Automatic Extraction of Emotive and Non-emotive Sentence Patterns
Michal E. Ptaszynski, Fumito Masui, Rafał Rzepka, Kenji Araki · 2014
In this research we focus on automatic extraction of patterns from emotive (emotionally loaded) sentences. We assume emotive sentences stand out both lexically and grammatically and verify this assumption experimentally by comparing two sets of such sentences. We use a novel pattern extraction method based on the idea of language combinatorics. Extracted patterns are applied in a text classification task of discriminating between emotive and non-emotive sentences. The method reached balanced F-score of 76% with Precision equal to 64% and Recall 93%.