Course "Machine Learning and Data Mining" for the degree of Computer Engineering at the Politecnico di Milano. In this lecture we overview text and web mining. The slides are mainly taken from Jiawei Han textbook.
Machine Learning and Data Mining: 19 Mining Text And Web Data
1. Text and Web Mining
Machine Learning and Data Mining (Unit 19)
Prof. Pier Luca Lanzi
2. References 2
Jiawei Han and Micheline Kamber,
quot;Data Mining: Concepts and
Techniquesquot;, The Morgan Kaufmann
Series in Data Management Systems
(Second Edition)
Chapter 10, part 2
Web Mining Course by
Gregory-Platesky Shapiro available at
www.kdnuggets.com
Prof. Pier Luca Lanzi
3. Mining Text Data: An Introduction 3
Data Mining / Knowledge Discovery
Structured Data Multimedia Free Text Hypertext
<a href>Frank Rizzo
Frank Rizzo bought
HomeLoan (
</a> Bought
his home from Lake
Loanee: Frank Rizzo
<a hef>this home</a>
View Real Estate in
Lender: MWF
from <a href>Lake
1992.
Agency: Lake View
View Real Estate</a>
He paid $200,000
Amount: $200,000
In <b>1992</b>.
under a15-year loan
Term: 15 years
Loans($200K,[map],...) <p>...
from MW Financial.
)
Prof. Pier Luca Lanzi
4. Bag-of-Tokens Approaches 4
Documents Token Sets
Four score and seven nation – 5
years ago our fathers brought civil - 1
forth on this continent, a new war – 2
nation, conceived in Liberty, Feature men – 2
and dedicated to the Extraction died – 4
proposition that all men are people – 5
created equal. Liberty – 1
Now we are engaged in a God – 1
great civil war, testing …
whether that nation, or …
Loses all order-specific information!
Severely limits context!
Prof. Pier Luca Lanzi
5. Natural Language Processing 5
A dog is chasing a boy on the playground Lexical
analysis
Det Noun Aux Verb Det Noun Prep Det Noun
(part-of-speech
tagging)
Noun Phrase
Noun Phrase Noun Phrase
Complex Verb
Prep Phrase
Semantic analysis Verb Phrase
Syntactic analysis
Dog(d1). (Parsing)
Boy(b1).
Verb Phrase
Playground(p1).
Chasing(d1,b1,p1).
Sentence
+
Scared(x) if Chasing(_,x,_).
A person saying this may
be reminding another person to
get the dog back…
Scared(b1)
Pragmatic analysis
Inference
(speech act)
(Taken from ChengXiang Zhai, CS 397cxz – Fall 2003)
Prof. Pier Luca Lanzi
6. General NLP—Too Difficult! 6
Word-level ambiguity
“design” can be a noun or a verb (Ambiguous POS)
“root” has multiple meanings (Ambiguous sense)
Syntactic ambiguity
“natural language processing” (Modification)
“A man saw a boy with a telescope.” (PP Attachment)
Anaphora resolution
“John persuaded Bill to buy a TV for himself.”
(himself = John or Bill?)
Presupposition
“He has quit smoking.” implies that he smoked before.
Humans rely on context to interpret (when possible).
This context may extend beyond a given document!
(Taken from ChengXiang Zhai, CS 397cxz – Fall 2003)
Prof. Pier Luca Lanzi
7. Shallow Linguistics 7
English Lexicon
Part-of-Speech Tagging
Word Sense Disambiguation
Phrase Detection / Parsing
Prof. Pier Luca Lanzi
8. WordNet 8
An extensive lexical network for the English language
Contains over 138,838 words.
Several graphs, one for each part-of-speech.
Synsets (synonym sets), each defining a semantic sense.
Relationship information (antonym, hyponym, meronym …)
Downloadable for free (UNIX, Windows)
Expanding to other languages (Global WordNet Association)
Funded >$3 million, mainly government (translation interest)
Founder George Miller, National Medal of Science, 1991.
watery parched
moist dry arid
wet
synonym
antonym
damp anhydrous
Prof. Pier Luca Lanzi
9. Part-of-Speech Tagging 9
Training data (Annotated text)
This sentence serves as an example of annotated text…
Det N V1 P Det N P V2 N
This is a new sentence.
POS Tagger
“This is a new sentence.”
