CROSSMARC: Cross-lingual multi-agent retail
The aim of CROSSMARC was to develop next-generation technology for electronic-retail product comparison, drawing on techniques from language engineering, machine learning and user modelling. I participated in the CROSSMARC project as a research associate from May 2002 to August 2003. My responsibility was to design and implement a cross-lingual and cross-thematic-domain named-entity recognition engine (NERC), based on various machine-learning techniques. The produced NERC engine has been successfully tested on all four languages (English, French, Greek and Italian), and three thematic domains (laptop advertisements, job offers and real-estate advertisements) of the project. The adaptation of the NERC engine to new languages/thematic domains was fully automated. The NERC engine was developed inside the Ellogon Language Engineering Platform, while some screenshots of the NERC engine can be found below.
Proceed to the official CROSSMARC web site.