Cs511 data-extraction

Chapter 9:
Structured Data Extraction
Supervised and unsupervised
wrapper generation

CS511, Bing Liu, UIC 2
Road map
 Introduction
 Data Model and HTML encoding
 Wrapper induction
 Automatic Wrapper Generation: Two Problems
 String Matching and Tree Matching
 Multiple Alignments
 Building DOM Trees
 Extraction Given a List Page: Flat Data Records
 Extraction Given a List Page: Nested Data Records
 Extraction Given Multiple Pages
 Summary

Introduction
 A large amount of information on the Web is
contained in regularly structured data objects.
 often data records retrieved from databases.
 Such Web data records are important: lists of
products and services.
 Applications: e.g.,
 Comparative shopping, meta-search, meta-query,
etc.
 We introduce:
 Wrapper induction (supervised learning)
 automatic extraction (unsupervised learning)

Two types of data rich pages
 List pages
 Each such page contains one or more lists of data
records.
 Each list in a specific region in the page
 Two types of data records: flat and nested
 Detail pages
 Each such page focuses on a single object.
 But can have a lot of related and unrelated
information

Extraction results

Road map
 Introduction
 Summary

The data model
 Most Web data can be modeled as nested
relations
 typed objects allowing nested sets and tuples.
 An instance of a type T is simply an element of
dom(T).

An example nested tuple type
 Classic flat relations are of un-nested or flat
set types.
 Nested relations are of arbitrary set types.

Type tree
 A basic type Bi is a leaf tree,
 A tuple type [T1, T2, …, Tn] is a tree rooted at a
tuple node with n sub-trees, one for each Ti.
 A set type {T} is a tree rooted at a set node with
one sub-tree.
Note: attribute names are not included in the type tree.
We introduce a labeling of a type tree, which is
defined recursively:
 If a set node is labeled φ, then its child is labeled
φ.0, a tuple node.
 If a tuple node is labeled φ, then its n children are
labeled φ.1, …, φ.n.

Instance tree
 An instance (constant) of a basic type is a leaf
tree.
 A tuple instance [v1, v2, …, vn] forms a tree rooted at
a tuple node with n children or sub-trees
representing attribute values v1, v2, …, vn.
 A set instance {e1, e2, …, en} forms a set node with
n children or sub-trees representing the set
elements e1, e2, …, and en.
Note: A tuple instance is usually called a data record
in data extraction research.

HTML mark-up encoding of data
 There are no designated tags for each type as
HTML was not designed as a data encoding
language. Any HTML tag can be used for any type.
 For a tuple type, values (also called data items) of
different attributes are usually encoded differently
to distinguish them and to highlight important
items.
 A tuple may be partitioned into several groups or
sub-tuples. Each group covers a disjoint subset of
attributes and may be encoded differently.

HTML encoding (cont …)

More on HTML encoding
 By no means, this mark-up encoding covers all
cases in Web pages.
 In fact, each group of a tuple type can be further divided.
 We must also note that in an actual Web page the
encoding may not be done by HTML tags alone.
 Words and punctuation marks can be used as well.

Road map
 Introduction
 Summary

Wrapper induction
 Using machine learning to generate extraction rules.
 The user marks the target items in a few training pages.
 The system learns extraction rules from these pages.
 The rules are applied to extract items from other pages.
 Many wrapper induction systems, e.g.,
 WIEN (Kushmerick et al, IJCAI-97),
 Softmealy (Hsu and Dung, 1998),
 Stalker (Muslea et al. Agents-99),
 BWI (Freitag and Kushmerick, AAAI-00),
 WL2
(Cohen et al. WWW-02).
 We will only focus on Stalker, which also has a
commercial version, Fetch.

Stalker: A hierarchical wrapper induction
system
 Hierarchical wrapper learning
 Extraction is isolated at different levels of hierarchy
 This is suitable for nested data records (embedded list)
 Each item is extracted independent of others.
 Each target item is extracted using two rules
 A start rule for detecting the beginning of the target item.
 A end rule for detecting the ending of the target item.

