Table Structure Understanding by Sibling Page Comparison

1
Table Structure Understanding by Sibling Page
Comparison
Cui Tao Data Extraction Group Department of
Computer Science Brigham Young University
Supported by NSF
2
Table Structure Understanding

• Motivation
• Many documents contain tables
• Data extraction
• Data integration
• Ontology evolution
• Solution
• Locate tables
• Locate table labels
• Locate table values
• Find label/value associations

3
Table Structure Understanding
4
Table Structure Understanding
2
(Gene Model, 1) F18H3.5a (Gene Model, 2)
F18H3.5b
5
(No Transcript)
6
(No Transcript)
7
Sibling Pages

• Generated output pages
• user query
• results in predefined page structure
• Same web site same structure

8
Problems

• Data rich area --- discard the irrelevant parts
• Find table correspondences
• Find mappings between table cells
• Find structure patterns

9
HTML Table Components
10
Data Rich Area
11
Table Unnesting
12
DOM Tree
13
Simple Tree Matching

• Simple Tree Matching (STM) Yang91
• Maximum matching pairs of nodes
• O(mn)

14
Table Structure Pattern
15
Table Structure Pattern
16
Experimental Results

• Initial Test
• General pattern extraction
• Molecular biology 95.6
• Car ad 100
• Dynamic adjustment
• Unseen structure
• Structure variations