Improving the classification of events in tweets using semantic enrichment

Contextual enrichment using external sources has been proposed as a means to deal with the poor textual contents of tweets for event classification. Related work performs contextual enrichment according to specific assumptions about the events. Furthermore, enrichment adds a significant amount of extra features, most of them with no discriminative contribution to the event classification task. In this paper, we propose an enrichment framework targeted at the classification of events in general, of which the key elements are: a) external enrichment using related web pages for extending the conceptual features contained within the tweets; b) semantic enrichment using the DBpedia to add related semantic features, and c) a pruning technique that selects the semantic features with discriminative potential. We compared the proposed approach against two distinct baselines based on textual features only and word embeddings, using seven different event datasets. Our experiments reveal that the proposed framework supports the classification of distinct event types, outperforming the textual baseline in 63.5% of the cases, and the word embeddings baseline in 96.5% of the cases.