Hierarchical representation: type tree

Data extraction based on EC tree
 The extraction is done using a tree structure called the
EC tree (embedded catalog tree).
 The EC tree is based on the type tree above.
 To extract each target item (a node), the wrapper
needs a rule that extracts the item from its parent.

Extraction using two rules
 Each extraction is done using two rules,
 a start rule and a end rule.
 The start rule identifies the beginning of the
node and the end rule identifies the end of
the node.
 This strategy is applicable to both leaf nodes
(which represent data items) and list nodes.
 For a list node, list iteration rules are
needed to break the list into individual data
records (tuple instances).

Rules use landmarks
 The extraction rules are based on the idea of
landmarks.
 Each landmark is a sequence of consecutive
tokens.
 Landmarks are used to locate the beginning
and the end of a target item.
 Rules use landmarks

An example
 Let us try to extract the restaurant name “Good Noodles”.
Rule R1 can to identify the beginning :
R1: SkipTo() // start rule
 This rule means that the system should start from the
beginning of the page and skip all the tokens until it sees the
first tag. is a landmark.
 Similarly, to identify the end of the restaurant name, we use:
R2: SkipTo() // end rule

Rules are not unique
 Note that a rule may not be unique. For example, we
can also use the following rules to identify the
beginning of the name:
R3: SkiptTo(Name _Punctuation_ _HtmlTag_)
orR4: SkiptTo(Name) SkipTo()
 R3 means that we skip everything till the word
“Name” followed by a punctuation symbol and then
a HTML tag. In this case, “Name _Punctuation_
_HtmlTag_” together is a landmark.
 _Punctuation_ and _HtmlTag_ are wildcards.

Extract area codes

Learning extraction rules
 Stalker uses sequential covering to learn
extraction rules for each target item.
 In each iteration, it learns a perfect rule that
covers as many positive examples as possible
without covering any negative example.
 Once a positive example is covered by a rule, it is
removed.
 The algorithm ends when all the positive
examples are covered. The result is an ordered
list of all learned rules.

The top level algorithm

Example: Extract area codes

Learn disjuncts

Example
 For the example E2 of Fig. 9, the following
candidate disjuncts are generated:
D1: SkipTo( ( )
D2: SkipTo(_Punctuation_)
 D1 is selected by BestDisjunct
 D1 is a perfect disjunct.
 The first iteration of LearnRule() ends. E2
and E4 are removed

The next iteration of LearnRule
 The next iteration of LearnRule() is left with
E1 and E3.
 LearnDisjunct() will select E1 as the Seed
Two candidates are then generated:
D3: SkipTo( )
D4: SkipTo( _HtmlTag_ )
 Both these two candidates match early in the
uncovered examples, E1 and E3. Thus, they
cannot uniquely locate the positive items.
 Refinement is needed.

Refinement
 To specialize a disjunct by adding more
terminals to it.
 A terminal means a token or one of its
matching wildcards.
 We hope the refined version will be able to
uniquely identify the positive items in some
examples without matching any negative item
in any example in E.
 Two types of refinement
 Landmark refinement
 Topology refinement

Landmark refinement
 Landmark refinement: Increase the size of a
landmark by concatenating a terminal.
 E.g.,
D5: SkipTo( - )
D6: SkipTo( _Punctuation_ )

Topology refinement
 Topology refinement: Increase the number of
landmarks by adding 1-terminal landmarks, i.e., t
and its matching wildcards

Refining, specializing

The final solution
 We can see that D5, D10, D12, D13, D14, D15, D18
and D21 match correctly with E1 and E3 and fail to
match on E2 and E4.
 Using BestDisjunct in Fig. 13, D5 is selected as the
final solution as it has longest last landmark (- ).
 D5 is then returned by LearnDisjunct().
 Since all the examples are covered, LearnRule()
returns the disjunctive (start) rule either D1 or D5
R7: either SkipTo( ( )
or SkipTo(- )

Summary
 The algorithm learns by sequential covering
 It is based on landmarks.
 The algorithm is by no mean the only
possible algorithm.
 Many variations are possible. There are
entirely different algorithms.
 In our discussion, we used only the SkipTo()
function in extraction rules.
 SkipUntil() is useful too.

