Making Your Tweets More Fancy: Emoji Insertion to Texts

Proceedings of Recent Advances in Natural Language Processing, pages 770–779 Sep 1–3, 2021. https://doi.org/10.26615/978-954-452-072-4_088 770 Making Your Tweets More Fancy: Emoji Insertion to Texts Jingun Kwon1, Naoki Kobayashi1, Hidetaka Kamigaito1, Hiroya Takamura2 and Manabu Okumura1 1Tokyo Institute of Technology 2National Institute of Advanced Industrial Science and Technology (AIST) kwon.j.ad@m.titech.ac.jp {kobayasi,kamigaito,oku}@lr.pi.titech.ac.jp {takamura.hiroya}@aist.go.jp Abstract In the social media, users frequently use small images called emojis in their posts. Although using emojis in texts plays a key role in re- cent communication systems, less attention has been paid on their positions in the given texts, despite that users carefully choose and put an emoji that matches their post. Exploring positions of emojis in texts will enhance un- derstanding of the relationship between emo- jis and texts. We extend an emoji label pre- diction task taking into account the informa- tion of emoji positions, by jointly learning the emoji position in a tweet to predict the emoji label. The results demonstrate that the position of emojis in texts is a good clue to boost the performance of emoji label prediction. Human evaluation validates that there exists a suitable emoji position in a tweet, and our proposed task is able to make tweets more fancy and natural. In addition, considering emoji posi- tion can further improve the performance for the irony detection task compared to the emoji label prediction. We also report the experi- mental results for the modified dataset, due to the problem of the original dataset for the first shared task to predict an emoji label in Se- mEval2018. 1 Introduction The advent of emojis has dramatically changed the style of human communication. Currently, emojis are widely used on social media platforms such as Twitter, Facebook, and Instagram. Owing to the prevalent use of emojis in the last few years, they become a target of recent researches. As reported by Kralj Novak et al. (2015), ex- ploring the interplay between emojis and texts can be a clue for better natural language understanding (NLU) in social media datasets with full of emojis. In response to this report, Barbieri et al. (2017) proposed a task to predict an emoji label given a Ex.1: was used instead of the word “tree”. Input : Don’t forget your Christmas TAN! Previous task : Proposed task : Don’t forget your Christmas TAN! Ex.2: was used instead of the character “A”. Input : @AIIAmericanGirI: U.S. judge sworn into office on Quran disgusting! Previous task : Proposed task : @AIIAmericanGirI: U.S. judge sworn into office on Quran disgusting! Ex.3: Two sentences have different emotions. Input : Grateful for some girl time yesterday with my favorites. Sad we won’t all be together again... Previous task : Proposed task : Grateful for some girl time yesterday with my favorites. Sad we won’t all be together again... Figure 1: Example tweets from the training dataset of SemEval2018 task2. Our goal is to insert a suitable emoji to a suitable position in a given tweet. text, which can contribute to various natural lan- guage processing …

Content truncated to 500 words