The availability of natural language content online has swelled in recent years which has increased the interest of many research fields as well as industry in using these services as data sources. Traditional classification methods have trouble dealing with the online nature of news streams and ensemble methods presented to date mainly focus on building model ensembles over time. I will present motivation and research results for building classifier ensembles that are distributed across topical distribution of the stream as opposed to a temporal one.
By: Luke Scott
Last updated: Friday, 23 January 2015