Identifying informative examples
 Wrapper learning needs manual labeling of training
examples.
 To ensure accurate learning, a large number of
training examples are needed.
 Manual labeling labor intensive and time consuming.
 Is it possible to automatically select (unlabelled)
examples that are informative for the user to label.
 Clearly, examples of the same formatting are of limited use.
 Examples that represent exceptions are informative as they
are different from already labeled examples.

Active learning
 help identify informative unlabeled examples in
learning automatically.

Active learning: co-testing
 Co-testing exploits the fact that there are often
multiple ways of extracting the same item.
 Thus, the system can learn different rules, forward
and backward rules, to locate the same item.
 Let us use learning of start rules as an example.
The rules learned in Section 8.2.2 are called forward
rules because they consume tokens from the
beginning of the example to the end.
 In a similar way, we can also learn backward rules
that consume tokens from the end of the example to
the beginning.

Co-testing (cont …)
 Given an unlabeled example, both the forward rule
and backward rule are applied.
 If the two rules disagree on the beginning of a target
item in the example, this example is given to the
user to label.
 Intuition: When the two rules agree, the extraction is
very likely to be correct.
 When the two rules do not agree on the example, one of
them must be wrong.
 By giving the user the example to label, we obtain an
informative training example.

Wrapper maintenance
 Wrapper verification: If the site changes, does the
wrapper know the change?
 Wrapper repair: If the change is correctly detected,
how to automatically repair the wrapper?
 One way to deal with both problems is to learn the
characteristic patterns of the target items.
 These patterns are then used to monitor the
extraction to check whether the extracted items
are correct.

Wrapper maintenance (cont …)
 Re-labeling: If they are incorrect, the same patterns
can be used to locate the correct items assuming
that the page changes are minor formatting
changes.
 Re-learning: re-learning produces a new wrapper.
 Difficult problems: These two tasks are extremely
difficult because it often needs contextual and
semantic information to detect changes and to find
the new locations of the target items.
 Wrapper maintenance is still an active research
area.

Road map
 Introduction
 Summary

Automatic wrapper generation
 Wrapper induction (supervised) has two main
shortcomings:
 It is unsuitable for a large number of sites due to
the manual labeling effort.
 Wrapper maintenance is very costly. The Web is a
dynamic environment. Sites change constantly.
Since rules learnt by wrapper induction systems
mainly use formatting tags, if a site changes its
formatting templates, existing extraction rules for
the site become invalid.

Unsupervised learning is possible
 Due to these problems, automatic (or
unsupervised) extraction has been studied.
 Automatic extraction is possible because
data records (tuple instances) in a Web site
are usually encoded using a very small
number of fixed templates.
 It is possible to find these templates by
mining repeated patterns.

Two data extraction problems
 In Sections 8.1.2 and 8.2.3, we described an
abstract model of structured data on the Web
(i.e., nested relations), and a HTML mark-up
encoding of the data model respectively.
 The general problem of data extraction is to
recover the hidden schema from the HTML
mark-up encoded data.
 We study two extraction problems, which are
really quite similar.

Problem 1: Extraction given a single list
page
 Input: A single HTML string S, which contain k non-
overlapping substrings s1, s2, …, sk with each si
encoding an instance of a set type. That is, each si
contains a collection Wi of mi (≥ 2) non-overlapping
sub-substrings encoding mi instances of a tuple
type.
 Output: k tuple types σ1, σ2, …, σk, and k collections
C1, C2, …, Ck, of instances of the tuple types such
that for each collection Ci there is a HTML encoding
function enci such that enci: Ci → Wi is a bijection.

Problem 2: Data extraction given multiple
pages
 Input: A collection W of k HTML strings,
which encode k instances of the same type.
 Output: A type σ, and a collection C of
instances of type σ, such that there is a
HTML encoding enc such that enc: C → W is
a bijection.