Det Aux Det Adj N
Pick the most1likely, ttag tsequence.
p(w ,..., wk 1,..., k )
⎧ p(t1 | w1 )... p(tk | wk ) p(w1 )... p(wk )
⎪
p(w1 ,..., wk , t1 ,..., tk ) = ⎨ k
⎪∏ p(wi | ti ) p(ti | ti−1 )
Independent assignment
⎩ =1
⎧ p(t1 | w1 )... p(tk | wk ) p(iw1 )... p(wk ) Most common tag
⎪
=⎨ k
⎪∏ p(wi | ti ) p(ti | ti −1 )
⎩ i =1 Partial dependency
(HMM)
Prof. Pier Luca Lanzi
10. Word Sense Disambiguation 10
?
“The difficulties of computational linguistics are rooted in ambiguity.”
N Aux V P N
Supervised Learning
Features:
Neighboring POS tags (N Aux V P N)
Neighboring words (linguistics are rooted in ambiguity)
Stemmed form (root)
Dictionary/Thesaurus entries of neighboring words
High co-occurrence words (plant, tree, origin,…)
Other senses of word within discourse
Algorithms:
Rule-based Learning (e.g. IG guided)
Statistical Learning (i.e. Naïve Bayes)
Unsupervised Learning (i.e. Nearest Neighbor)
(Adapted from ChengXiang Zhai, CS 397cxz – Fall 2003)
Prof. Pier Luca Lanzi
11. Parsing 11
(Adapted from ChengXiang Zhai, CS 397cxz – Fall 2003)
Choose most likely parse tree… S Probability of this tree=0.000015
NP VP
Probabilistic CFG
Det
S→ NP VP 1.0 BNP VP PP
NP → Det BNP 0.3
NP → BNP A
0.4 N Aux V NP P NP
0.3
NP→ NP PP
Grammar
is chasing on
BNP→ N dog
…
VP → V a boy
the playground
.
VP → Aux V NP .
.
…
VP → VP PP Probability of this tree=0.000011
S
PP → P NP 1.0
NP VP
V → chasing 0.01 Det NP
BNP Aux V
Aux→ is
PP
N → dog 0.003 A chasing NP
N is
N → boy
Lexicon NP
P
N→ playground …
dog a boy
Det→ the on
…
Det→ a
P → on the playground
Prof. Pier Luca Lanzi
12. Obstacles 12
Ambiguity
A man saw a boy with a telescope.
Computational Intensity
Imposes a context horizon.
Text Mining NLP Approach
Locate promising fragments using fast IR methods
(bag-of-tokens)
Only apply slow NLP techniques to promising fragments
Prof. Pier Luca Lanzi
13. Summary: Shallow NLP 13
However, shallow NLP techniques are feasible and useful:
Lexicon – machine understandable linguistic knowledge
possible senses, definitions, synonyms, antonyms, typeof,
etc.
POS Tagging – limit ambiguity (word/POS), entity extraction
WSD – stem/synonym/hyponym matches (doc and query)
Query: “Foreign cars” Document: “I’m selling a 1976
Jaguar…”
Parsing – logical view of information (inference?,
translation?)
A man saw a boy with a telescope.”
Even without complete NLP, any additional knowledge
extracted from text data can only be beneficial.
Ingenuity will determine the applications.
Prof. Pier Luca Lanzi
14. References for NLP 14
C. D. Manning and H. Schutze, “Foundations of Natural Language
Processing”, MIT Press, 1999.
S. Russell and P. Norvig, “Artificial Intelligence: A Modern
Approach”, Prentice Hall, 1995.
S. Chakrabarti, “Mining the Web: Statistical Analysis of Hypertext
and Semi-Structured Data”, Morgan Kaufmann, 2002.
G. Miller, R. Beckwith, C. FellBaum, D. Gross, K. Miller, and R.
Tengi. Five papers on WordNet. Princeton University, August 1993.
C. Zhai, Introduction to NLP, Lecture Notes for CS 397cxz, UIUC,
Fall 2003.
M. Hearst, Untangling Text Data Mining, ACL’99, invited paper.
http://www.sims.berkeley.edu/~hearst/papers/acl99/acl99-
tdm.html
R. Sproat, Introduction to Computational Linguistics, LING 306,
UIUC, Fall 2003.
A Road Map to Text Mining and Web Mining, University of Texas
resource page. http://www.cs.utexas.edu/users/pebronia/text-
mining/
Computational Linguistics and Text Mining Group, IBM Research,
http://www.research.ibm.com/dssgrp/
Prof. Pier Luca Lanzi
15. Text Databases and Information Retrieval 15
Text databases (document databases)
Large collections of documents from various sources:
news articles, research papers, books, digital libraries,
E-mail messages, and Web pages, library database, etc.
