PDF-Inspired Visual Data Extraction for HTML
Abstract
Background: The work will be done in the context of the large ERC project DIADEM: Domain-centric Intelligent Automated Data Extraction Methodology whose goal is to automate web data extraction in specific application domains such as real estate, restaurants, and so on.
Principal goal of the MSc or Honour School project:
In PDF data extraction, several techniques for data extraction has been proposed mainly based on visual aspects of a document
such as layout, text alignment, bitmap information. However, very few of them has been applied to HTML documents, mainly because
the difference between html and pdf.
This proposal aims at (i) studying the state-of-the-art pdf data extraction techniques
based on visual information, and (ii) applying the most promising ones to HTML documents.
Also, computer vision approaches
could be investigated.
Skills Needed: This project requires good theoretical, analytic and software engineering skills.Also, good knowledge of Java is essential.
Supervision: This project will be co-supervised by Dr. Giorgio Orsi.