Title: Information extraction from the web by matching visual presentation patterns
Authors: Minárik, Matej
Burget, Radek
Citation: STEINBERGER, Josef ed.; ZÍMA, Martin ed.; FIALA, Dalibor ed.; DOSTAL, Martin ed.; NYKL, Michal ed. Data a znalosti 2017: sborník konference, Plzeň, Hotel Angelo 5. - 6. října 2017. 1. vyd. Plzeň: Západočeská univerzita v Plzni, 2017, s. 227-231. ISBN 978-80-261-0720-0.
Issue Date: 2017
Publisher: Západočeská univerzita v Plzni
Document type: konferenční příspěvek
URI: https://www.zcu.cz/export/sites/zcu/pracoviste/vyd/online/DataAZnalosti2017.pdf
ISBN: 978-80-261-0720-0
Keywords: integrace webových dat;extrakce informací;strukturovaná extrakce záznamů;segmentace stránek;klasifikace obsahu;mapování ontologií
Keywords in different language: web data integration;information extraction;structured record extraction;page segmentation;content classification;ontology mapping
Abstract in different language: There is a large amount of data available on the Web. Data are often represented as text, enriched with tables, lists, images or other visual structures. These data are usually coded in HTML without any additional semantics, which makes them nigh impossible to automatically process and extract. There are ap-proaches based on top-down document segmentation according to visual infor-mation and layout. We present a bottom-up approach which starts with the smallest consistent elements and matches the visual relationships among these elements to a pre-defined ontological structure of extracted records. This meth-od considers not only the visual attributes of a particular segment, but also its position amongst other segments.
Rights: © Západočeská univerzita v Plzni
Appears in Collections:Data a znalosti 2017
Data a znalosti 2017

Files in This Item:
File Description SizeFormat 
Minarik.pdfPlný text376,7 kBAdobe PDFView/Open

Please use this identifier to cite or link to this item: http://hdl.handle.net/11025/26368

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.