Data stored is usually semi-structured
Traditional information retrieval techniques become
inadequate for the increasingly vast amounts of text data
Information retrieval
A field developed in parallel with database systems
Information is organized into (a large number of)
documents
Information retrieval problem: locating relevant
documents based on user input, such as keywords or
example documents
Prof. Pier Luca Lanzi
16. Information Retrieval (IR) 16
Typical Information Retrieval systems
Online library catalogs
Online document management systems
Information retrieval vs. database systems
Some DB problems are not present in IR, e.g.,
update, transaction management, complex objects
Some IR problems are not addressed well in DBMS, e.g.,
unstructured documents, approximate search using
keywords and relevance
Prof. Pier Luca Lanzi
17. Basic Measures for Text Retrieval 17
Relevant Relevant &
Retrieved
Retrieved
All Documents
Precision: the percentage of retrieved documents that are in fact
relevant to the query (i.e., “correct” responses)
| {Relevant} ∩ {Retrieved } |
precision =
| {Retrieved} |
Recall: the percentage of documents that are relevant to the query
and were, in fact, retrieved
| {Relevant} ∩ {Retrieved } |
precision =
| {Relevant} |
Prof. Pier Luca Lanzi
18. Information Retrieval Techniques 18
Basic Concepts
A document can be described by a set of representative
keywords called index terms.
Different index terms have varying relevance when used
to describe document contents.
This effect is captured through the assignment of
numerical weights to each index term of a document.
(e.g.: frequency, tf-idf)
DBMS Analogy
Index Terms Attributes
Weights Attribute Values
Prof. Pier Luca Lanzi
19. Information Retrieval Techniques 19
Index Terms (Attribute) Selection:
Stop list
Word stem
Index terms weighting methods
Terms Documents Frequency Matrices
Information Retrieval Models:
Boolean Model
Vector Model
Probabilistic Model
Prof. Pier Luca Lanzi
20. Boolean Model 20
Consider that index terms are either present
or absent in a document
As a result, the index term weights are assumed
to be all binaries
A query is composed of index terms linked by three
connectives: not, and, and or
E.g.: car and repair, plane or airplane
The Boolean model predicts that each document is either
relevant or non-relevant based on the match of a document
to the query
Prof. Pier Luca Lanzi
21. Keyword-Based Retrieval 21
A document is represented by a string, which can be
identified by a set of keywords
Queries may use expressions of keywords
E.g., car and repair shop, tea or coffee,
DBMS but not Oracle
Queries and retrieval should consider synonyms,
e.g., repair and maintenance
Major difficulties of the model
Synonymy: A keyword T does not appear anywhere in the
document, even though the document is closely related to
T, e.g., data mining
Polysemy: The same keyword may mean different things
in different contexts, e.g., mining
Prof. Pier Luca Lanzi
22. Similarity-Based Retrieval in Text Data 22
Finds similar documents based on a set of common keywords
Answer should be based on the degree of relevance based on
the nearness of the keywords, relative frequency of the
keywords, etc.
Basic techniques
Stop list
Set of words that are deemed “irrelevant”, even though
they may appear frequently
E.g., a, the, of, for, to, with, etc.
Stop lists may vary when document set varies
Prof. Pier Luca Lanzi
23. Similarity-Based Retrieval in Text Data 23
Word stem
Several words are small syntactic variants of each other since
they share a common word stem
E.g., drug, drugs, drugged
A term frequency table
Each entry frequent_table(i, j) = # of occurrences of the word
ti in document di
Usually, the ratio instead of the absolute number of occurrences
is used
Similarity metrics: measure the closeness of a document to a query
(a set of keywords)
Relative term occurrences
Cosine distance:
v1 ⋅ v2
sim(v1 , v2 ) =
| v1 || v2 |
Prof. Pier Luca Lanzi
24. Indexing Techniques 24
Inverted index
Maintains two hash- or B+-tree indexed tables:
• document_table: a set of document records <doc_id,
postings_list>
• term_table: a set of term records, <term, postings_list>
Answer query: Find all docs associated with one or a set of
terms
+ easy to implement
– do not handle well synonymy and polysemy, and posting lists
could be too long (storage could be very large)
Signature file
Associate a signature with each document
A signature is a representation of an ordered list of terms that
describe the document
Order is obtained by frequency analysis, stemming and stop lists
Prof. Pier Luca Lanzi
25. Vector Space Model 25
Documents and user queries are represented as m-dimensional
vectors, where m is the total number of index terms in the
document collection.
