Poster Submissions


Cross-Media Knowledge Aquisition in X-Media Project

Joao Magalhaes, Jose' Iria, Lei Xi, Mark Greenwood, Fabio Ciravegna

Humans can easily understand information represented in different media. Articles, books and other formatted documents fit our mental process of acquiring knowledge: split information into meaningful parts and follow a logical flow or structure of ideas. We can understand an idea if it is written in scrap paper, in a formal letter, in an image, or even if the image is used to disambiguate the text. Unfortunately information processing systems are not comparable to humans in this task: information extraction from multiple document formats is usually a nightmare addressed by a vast number different libraries; information extraction have always focused on single-media data and ignore the structure of a document. In this context, the objective of the proposed Cross-Media Knowledge Acquisition framework addresses the following issues: (1) process information from different formats and media-types; (2) design cross-media features extractors that explore the rich structure of documents; (3) algorithms capable of exploiting heterogenous sources of data; (4) use background knowledge to improve information extraction.

Hosted by the Knowledge Media Institute