Data Augmentation for Robust Character Detection in Fantasy Novels
This work addresses character detection in fantasy novels, an incremental improvement for NLP applications in literary analysis.
The paper tackled the problem of false negatives in Named Entity Recognition for character detection in fantasy novels by applying a straightforward data augmentation technique, resulting in higher recall at the cost of some precision, which was mitigated by adding more local context.
Named Entity Recognition (NER) is a low-level task often used as a foundation for solving higher level NLP problems. In the context of character detection in novels, NER false negatives can be an issue as they possibly imply missing certain characters or relationships completely. In this article, we demonstrate that applying a straightforward data augmentation technique allows training a model achieving higher recall, at the cost of a certain amount of precision regarding ambiguous entities. We show that this decrease in precision can be mitigated by giving the model more local context, which resolves some of the ambiguities.