Discovery of Weather Forecast Web Resources Based on Ontology and Content-Driven Hierarchical Classification

Monitoring of environmental information is critical both for the evolvement of important environmental events, as well as for everyday life activities. In this work, we focus on the discovery of web resources that provide weather forecasts. To this end we submit domain-specific queries to a general purpose search engine and post process the results by introducing a hierarchical two layer classification scheme. The top layer includes two classification models: a) the first is trained using ontology concepts as textual features; b) the second is trained using textual features that are learned from a training corpus. The bottom layer includes a hybrid classifier that combines the results of the top layer. We evaluate the proposed technique by discovering weather forecast websites for cities of Finland and compare the results with previous works.