Ensemble Learning for Keyword Extraction

Abstract

Nowadays, the most relevant events occurring in the city are advertised on-line, generally through small textual descriptions. The exponential growth of the Web often hampers the task of finding relevant information, turning the existence of good information extraction and summarization methods in a necessity. As such, the main goal of this dissertation is to develop an ensemble learning application for automatically extracting keywords from those event textual descriptions, since using human indexers is slow and expensive. Through rich information on events, one should be able to understand its mobility implications and possibly correlate both, allowing to foreseeing eventual repercussions that a specific event may cause in the city’s normal behavior. The proposed application intends to apply Supervised Machine Learning approaches, namely from known automatic keyword extraction systems, retrieving a set of keywords as output from the event descriptions usually found in theWeb