Templates as regular expressions
 A regular expression can be naturally used to
model the HTML encoded version of a nested
type.
 Given an alphabet of symbols Σ and a special
token "#text" that is not in Σ,
 a regular expression over Σ is a string over Σ ∪
{#text, *, ?, |, (, )} defined as follows:

Regular expressions
 The empty string ε and all elements of Σ ∪ {#text}
are regular expressions.
 If A and B are regular expressions, then AB, (A|B)
and (A)?
are regular expressions, where (A|B)
stands for A or B and (A)?
stands for (A|ε).
 If A is a regular expression, (A)* is a regular
expression, where (A)* stands for ε or A or AA
or ...
We also use (A)+ as a shortcut for A(A)*, which can be
used to model the set type of a list of tuples. (A)?
indicates that A is optional. (A|B) represents a
disjunction.

Regular expressions and extraction
 Regular expressions are often employed to
represent templates (or encoding functions).
 However, templates can also be represented
as string or tree patterns as we will see later.
 Extraction:
 Given a regular expression, a nondeterministic
finite-state automaton can be constructed and
employed to match its occurrences in string
sequences representing Web pages.
 In the process, data items can be extracted, which
are text strings represented by #text.

Road map
 Introduction
 Summary

Some useful algorithms
 The key is to finding the encoding template
from a collection of encoded instances of the
same type.
 A natural way to do this is to detect repeated
patterns from HTML encoding strings.
 String edit distance and tree edit distance are
obvious techniques for the task. We describe
these techniques.

String edit distance
 String edit distance: the most widely used
string comparison technique.
 The edit distance of two strings, s1 and s2,
is defined as the minimum number of point
mutations required to change s1 into s2,
where a point mutation is one of:
 (1) change a letter,
 (2) insert a letter, and
 (3) delete a letter.

String edit distance (definition)

Dynamic programming

An example
 The edit distance matrix and
back trace path
 alignment

Tree Edit Distance
 Tree edit distance between two trees A and B
(labeled ordered rooted trees) is the cost
associated with the minimum set of
operations needed to transform A into B.
 The set of operations used to define tree edit
distance includes three operations:
 node removal,
 node insertion, and
 node replacement.
A cost is assigned to each of the operations.

Definition

Simple tree matching
 In the general setting,
 mapping can cross levels, e.g., node a in tree A and node a
in tree B.
 Replacements are also allowed, e.g., node b in A and node
h in B.
 We describe a restricted matching algorithm, called
simple tree matching (STM), which has been
shown quite effective for Web data extraction.
 STM is a top-down algorithm.
 Instead of computing the edit distance of two trees, it
evaluates their similarity by producing the maximum
matching through dynamic programming.

Simple Tree Matching algo

An example

Road map
 Introduction
 Summary

Multiple alignment
 Pairwise alignment is not sufficient because a
web page usually contain more than one data
records.
 We need multiple alignment.
 We discuss two techniques
 Center Star method
 Partial tree alignment.

Center star method
 This is a classic technique, and quite simple. It
commonly used for multiple string alignments, but
can be adopted for trees.
 Let the set of strings to be aligned be S. In the
method, a string sc that minimizes,
 is first selected as the center string. d(sc, si) is the
distance of two strings.
 The algorithm then iteratively computes the
alignment of rest of the strings with sc.
∑ ∈Ss ic
i
ssd ),( (3)

The algorithm

An example

The shortcomings
 Assume there are k strings in S and all strings have
length n, finding the center takes O(k2
n2
) time and
the iterative pair-wise alignment takes O(kn2
) time.
Thus, the overall time complexity is O(k2
n2
).

Shortcomings (cont …)
 Giving the cost of 1 for “changing a letter” in edit
distance is problematic (e.g., A and X in the first and
second strings in the final result) because of
optional data items in data records.
 The problem can be partially dealt with by
disallowing “changing a letter” (e.g., giving it a larger
cost). However, this introduces another problem.
 For example, if we align only ABC and XBC, it is not
clear which of the following alignment is better.