The degree of similarity of the document d with regard to the query
q is calculated as the correlation between the vectors that represent
them, using measures such as the Euclidian distance or the cosine
of the angle between these two vectors.
Prof. Pier Luca Lanzi
26. Latent Semantic Indexing (1) 26
Basic idea
Similar documents have similar word frequencies
Difficulty: the size of the term frequency matrix is very large
Use a singular value decomposition (SVD) techniques to reduce
the size of frequency table
Retain the K most significant rows of the frequency table
Method
Create a term x document weighted frequency matrix A
SVD construction: A = U * S * V’
Define K and obtain Uk ,, Sk , and Vk.
Create query vector q’ .
Project q’ into the term-document space: Dq = q’ * Uk * Sk-1
Calculate similarities: cos α = Dq . D / ||Dq|| * ||D||
Prof. Pier Luca Lanzi
27. Latent Semantic Indexing (2) 27
Weighted Frequency Matrix
Query Terms:
- Insulation
- Joint
Prof. Pier Luca Lanzi
28. Probabilistic Model 28
Basic assumption: Given a user query, there is a set of
documents which contains exactly the relevant documents
and no other (ideal answer set)
Querying process as a process of specifying the properties of
an ideal answer set. Since these properties are not known at
query time, an initial guess is made
This initial guess allows the generation of a preliminary
probabilistic description of the ideal answer set which is used
to retrieve the first set of documents
An interaction with the user is then initiated with the purpose
of improving the probabilistic description of the answer set
Prof. Pier Luca Lanzi
29. Types of Text Data Mining 29
Keyword-based association analysis
Automatic document classification
Similarity detection
Cluster documents by a common author
Cluster documents containing information
from a common source
Link analysis: unusual correlation between entities
Sequence analysis: predicting a recurring event
Anomaly detection: find information that violates usual
patterns
Hypertext analysis
Patterns in anchors/links
Anchor text correlations with linked objects
Prof. Pier Luca Lanzi
30. Keyword-Based Association Analysis 30
Motivation
Collect sets of keywords or terms that occur frequently
together and then find the association or correlation
relationships among them
Association Analysis Process
Preprocess the text data by parsing, stemming,
removing stop words, etc.
Evoke association mining algorithms
• Consider each document as a transaction
• View a set of keywords in the document as
a set of items in the transaction
Term level association mining
• No need for human effort in tagging documents
• The number of meaningless results and
the execution time is greatly reduced
Prof. Pier Luca Lanzi
31. Text Classification (1) 31
Motivation
Automatic classification for the large number of on-line
text documents (Web pages, e-mails, corporate intranets,
etc.)
Classification Process
Data preprocessing
Definition of training set and test sets
Creation of the classification model using the selected
classification algorithm
Classification model validation
Classification of new/unknown text documents
Text document classification differs
from the classification of relational data
Document databases are not structured according to
attribute-value pairs
Prof. Pier Luca Lanzi
32. Text Classification (2) 32
Classification Algorithms:
Support Vector Machines
K-Nearest Neighbors
Naïve Bayes
Neural Networks
Decision Trees
Association rule-based
Boosting
Prof. Pier Luca Lanzi
33. Document Clustering 33
Motivation
Automatically group related documents based on their contents
No predetermined training sets or taxonomies
Generate a taxonomy at runtime
Clustering Process
Data preprocessing:
remove stop words, stem, feature extraction, lexical analysis,
etc.
Hierarchical clustering:
compute similarities applying clustering algorithms.
