2018/04/18 by Christos Baziotis, Baziotis, Christos, Nikos Athanasiou +9
Computer Science · #Digital Communication and Language #Natural Language Processing Techniques #Sentiment Analysis and Opinion Mining
paper · pdf · doi:10.48550/arxiv.1804.06657
In this paper we present a deep-learning model that competed at SemEval-2018\nTask 2 "Multilingual Emoji Prediction". We participated in subtask A, in which\nwe are called to predict the most likely associated emoji in English tweets.\nThe proposed architecture relies on a Long Short-Term Memory network, augmented\nwith an attention mechanism, that conditions the weight of each word, on a\n"context vector" which is taken as the aggregation of a tweet's meaning.\nMoreover, we initialize the embedding layer of our model, with word2vec word\nembeddings, pretrained on a dataset of 550 million English tweets. Finally, our\nmodel does not rely on hand-crafted features or lexicons and is trained\nend-to-end with back-propagation. We ranked 2nd out of 48 teams.\n