The partial tree alignment method
 Choose a seed tree: A seed tree, denoted by Ts, is
picked with the maximum number of data items.
 The seed tree is similar to center string, but without
the O(k2
n2
) pair-wise tree matching to choose it.
 Tree matching:
For each unmatched tree Ti (i ≠ s),
 match Ts and Ti.
 Each pair of matched nodes are linked (aligned).
 For each unmatched node nj in Ti do
 expand Ts by inserting nj into Ts if a position for insertion
can be uniquely determined in Ts.
The expanded seed tree Ts is then used in
subsequent matching.

p p
a b e dc eb
dc e
pNew part of Ts
e ab x
p pTs
Ti
a e
ba
Ts Ti
Insertion is possible
Insertion is not
possible
Partial tree alignment of two trees

Partial alignment of two trees

dx… b
p
c k gn
p
b
dx… b
p
kcx
…
b
p
d h
c k gn
p
b
nx… b
p
c d h k
No node inserted
T2 T3
T2
g
Ts
New Ts
d h kc
p
b
c, h, and k inserted
Ts = T1
T2 is matched again
A complete
example

Output Data Table
… x b n c d h k g
T1
… 1 1 1
T2
1 1 1 1 1
T3
1 1 1 1 1

Road map
 Introduction
 Summary

Building DOM trees
 We now start to talk about actual data extraction.
 The usual first step is to build a DOM tree (tag tree)
of a HTML page.
 Most HTML tags work in pairs. Within each corresponding
tag-pair, there can be other pairs of tags, resulting in a
nested structure.
 Building a DOM tree from a page using its HTML code is
thus natural.
 In the tree, each pair of tags is a node, and the
nested tags within it are the children of the node.

Two steps to build a tree
 HTML code cleaning:
 Some tags do not require closing tags (e.g., <li>, <hr> and
) although they have closing tags.
 Additional closing tags need to be inserted to ensure all
tags are balanced.
 Ill-formatted tags need to be fixed. One popular program is
called Tidy, which can be downloaded from http://
tidy.sourceforge.net/.
 Tree building: simply follow the nested blocks of
the HTML tags in the page to build the DOM tree. It
is straightforward.

Building tree using tags & visual cues
 Correcting errors in HTML can be hard.
 There are also dynamically generated pages
with scripts.
 Visual information comes to the rescue.
 As long as a browser can render a page
correct, a tree can be built correctly.
 Each HTML element is rendered as a rectangle.
 Containments of rectangles representing nesting.

An example

Road map
 Introduction
 Summary

Extraction Given a List Page: Flat Data
Records
 Given a single list page with multiple data
records,
 Automatically segment data records
 Extract data from data records.
 Since the data records are flat (no nested
lists), string similarity or tree matching can be
used to find similar structures.
 Computation is a problem
 A data record can start anywhere and end
anywhere

Two important observations
 Observation 1: A group of data records that
contains descriptions of a set of similar
objects are typically presented in a
contiguous region of a page and are
formatted using similar HTML tags. Such a
region is called a data region.
 Observation 2: A set of data records are
formed by some child sub-trees of the same
parent node.

An example

The DOM tree

The Approach
Given a page, three steps:
 Building the HTML Tag Tree
 Erroneous tags, unbalanced tags, etc
 Mining Data Regions
 Spring matching or tree matching
 Identifying Data Records
Rendering (or visual) information is very useful
in the whole process

Mining a set of similar structures
 Definition: A generalized node (a node
combination) of length r consists of r (r ≥ 1) nodes in
the tag tree with the following two properties:
 the nodes all have the same parent.
 the nodes are adjacent.
 Definition: A data region is a collection of two or
more generalized nodes with the following
properties:
 the generalized nodes all have the same parent.
 the generalized nodes all have the same length.
 the generalized nodes are all adjacent.
 the similarity between adjacent generalized nodes is
greater than a fixed threshold.