Model-Based clustering (Neural Network Approach):
clusters are represented by “exemplars”. (e.g.: SOM)
Prof. Pier Luca Lanzi
34. Text Categorization 34
Pre-given categories and labeled document examples
(Categories may form hierarchy)
Classify new documents
A standard classification (supervised learning ) problem
Sports
Categorization
Business
System
Education
… …
Science
Sports
Business
Education
Prof. Pier Luca Lanzi
35. Applications 35
News article classification
Automatic email filtering
Webpage classification
Word sense disambiguation
……
Prof. Pier Luca Lanzi
36. Categorization Methods 36
Manual: Typically rule-based
Does not scale up (labor-intensive, rule inconsistency)
May be appropriate for special data on a particular
domain
Automatic: Typically exploiting machine learning techniques
Vector space model based
• Prototype-based (Rocchio)
• K-nearest neighbor (KNN)
• Decision-tree (learn rules)
• Neural Networks (learn non-linear classifier)
• Support Vector Machines (SVM)
Probabilistic or generative model based
• Naïve Bayes classifier
Prof. Pier Luca Lanzi
37. Vector Space Model 37
Represent a doc by a term vector
Term: basic concept, e.g., word or phrase
Each term defines one dimension
N terms define a N-dimensional space
Element of vector corresponds to term weight
E.g., d = (x1,…,xN), xi is “importance” of term I
New document is assigned to the most likely category based
on vector similarity
Prof. Pier Luca Lanzi
38. Illustration of the Vector Space Model 38
Starbucks
C2 Category 2
Category 3
C3
Java
new doc
C1 Category 1
Microsoft
Prof. Pier Luca Lanzi
39. What VS Model Does Not Specify? 39
How to select terms to capture “basic concepts”
Word stopping
• e.g. “a”, “the”, “always”, “along”
Word stemming
• e.g. “computer”, “computing”, “computerize” => “compute”
Latent semantic indexing
How to assign weights
Not all words are equally important: Some are more
indicative than others
• e.g. “algebra” vs. “science”
How to measure the similarity
Prof. Pier Luca Lanzi
40. How to Assign Weights 40
Two-fold heuristics based on frequency
TF (Term frequency)
More frequent within a document more relevant to
semantics
E.g., “query” vs. “commercial”
IDF (Inverse document frequency)
Less frequent among documents more discriminative
E.g. “algebra” vs. “science”
Prof. Pier Luca Lanzi
41. Term Frequency Weighting 41
Weighting
The more frequent, the more relevant to topic
E.g. “query” vs. “commercial”
Raw TF= f(t,d): how many times term t appears in doc d
Normalization
When document length varies,
then relative frequency is preferred
E.g., Maximum frequency normalization
Prof. Pier Luca Lanzi
42. Inverse Document Frequency Weighting 42
Idea: The less frequent terms among documents are the
more discriminative
Formula:
Given
n, total number of docs
k, the number of docs with term t appearing
(the DF document frequency)
Prof. Pier Luca Lanzi
43. TF-IDF Weighting 43
TF-IDF weighting: weight(t, d) = TF(t, d) * IDF(t)
Frequent within doc high tf high weight
Selective among docs high idf high weight
Recall VS model
Each selected term represents one dimension
Each doc is represented by a feature vector
Its t-term coordinate of document d is the TF-IDF weight
This is more reasonable
Just for illustration …
Many complex and more effective weighting variants exist
in practice
Prof. Pier Luca Lanzi
44. How to Measure Similarity? 44
Given two document
Similarity definition
dot product
normalized dot product (or cosine)
Prof. Pier Luca Lanzi
45. Illustrative Example 45
text
doc1 mining Sim(newdoc,doc1)=4.8*2.4+4.5*4.5
search
engine
Sim(newdoc,doc2)=2.4*2.4
text
To whom is newdoc
more similar?
travel
text
Sim(newdoc,doc3)=0
map
doc2 travel
text mining travel map search engine govern president congress
IDF(faked) 2.4 4.5 2.8 3.3 2.1 5.4 2.2 3.2 4.3
government doc1 2(4.8) 1(4.5) 1(2.1) 1(5.4)
president doc2 1(2.4 ) 2 (5.6) 1(3.3)
congress
doc3 doc3 1 (2.2) 1(3.2) 1(4.3)
newdoc 1(2.4) 1(4.5)
……
Prof. Pier Luca Lanzi
46. Vector Space Model-Based Classifiers 46
What do we have so far?
A feature space with similarity measure
This is a classic supervised learning problem
Search for an approximation to classification hyper plane
VS model based classifiers
K-NN
Decision tree based
Neural networks
Support vector machine
Prof. Pier Luca Lanzi
47. Probabilistic Model 47
Category C is modeled as a probability distribution
of predefined random events
Random events model the process
of generating documents
Therefore, how likely a document d belongs
to category C is measured through the probability
for category C to generate d.
