Stance and Gender Detection in Spanish Tweets

Abstract

AbstractIn this paper, we present a deep learning based system for the user profiling and stance detection tasks in Twitter. Stance detection consists in automatically determining from text whether the author is in favor of a given target, against this target, or whether neither inference is likely. The proposed system assembles Convolutional Neural Networks and Long Short-Term Memory neural networks. We use this system to address, with minor changes, both problems. We explore embeddings and one-hot vectors at character level to select the best tweet representation.We test our approach in the Stance and Gender Detection in Tweets on Catalan Independence track proposed at IberEval 2017 workshop. With the proposed approach, we achieve state-of-the-art results for the Stance detection subtask and the best results published until now for the Gender detection subtask.KeywordsStance detectionGender detectionDeep learningTwitter