Datasets to train models for abusive language detection are both necessary and scarce. One reason for their limited availability is the cost of their creation. Manual annotation is expensive, and on top of it, the phenomenon itself is sparse, causing human annotators having to go through a large number of irrelevant examples in order to obtain some significant data. Strategies used until now to increase density of abusive language and obtain more meaningful data, include data filtering on the basis of pre-selected keywords and hate-rich sources of data. We suggest a recipe that at the same time can provide meaningful data with possibly higher density of abusive language and also reduce top-down biases imposed by corpus creators in the selection of the data to annotate. More specifically, we exploit the controversy channel on Reddit to obtain keywords that are used to filter a Twitter dataset. While the method needs further validation and refinement, our preliminary experiments show a higher density of abusive tweets in the filtered vj. unfiltered datasets, and a more meaningful topic distribution after filtering.
展开▼