Prof. Pier Luca Lanzi
48. Quick Revisit of Bayes’ Rule 48
Category Hypothesis space: H = {C1 , …, Cn}
One document: D
P ( D | Ci ) P(Ci )
P (Ci | D) =
P( D)
As we want to pick the most likely category C*, we can drop
p(D)
Posterior probability of Ci
C* = arg max C P(C | D) = arg max C P ( D | C ) P(C )
Document model for category C
Prof. Pier Luca Lanzi
49. Probabilistic Model: Multi-Bernoulli 49
Event: word presence or absence
D = (x1, …, x|V|)
xi =1 for presence of word wi
xi =0 for absence
Parameters
{p(wi=1|C), p(wi=0|C)}
p(wi=1|C)+ p(wi=0|C)=1
|V | |V | |V |
p( D = ( x1 ,..., x|V | ) | C )= ∏ p( wi = xi | C ) = ∏ ∏
p(wi = 1| C ) p( wi = 0 | C )
i =1 i =1, xi =1 i =1, xi =0
Prof. Pier Luca Lanzi
50. Probabilistic Model: Multinomial 50
Event: word selection/sampling
D = (n1, …, n|V|)
ni: frequency of word wi
n=n1,+…+ n|V|
Parameters
{p(wi|C)}
p(w1|C)+… p(w|v||C) = 1
⎛ ⎞ |V |
n
n1 ... n|V | ⎠ ∏
p ( D = ( n1 , ..., n|v | ) | C )= p ( n | C ) ⎜ p ( w i | C ) ni
⎟
⎝ i =1
Prof. Pier Luca Lanzi
51. Parameter Estimation 51
Category prior
Training examples:
| E (Ci ) |
p (Ci ) =
E(C2)
E(C1) k
∑ | E (C ) | j
j =1
C1 C2
Multi-Bernoulli Doc model
Ck ∑ δ(w ,d)+0.5 ⎧1 if wj occursin d
j
d∈E(Ci )
δ(wj ,d) = ⎨
p(wj =1| Ci ) =
| E(Ci )| +1 ⎩0 otherwise
E(Ck)
Multinomial doc model
Vocabulary: V = {w1, …, w|V|}
∑ c(w ,d)+1 j
d∈E(Ci )
p(wj | Ci ) = |V| c(wj ,d) =counts of wj in d
∑ ∑ c(w ,d)+|V | m
m=1 d∈E(Ci )
Prof. Pier Luca Lanzi
52. Classification of New Document 52
Multi-Bernoulli Multinomial
d = (x1,..., x|V| ) x ∈{0,1 d = (n1,..., n|V| ) | d |= n = n1 +... + n|V|
}
C* = argmaxC P(D | C)P(C)
C* = argmaxC P(D | C)P(C)
|V|
= argmaxC p(n | C)∏p(wi | C)ni P(C)
|V|
= argmaxC ∏p(wi = xi | C)P(C) i=1
i=1 |V|
= argmaxC log p(n | C) + log p(C) + ∑ni log p(wi | C)
|V|
= argmaxC log p(C) + ∑log p(wi = xi | C) i=1
|V|
≈ argmaxC log p(C) + ∑ni log p(wi | C)
i=1
i=1
Prof. Pier Luca Lanzi
53. Categorization Methods 53
Vector space model
K-NN
Decision tree
Neural network
Support vector machine
Probabilistic model
Naïve Bayes classifier
Many, many others and variants exist
e.g. Bim, Nb, Ind, Swap-1, LLSF, Widrow-Hoff, Rocchio,
Gis-W, … …
Prof. Pier Luca Lanzi
55. Evaluation (2) 55
Benchmarks
Classic: Reuters collection
• A set of newswire stories classified under categories related
to economics.
Effectiveness
Difficulties of strict comparison
• different parameter setting
• different “split” (or selection) between training and testing
• various optimizations … …
However widely recognizable
• Best: Boosting-based committee classifier & SVM
• Worst: Naïve Bayes classifier
Need to consider other factors, especially efficiency
Prof. Pier Luca Lanzi
56. Summary: Text Categorization 56
Wide application domain
Comparable effectiveness to professionals
Manual Text Classification is not 100% and unlikely to
improve substantially
Automatic Text Classification is growing at a steady pace
Prospects and extensions
Very noisy text, such as text from O.C.R.
Speech transcripts
Prof. Pier Luca Lanzi
57. Research Problems in Text Mining 57
Google: what is the next step?
How to find the pages that match approximately the
sophisticated documents, with incorporation of user-profiles
or preferences?
Look back of Google: inverted indicies
Construction of indicies for the sophisticated documents,
with incorporation of user-profiles or preferences
Similarity search of such pages using such indicies
Prof. Pier Luca Lanzi
58. References for Text Mining 58
Fabrizio Sebastiani, “Machine Learning in Automated Text
Categorization”, ACM Computing Surveys, Vol. 34, No.1,
March 2002
Soumen Chakrabarti, “Data mining for hypertext: A tutorial
survey”, ACM SIGKDD Explorations, 2000.