Mining Data Regions
1
3
10
2
7 8 9
Region 2
5 6
4
11 12
14 15 16 17 191813 20
Region 1
Region 3

Mining data regions
 We need to find where each generalized
node starts and where it ends.
 perform string or tree matching
 Computation is not a problem anymore
 Due to the two observations, we only need to
perform comparisons among the children nodes
of a parent node.
 Some comparisons done for earlier nodes are
the same as for later nodes (see the example
below).

Comparison

Comparison (cont …)

The MDR algorithm

Find data records from generalized nodes
 A generalized node may
not represent a data
record.
 In the example on the right,
each row is found as a
generalized node.
 This step needs to identify
each of the 8 data record.
 Not hard
 We simply run the MDR
algorithm given each
generalized node as input
 There are some
complications (read the
notes)

2. Extract Data from Data Records
 Once a list of data records is identified, we
can align and extract data items from them.
 Approaches (align multiple data records):
 Multiple string alignment
 Many ambiguities due to pervasive use of table related
tags.
 Multiple tree alignment (partial tree alignment)
 Together with visual information is effective

Generating extraction patterns and data
extraction
 Once data records in each data region are
discovered, we align them to produce an
extraction pattern that can be used to extract
data from the current page and also other
pages that use the same encoding template.
 Partial tree alignment algorithm is just for
the purpose.
 Visual information can help in various ways
(read the notes)

Road map
 Introduction
 Extraction Given a List Page: Nested Data
Records
 Summary

Extraction Given a List Page:
Nested Data Records
 We now deal with the most general case
 Nested data records
 Problem with the previous method
 not suitable for nested data records, i.e., data
records containing nested lists.
 Since the number of elements in the list of each
data record can be different, using a fixed
threshold to determine the similarity of data
records will not work.

Solution idea
 The problem, however, can be dealt with as follows.
 Instead of traversing the DOM tree top down, we can
traverse it post-order.
 This ensures that nested lists at lower levels are found first
based on repeated patterns before going to higher levels.
 When a nested list is found, its records are collapsed to
produce a single template.
 This template replaces the list of nested data records.
 When comparisons are made at a higher level, the
algorithm only sees the template. Thus it is treated
as a flat data record.

The NET algorithm

The MATCH algorithm
 It performs tree matching on child sub-trees of Node and
template generation. τ is the threshold for a match of two
trees to be considered sufficiently similar.

An example

GenNodeTemplate
 It generates a node template for all the
nodes (including their sub-trees) that match
ChildFirst.
 It first gets the set of matched nodes ChildRs
 then calls PartialTreeAlignment to produce a
template which is the final seed tree.
 Note: AlignAndLink aligns and links all
matched data items in ChildFirst and ChildR.

GenRecordPattern
 This function produces a regular expression
pattern for each data record.
 This is a grammar induction problem.
 Grammar induction in our context is to infer a
regular expression given a finite set of
positive and negative example strings.
 However, we only have a single positive example.
Fortunately, structured data in Web pages are
usually highly regular which enables heuristic
methods to generate “simple” regular expressions.
 We need to make some assumptions

Assumptions
 Three assumptions
 The nodes in the first data record at each level
must be complete.
 The first node of every data record at each level
must be present.
 Nodes within a flat data record (no nesting) do not
match one another.
 On the Web, these are not strong
assumptions. In fact, they work well in
practice.

Generating NFA

An example
 Line 1 simply produces a string for generating a
regular expression.
 The final NFA and the regular expression

Example (cont …)
 We finally obtain the following

Data extraction
 The function PutDataInTables (line 3 of NET)
outputs data items in a table, which is simple after
the data record templates are found.
 An example

An more complete example

Road map
 Introduction
 Summary

Extraction Given Multiple Pages
 We now discuss the second extraction problem
described in Section 8.3.1.
 Given multiple pages with the same encoding template, the
system finds patterns from them to extract data from other
similar pages.
 The collection of input pages can be a set of list pages or
detail pages.
 Below, we first see how the techniques described so
far can be applied in this setting, and then
 describe a technique specifically designed for this setting.

