The Fourth International Workshop on “Data Extraction and Object Search” (DEOS 2013) will take place as a satellite event of WWW 2014 in Seoul, South Korea, on April 7th, 2014. Web data extraction is witnessing a renaissance. In an increasing number of applications such as price intelligence or predictive analytics, the value of data-driven approaches has been conclusively proven. However, the necessary data is often available only as HTML, e.g., in form of online shops of competitors that can serve as sources for pricing and offer data. DEOS is a regular forum for researchers and practitioners in data extraction and object search, to present and discuss ongoing work on data extraction and object search for products, events, reviews, and other types of structured data on the web.
This year’s DEOS focuses on the challenges in scaling data extraction to the variety and volume of different data sources available only as HTML on the web. Classical data extraction has been largely site-specific, requiring some manual supervision for every site. Where data is to be sourced from more than a handful of websites, this approach fails. To address this challenge, we are witnessing a paradigm shift in data extraction away from manual supervision by experts.
This shift has seen two primary directions emerge: Some approaches have considered how to allow non-experts to provide the necessary per-site supervision and turned to crowdsourcing. Some approaches employ automatic entity extraction to replace human annotation of data to be extracted and techniques to deal with the noise in such automatic annotations. Either direction poses major challenges and changes to existing data extraction technology. In this workshop, we bring together researchers from both directions.
Source:http://diadem.cs.ox.ac.uk/deos14/
This year’s DEOS focuses on the challenges in scaling data extraction to the variety and volume of different data sources available only as HTML on the web. Classical data extraction has been largely site-specific, requiring some manual supervision for every site. Where data is to be sourced from more than a handful of websites, this approach fails. To address this challenge, we are witnessing a paradigm shift in data extraction away from manual supervision by experts.
This shift has seen two primary directions emerge: Some approaches have considered how to allow non-experts to provide the necessary per-site supervision and turned to crowdsourcing. Some approaches employ automatic entity extraction to replace human annotation of data to be extracted and techniques to deal with the noise in such automatic annotations. Either direction poses major challenges and changes to existing data extraction technology. In this workshop, we bring together researchers from both directions.
Source:http://diadem.cs.ox.ac.uk/deos14/
No comments:
Post a Comment