Exploiting the language of moderated sources for cross-classification of user generated content

authored by

Avaré Stewart, Wolfgang Nejdl

Abstract

Recent pandemics such as Swine Flu have caused concern for public health officials. Given the ever increasing pace at which infectious diseases can spread globally, officials must be prepared to react sooner and with greater epidemic intelligence gathering capabilities. However, state-of-the-art systems for Epidemic Intelligence have not kept the pace with the growing need for more robust public health event detection. Existing systems are limited in that they rely on template-driven approaches to extract information about public health events from human language text. In this paper, we propose a new approach to support Epidemic Intelligence. We tackle the problem of detecting relevant information from unstructured text from a statistical pattern recognition viewpoint. In doing so, we also address the problems associated with the noisy and dynamic nature of blogs by exploiting the language in moderated sources, to train a classifier for detecting victim reporting sentences in blog social media. We refer to this as Cross-Classification. Our experiments show that without using manually labeled data, and with a simple set of features, we are able to achieve a precision as high as 88% and an accuracy of 77%, comparable with the state-of-the-art approaches for the same task.

Organisation(s)

L3S Research Centre

Type

Conference contribution

Pages

571-576

No. of pages

6

Publication date

14.09.2011

Publication status

Published

Peer reviewed

Yes

ASJC Scopus subject areas

Computer Networks and Communications, Information Systems

Sustainable Development Goals

SDG 3 - Good Health and Well-being