● 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
4
Outline
●
●
●
●
●
●
●
●
Definition and Paper Overview
Motivation
Methodology
Feature Extraction
Clustering and Categorizing
Some Applications
Comparison with Data Mining
Conclusion & Exam Questions
5
Motivation
● A large portion of a company’s data is
unstructured or semi-structured – about 90%
in 1999!
•
•
•
•
Letters
Emails
Phone transcripts
Contracts
•
•
•
•
Technical documents
Patents
Web pages
Articles
6
Typical Applications
● Summarizing documents
● Discovering/monitoring relations among
people, places, organizations, etc
● Customer profile analysis
● Trend analysis
● Document summarization
● Spam Identification
● Public health early warning
● Event tracks
7
Outline
●
●
●
●
●
●
●
●
Definition and Paper Overview
Motivation
Methodology
Comparison with Data Mining
Feature Extraction
Clustering and Categorizing
Some Applications
Conclusion & Exam Questions
8
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
9
Google vs Watson
● Google justifies the ● Watson tries to
answer by returning
understand the
the text documents
semantics behind a
where it found the
given key phrase or
evidence.
question.
● Google finds
● Then Watson will
documents that are
use its huge
most suitable to a
knowledge base to
given Keyword.
find the correct
answer.
10
Methodology: Two Aspects
● Knowledge Discovery
○
○
Extraction of codified information
■ Feature Extraction
Mining proper; determining some structure
● Information Distillation
○
Analysis of feature distribution
11
Two Text Mining
Approaches
● Extraction
○
Extraction of codified information from a single
document
● Analysis
○
Analysis of the features to detect patterns, trends, and
other similarities over whole collections of documents
12
Outline
●
●
●
●
●
●
●
●
Definition and Paper Overview
Motivation
Methodology
Feature Extraction
Clustering and Categorizing
Some Applications
Comparison with Data Mining
Conclusion & Exam Questions
13
Feature Extraction
● Recognize and classify “significant”
vocabulary items from the text
● Categories of vocabulary
○
○
○
○
○
Proper names – Mrs. Albright or Dheli, India
Multiword terms – Joint venture, online document
Abbreviations – CPU, CEO
Relations – Jack Smith-age-42
Other useful things: numerical forms of numbers,
percentages, money, dates, and many other
14
Canonical Form Examples
● Normalize numbers, money
○
Four = 4, five-hundred dollars = $500
● Conversion of date to normal form
○
8/17/1992 = August 18 1992
● Morphological variants
○
Drive, drove, driven = drive
● Proper names and other forms
○
Mr. Johnson, Bob Johnson, The author = Bob
Johnson
15
Feature Extraction
Approach
●
●
●
●
Linguistically motivated heuristics
Pattern matching
Limited lexical information (part-of-speech)
Avoid analyzing with too much depth
○
○
Does not use too much lexical information
No in-depth syntactic or semantic analysis
16
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)
17
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
18
Outline
●
●
●
●
●
●
●
●
Definition and Paper Overview
Motivation
Methodology
Comparison with Data Mining
Feature Extraction
Clustering and Categorizing
Some Applications
Conclusion & Exam Questions
19
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
20
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
21
Clustering Model
22
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
23
Categorization Model
24
Outline
●
●
●
●
●
●
●
●
Definition and Paper Overview
Motivation
Methodology
Feature Extraction
Clustering and Categorizing
Some Applications
Comparison with Data Mining
Conclusion & Exam Questions
25
Applications
● Customer Relationship Management
application provided by IBM Intelligent Miner
for Text called “Customer Relationship
Intelligence” or CRI
○
“Help companies better understand what their
customers want and what they think about the
company itself”
26
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
27
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
28
Outline
●
●
●
●
●
●
●
●
Definition and Paper Overview
Motivation
Methodology
Feature Extraction
Clustering and Categorizing
Some Applications
Comparison with Data Mining
Conclusion & Exam Questions
29
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
30
Outline
●
●
●
●
●
●
●
●
Definition and Paper Overview
Motivation
Methodology
Feature Extraction
Clustering and Categorizing
Some Applications
Comparison with Data Mining
Conclusion & Exam Questions
31
Conclusion
● Text mining can be used as an effective
business tool that supports
○
Creation of knowledge by preparing and organizing
unstructured textual data [Knowledge Discovery]
○ Extraction of relevant information from large amounts
of unstructured textual data through automatic preselection based on user defined criteria [Information
Distillation]
32
Exam Question #1
● 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
Questions?
34
pg 01
Web Mining Research:
A Survey
Authors: Raymond Kosala & Hendrik Blockeel
Presenter: Nick Romano Slides adapted from: Ryan Patterson
April 23rd 2014
CS332 Data Mining
pg
03
outline
•
•
•
•
•
•
•
Introduction
Web Mining
Web Content Mining
Web Structure Mining
Web Usage Mining
Review
Exam Questions
pg
04
Introduction
“The Web is huge, diverse, and dynamic . . . we
are currently drowning in information and facing
information overload.”
Web users encounter problems:
• Finding relevant information
• Creating new knowledge out of the information available
on the Web
• Personalization of the information
• Learning about consumers or individual users
pg
05
outline
•
•
•
•
•
•
•
Introduction
Web Mining
Web Content Mining
Web Structure Mining
Web Usage Mining
Review
Exam Questions
pg
06
Web Mining
“Web mining is the use of data mining
techniques to automatically discover and
extract information from Web documents and
services.”