Cleverdon, “Optimizing convenient online accesss to
bibliographic databases”, Information Survey, Use4, 1, 37-
47, 1984
Yiming Yang, “An evaluation of statistical approaches to text
categorization”, Journal of Information Retrieval, 1:67-88,
1999.
Yiming Yang and Xin Liu “A re-examination of text
categorization methods”. Proceedings of ACM SIGIR
Conference on Research and Development in Information
Retrieval (SIGIR'99, pp 42--49), 1999.
Prof. Pier Luca Lanzi
59. World Wide Web: a brief history 59
Who invented the wheel is unknown
Who invented the World-Wide Web ?
(Sir) Tim Berners-Lee
in 1989, while working at CERN,
invented the World Wide Web,
including URL scheme, HTML, and in
1990 wrote the first server and the
first browser
Mosaic browser developed by Marc
Andreessen and Eric Bina at NCSA
(National Center for Supercomputing
Applications) in 1993; helped rapid
web spread
Mosaic was basis for Netscape …
Prof. Pier Luca Lanzi
60. What is Web Mining? 60
Discovering interesting and useful information
from Web content and usage
Examples
Web search, e.g. Google, Yahoo, MSN, Ask, …
Specialized search: e.g. Froogle (comparison shopping),
job ads (Flipdog)
eCommerce
Recommendations (Netflix, Amazon, etc.)
Improving conversion rate: next best product to offer
Advertising, e.g. Google Adsense
Fraud detection: click fraud detection, …
Improving Web site design and performance
Prof. Pier Luca Lanzi
61. How does it differ from “classical” 61
Data Mining?
The web is not a relation
Textual information and linkage structure
Usage data is huge and growing rapidly
Google’s usage logs are bigger than their web crawl
Data generated per day is comparable to largest
conventional data warehouses
Ability to react in real-time to usage patterns
No human in the loop
Reproduced from Ullman & Rajaraman with permission
Prof. Pier Luca Lanzi
62. How big is the Web? 62
The number of pages is technically, infinite
Because of dynamically generated content
Lots of duplication (30-40%)
Best estimate of “unique” static HTML pages comes from
search engine claims
Google = 8 billion
Yahoo = 20 billion
Lots of marketing hype
Reproduced from Ullman & Rajaraman with permission
Prof. Pier Luca Lanzi
63. Netcraft Survey: 63
76,184,000 web sites (Feb 2006)
Netcraft survey
http://news.netcraft.com/archives/web_server_survey.html
Prof. Pier Luca Lanzi
64. Mining the World-Wide Web 64
The WWW is huge, widely distributed, global information
service center for
Information services: news, advertisements, consumer
information, financial management, education,
government, e-commerce, etc.
Hyper-link information
Access and usage information
WWW provides rich sources for data mining
Challenges
Too huge for effective data warehousing and data mining
Too complex and heterogeneous: no standards and
structure
Prof. Pier Luca Lanzi
65. Web Mining: A more challenging task 65
Searches for
Web access patterns
Web structures
Regularity and dynamics of Web contents
Problems
The “abundance” problem
Limited coverage of the Web: hidden Web sources,
majority of data in DBMS
Limited query interface based on keyword-oriented search
Limited customization to individual users
Prof. Pier Luca Lanzi
66. Web Mining Taxonomy 66
Web Mining
Web Content Web Structure Web Usage
Mining Mining Mining
Web Page General Access Customized
Search Result
Content Mining Pattern Tracking Usage Tracking
Mining
Prof. Pier Luca Lanzi
67. Mining the World-Wide Web 67
Web Mining
Web Content
Mining
Web Structure
Web Usage
Mining
Mining
Web Page Content Mining
Web Page Summarization
General Access Customized
Search Result
WebLog , WebOQL …: Pattern Tracking Usage Tracking
Mining
Web Structuring query languages;
Can identify information within given web
pages
Ahoy!:Uses heuristics to distinguish personal
home pages from other web pages
ShopBot: Looks for product prices within web
pages
Prof. Pier Luca Lanzi
68. Mining the World-Wide Web 68
Web Mining
Web Content
Web Structure
Mining
Web Usage
Mining
Mining
Web Page
Content Mining
Search Result Mining General Access Customized
Pattern Tracking Usage Tracking
Search Engine Result
Summarization
Clustering Search Result :
Categorizes documents using
phrases in titles and snippets
Prof. Pier Luca Lanzi
69. Mining the World-Wide Web 69
Web Mining
Web Content Web Usage
Mining Mining
Web Structure Mining
Using Links
PageRank
CLEVER
Use interconnections between web pages to give
General Access
Search Result weight to pages.
