LIS at SemEval-2018 Task 2: Mixing Word Embeddings and Bag of Features for Multilingual Emoji Prediction

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 502–506 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics LIS at SemEval-2018 Task 2: Mixing Word Embeddings and Bag of Features for Multilingual Emoji Prediction Ga¨el Guibon LIS UMR 7020 Aix Marseille Universit´e CNRS Cal´ea Solutions gael.guibon@lis-lab.fr Magalie Ochs LIS UMR 7020 Aix Marseille Universit´e CNRS magalie.ochs@lis-lab.fr Patrice Bellot LIS UMR 7020 Aix Marseille Universit´e CNRS patrice.bellot@lis-lab.fr Abstract In this paper we present the system submit- ted to the SemEval2018 task2 : Multilingual Emoji Prediction. Our system approaches both languages as being equal by first; considering word embeddings associated to automatically computed features of different types, then by applying bagging algorithm RandomForest to predict the emoji of a tweet. 1 Introduction Emojis were first used to emphasize conversations before becoming representations of specific emo- tions, objects or ideas. They are now used in al- most every social medium and conversation de- vices, such as messenging applications or even emails1. Tweets and their emoticons were used as la- bels to predict polarity at first (Pak and Paroubek, 2010). However, emojis are not used the same way as emoticons in messaging applications. They can convey further information, even more when com- bined. The advantage of emojis is that they are becoming more standardized, even though exist- ing emojis are still growing quickly2. This is why emoji prediction is a relatively new task. It can be considered as a composite task mixing emotion prediction for face emojis, aspect/subject detec- tion for object emojis, and other metadata predic- tion for more abstract emojis, representing ideas for instance. This year, SemEval started the first emoji pre- diction task (Barbieri et al., 2018). It consists of a multiclass classification task for a total of 20 pos- sible classes, i.e. emojis. This task is interesting in several ways. Firstly, it is a relatively new task that only a few studies did focus on. Secondly, it is 1http://cdn.emogi.com/docs/reports/ 2015_emoji_report.pdf 2https://goo.gl/jbeRYW quite important not only for research, but also for companies willing to embrace the current trend of social network and interaction analysis. Both are important topics for Natural Language Processing (NLP) and Information Retrieval (IR). Our system obtained good results (63.65% f1- score) while using the trial dataset, and lower re- sults (13.53% f1-score) on the test dataset. Be- cause this pattern occurred for both English and Spanish, and for all participants, we try to explain it. The paper is organized as follows: we first sum- marize the existing work related to this task and to our approach (Section 2). Then we present what we identified as the most challenging areas from this task and the dataset used (Section 3). We go on by describing our system (Section 4) and de- tailing the pre processing and prediction steps. Fi- nally, we conclude by discussing the performance limits and show the benefits of our participation in this task (Section 5). 2 Related Work Several research studies focus on emoji prediction. Most of them use word …

Content truncated to 500 words