Web mining subtasks:
1.
2.
3.
4.
Resource finding
Information selection and pre-processing
Generalization
Analysis
Information Retrieval &
Information Extraction
• Information Retrieval (IR)
o
the automatic retrieval of all relevant documents
while at the same time retrieving as few of the nonrelevant as possible
• Information Extraction (IE)
o
transforming a collection of documents into
information that is more readily digested and
analyzed
pg
07
pg
09
outline
•
•
•
•
•
•
•
Introduction
Web Mining
Web Content Mining
Web Structure Mining
Web Usage Mining
Review
Exam Questions
Web Content Mining
Information Retrieval View
Unstructured Documents
• Most utilizes “bag of words” representation to generate documents features
o ignores the sequence in which the words occur
• Document features can be reduced with selection algorithms
o ie. information gain
• Possible alternative document feature representations:
o word positions in the document
o phrases/terms (ie. “annual interest rate”)
Semi-Structured Documents
• Utilize additional structural information gleaned from the document
o HTML markup (intra-document structure)
o HTML links (inter-document structure)
pg
10
pg 11
Web content mining, IR unstructured documents
pg 12
Web content mining, IR semi structured documents
Web Content Mining
Database View
“the Database view tries . . . to transform a Web site to become a database so
that . . . querying on the Web become[s] possible.”
• Uses Object Exchange Model (OEM)
o represents semi-structured data by a labeled graph
• Database view algorithms typically start from manually
selected Web sites
o site-specific parsers
• Database view algorithms produce:
o extract document level schema or DataGuides
▪ structural summary of semi-structured data
o extract frequent substructures (sub-schema)
o multi-layered database
▪ each layer is obtained by generalizations on lower layers
pg
13
pg 14
Web content mining, Database view
pg
15
outline
•
•
•
•
•
•
•
Introduction
Web Mining
Web Content Mining
Web Structure Mining
Web Usage Mining
Review
Exam Questions
pg
16
Web Structure Mining
“. . . we are interested in the structure of the hyperlinks within the Web itself”
• Inspired by the study of social networks and citation
analysis
o based on incoming & outgoing links we could discover specific types
of pages (such as hubs, authorities, etc)
• Some algorithms calculate the quality/relevancy of each
Web page
o ie. Page Rank
• Others measure the completeness of a Web site
o measuring frequency of local links on the same server
o interpreting the nature of hierarchy of hyperlinks on one domain
pg
17
outline
•
•
•
•
•
•
•
Introduction
Web Mining
Web Content Mining
Web Structure Mining
Web Usage Mining
Review
Exam Questions
pg
18
Web Usage Mining
“. . . focuses on techniques that could predict user behavior while the user
interacts with the Web.”
• Web usage is mined by parsing Web server logs
o mapped into relational tables → data mining techniques applied
o log data utilized directly
• Users connecting through proxy servers and/or users or
ISP’s utilizing caching of Web data results in decreased
server log accuracy
• Two applications:
o personalized - user profile or user modeling in adaptive interfaces
o impersonalized - learning user navigation patterns
pg
19
outline
•
•
•
•
•
•
•
Introduction
Web Mining
Web Content Mining
Web Structure Mining
Web Usage Mining
Review
Exam Questions
pg
20
Review
• Web mining
o
o
4 subtasks
IR & IE
• Web content mining
o
o
primarily intra-page analysis
IR view vs DB view
• Web structure mining
o
primarily inter-page analysis
• Web usage mining
o
primarily analysis of server activity logs
pg 21
Web Mining
Web Content Mining
Web Structure Mining
IR View
Web Usage Mining
DB View
- Unstructured
- Semi structured
- Semi structured
- Web site as DB
- Links structure
- Interactivity
Main Data
- Text documents
- Hypertext documents
- Hypertext documents
- Links structure
- Server logs
- Browser logs
Representation
- Bag of word, n-grams
- Terms, phrases
- Concepts of ontology
- Relational
- Edge-labeled graph (OEM)
- Relational
- Graph
- Relational table
- Graphs
- TFIDF and variants
- Machine learning
- Statistical (incl. NLP)
- Proprietary algorithms
- ILP
- (modified) association
rules
- Proprietary algorithms
- Machine Learning
- Statistical
- (modified) association rules
- Categorization
- Clustering
- Finding extraction rules
- Finding patterns in text
- User modeling
- Finding frequent substructures
- Web site schema
discovery
- Categorization
- Clustering
- Site construction, adaptation,
and management
- Marketing
- User modeling
View of Data
Method
Application
Categories
Web mining categories
pg
22
outline
•
•
•
•
•
•
•
Introduction
Web Mining
Web Content Mining
Web Structure Mining
Web Usage Mining
Review
Exam Questions
pg
24
Exam Question 2
Q:
Of the following Web mining paradigms:
• Information Retrieval
• Information Extraction
Which does a traditional Web search engine (google.com,
bing.com, etc.) attempt to accomplish? Briefly support your
answer.
A:
Information Retrieval, the search engine attempts
provides a list of documents ranked by their relevancy to
the search query.
pg
26
Exam Question 3
Q:
State one common problem hampering accurate
Web usage mining? Briefly support your answer.
A:
• Users connecting to a Web site though a proxy server,
• Users (or their ISP’s) utilizing Web data caching,will
result in decreased server log accuracy. Accurate
server logs are required for accurate Web usage
mining.
Exam Question 1 (Again)
● 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
57
Questions?
58