Date of Award

2016

Publication Type

Master Thesis

Degree Name

M.Sc.

Department

Computer Science

Keywords

Automatic Web Data Extraction; Data integration; Web Content Mining; Wrappers

Supervisor

Ezeife, Christie

Rights

info:eu-repo/semantics/openAccess

Abstract

Discovering potentially useful and previously unknown historical knowledge from heterogeneous E-Commerce (B2C) web site contents to answer comparative queries such as “list all laptop prices from Walmart and Staples between 2013 and 2015 including make, type, screen size, CPU power, year of make”, would require the difficult task of finding the schema of web documents from different web pages, extracting target information and performing web content data integration, building their virtual or physical data warehouse and mining from it. Automatic data extractors (wrappers) such as the WebOMiner system use data extraction techniques based on parsing the web page html source code into a document object model (DOM) tree, traversing the DOM for pattern discovery to recognize and extract different web data types (e.g., text, image, links, and lists). Some limitations of the existing systems include using complicated matching techniques such as tree matching, non-deterministic finite state automata (NFA), domain ontology and inability to answer complex comparative historical and derived queries. This thesis proposes building the WebOMiner_S which uses web structure and content mining approaches on the DOM¬ tree html code to simplify and make more easily extendable the WebOMiner system data extraction process. We propose to replace the use of NFA in the WebOMiner with a frequent structure finder algorithm which uses regular expression matching in Java XPATH parser with its methods to dynamically discover the most frequent structure (which is the most frequently repeated blocks in the html code represented as tags < div class = “ ′′ >) in the Dom tree. This approach eliminates the need for any supervised training or updating the wrapper for each new B2C web page making the approach simpler, more easily extendable and automated. Experiments show that the WebOMiner_S achieves a 100% precision and 100% recall in identifying the product records, 95.55% precision and 100% recall in identifying the data columns.

Download

Share

COinS