Using previous techniques
 Given a set of list pages
 The techniques described in previous sections are
for a single list page.
 They can clearly be used for multiple list pages.
 If multiple list pages are available, they may
help improve the extraction.
 For example, templates from all input pages may
be found separately and merged to produce a
single refined pattern.
 This can deal with the situation where a single
page may not contain the complete information.

Given a set of detail pages
 In some applications, one needs to extract data from
detail pages as they contain more information on the
object. Information in list pages are quite brief.
 For extraction, we can treat each detail page as a
data record, and extract using the algorithm
described in Section 8.7 and/or Section 8.8.
 For instance, to apply the NET algorithm, we simply create
a rooted tree as the input to NET as follows:
 create an artificial root node, and
 make the DOM tree of each page as a child sub-tree of the
artificial root node.

Difficulty with many detail pages
 Although a detail page focuses on a single object,
the page may contain a large amount of “noise”, at
the top, on the left and right and at the bottom.
 Finding a set of detail pages automatically is non-
trivial.
 List pages can be found automatically due to repeated
patterns in each page.
 Some domain heuristics may be used to find detail pages.
 We can find list pages and go to detail pages from there

An example page (a lot of noise)

The RoadRunner System
 Given a set of positive examples (multiple sample
pages). Each contains one or more data records.
 From these pages, generate a wrapper as a union-
free regular expression (i.e., no disjunction).
 Support nested data records.
The approach
 To start, a sample page is taken as the wrapper.
 The wrapper is then refined by solving mismatches
between the wrapper and each sample page, which
generalizes the wrapper.
 A mismatch occurs when some token in the sample does
not match the grammar of the wrapper.

Different types of mismatches and
wrapper generalization
 Text string mismatches: indicate data fields
(or items).
 Tag mismatches: indicate
 optional elements, or
 Iterators, list of repeated patterns
 Mismatch occurs at the beginning of a repeated pattern
and the end of the list.
 Find the last token of the mismatch position and identify
some candidate repeated patterns from the wrapper and
sample by searching forward.
 Compare the candidates with upward portion of the
sample to confirm.

Computation issues
 The match algorithm is exponential in the
input string length as it has to explore all
different alternatives.
 Heuristic pruning strategies are used to lower
the complexity.
 Limit the space to explore
 Limit backtracking
 Pattern (iterator or optional) cannot be delimited
on either side by an optional pattern (the
expressiveness is reduced).

Many other issues in data extraction
 Extraction from other pages.
 Disjunction or optional
 A set type or a tuple type
 Labeling and Integration
(Read the notes)

Road map
 Introduction
 Summary

Summary
Wrapper induction
 Advantages:
 Only the target data are extracted as the user can label
only data items that he/she is interested in.
 Due to manual labeling, there is no integration issue for
data extracted from multiple sites as the problem is solved
by the user.
 Disadvantages:
 It is not scalable to a large number of sites due to
significant manual efforts. Even finding the pages to label is
non-trivial.
 Wrapper maintenance (verification and repair) is very costly
if the sites change frequently.

Summary (cont …)
Automatic extraction
 Advantages:
 It is scalable to a huge number of sites due to the
automatic process.
 There is little maintenance cost.
 Disadvantages:
 It may extract a large amount of unwanted data because
the system does not know what is interesting to the user.
Domain heuristics or manual filtering may be needed to
remove unwanted data.
 Extracted data from multiple sites need integration, i.e.,
their schemas need to be matched.

Summary (cont…)
 In terms of extraction accuracy, it is reasonable to
assume that wrapper induction is more accurate
than automatic extraction. However, there is no
reported comparison.
 Applications
 Wrapper induction should be used in applications in which
the number of sites to be extracted and the number of
templates in these sites are not large.
 Automatic extraction is more suitable for large scale
extraction tasks which do not require accurate labeling or
integration.
 Still an active research area.

Cs511 data-extraction

More Related Content

What's hot (20)

Similar to Cs511 data-extraction (20)

More from Borseshweta (7)

Recently uploaded (20)

Cs511 data-extraction