Pattern Tracking
Mining
Using Generalization
Web Page Customized
MLDB, VWV
Content Mining Usage Tracking
Uses a multi-level database representation of the
Web. Counters (popularity) and link lists are used
for capturing structure.
Prof. Pier Luca Lanzi
70. Mining the World-Wide Web 70
Web Mining
Web Content Web Structure Web Usage
Mining Mining Mining
General Access Pattern Tracking
Web Page Customized
Content Mining Usage Tracking
Web Log Mining
Search Result Uses KDD techniques to understand
Mining
general access patterns and trends.
Can shed light on better structure and
grouping of resource providers.
Prof. Pier Luca Lanzi
71. Mining the World-Wide Web 71
Web Mining
Web Content Web Structure Web Usage
Mining Mining Mining
Customized Usage Tracking
Web Page General Access
Content Mining Pattern Tracking
Adaptive Sites
Analyzes access patterns of each user at a
Search Result
time. Web site restructures itself automatically
Mining
by learning from user access patterns.
Prof. Pier Luca Lanzi
72. Web Usage Mining 72
Mining Web log records to discover user access patterns of
Web pages
Applications
Target potential customers for electronic commerce
Enhance the quality and delivery of Internet information
services to the end user
Improve Web server system performance
Identify potential prime advertisement locations
Web logs provide rich information about Web dynamics
Typical Web log entry includes the URL requested, the IP
address from which the request originated, and a
timestamp
Prof. Pier Luca Lanzi
73. Web Usage Mining 73
Understanding is a pre-requisite to improvement
1 Google, but 70,000,000+ web sites
What Applications?
Simple and Basic
• Monitor performance, bandwidth usage
• Catch errors (404 errors- pages not found)
• Improve web site design (shortcuts for frequent paths,
remove links not used, etc)
•…
Advanced and Business Critical
• eCommerce: improve conversion, sales, profit
• Fraud detection: click stream fraud, …
•…
Prof. Pier Luca Lanzi
74. Techniques for Web usage mining 74
Construct multidimensional view on the Weblog database
Perform multidimensional OLAP analysis to find the top N
users, top N accessed Web pages, most frequently
accessed time periods, etc.
Perform data mining on Weblog records
Find association patterns, sequential patterns, and trends
of Web accessing
May need additional information,e.g., user browsing
sequences of the Web pages in the Web server buffer
Conduct studies to
Analyze system performance, improve system design by
Web caching, Web page prefetching, and Web page
swapping
Prof. Pier Luca Lanzi
75. References 75
Deng Cai, Shipeng Yu, Ji-Rong Wen and Wei-Ying Ma, “Extracting Content Structure for
Web Pages based on Visual Representation”, The Fifth Asia Pacific Web Conference,
2003.
Deng Cai, Shipeng Yu, Ji-Rong Wen and Wei-Ying Ma, “VIPS: a Vision-based Page
Segmentation Algorithm”, Microsoft Technical Report (MSR-TR-2003-79), 2003.
Shipeng Yu, Deng Cai, Ji-Rong Wen and Wei-Ying Ma, “Improving Pseudo-Relevance
Feedback in Web Information Retrieval Using Web Page Segmentation”, 12th
International World Wide Web Conference (WWW2003), May 2003.
Ruihua Song, Haifeng Liu, Ji-Rong Wen and Wei-Ying Ma, “Learning Block Importance
Models for Web Pages”, 13th International World Wide Web Conference (WWW2004),
May 2004.
Deng Cai, Shipeng Yu, Ji-Rong Wen and Wei-Ying Ma, “Block-based Web Search”,
SIGIR 2004, July 2004 .
Deng Cai, Xiaofei He, Ji-Rong Wen and Wei-Ying Ma, “Block-Level Link Analysis”, SIGIR
2004, July 2004 .
Deng Cai, Xiaofei He, Wei-Ying Ma, Ji-Rong Wen and Hong-Jiang Zhang, “Organizing
WWW Images Based on The Analysis of Page Layout and Web Link Structure”, The
IEEE International Conference on Multimedia and EXPO (ICME'2004) , June 2004
Deng Cai, Xiaofei He, Zhiwei Li, Wei-Ying Ma and Ji-Rong Wen, “Hierarchical Clustering
of WWW Image Search Results Using Visual, Textual and Link Analysis”,12th ACM
International Conference on Multimedia, Oct. 2004 .
Prof. Pier Luca Lanzi