Peperomia at SemEval-2018 Task 2: Vector Similarity Based Approach for Emoji Prediction

Proceedings of the 12th International Workshop on Semantic Evaluation (SemEval-2018), pages 428–432 New Orleans, Louisiana, June 5–6, 2018. ©2018 Association for Computational Linguistics Peperomia at SemEval-2018 Task 2: Vector Similarity Based Approach for Emoji Prediction Jing Chen , Dechuan Yang , Xilian Li , Wei Chen ∗ and Tengjiao Wang Key Lab of High Confidence Software Technologies (MOE), School of EECS, Peking University, Beijing, 100871, China {chenjing.amy, yangdechuan, xilianli, pekingchenwei, tjwang}@pku.edu.cn Abstract This paper describes our participation in Se- mEval 2018 Task 2: Multilingual Emoji Pre- diction, in which participants are asked to pre- dict a tweet’s most associated emoji from 20 emojis. Instead of regarding it as a 20-class classification problem we regard it as a text similarity problem. We propose a vector sim- ilarity based approach for this task. First the distributed representation (tweet vector) for each tweet is generated, then the similarity be- tween this tweet vector and each emoji’s em- bedding is evaluated. The most similar emoji is chosen as the predicted label. Experimental results show that our approach performs com- parably with the classification approach and shows its advantage in classifying emojis with similar semantic meaning. 1 Introduction Participants for SemEval 2018 Task 2 (Barbieri et al.) are asked to predict the most likely asso- ciated emoji given the tweet. For simplicity pur- poses, each tweet contains one and only one emoji, which belongs to the 20 most frequent emojis. We participate in its subtask 1: Emoji Prediction in English. With the wide-spread use on many social plat- forms, emoji has attracted more and more atten- tion of researchers recently. Miller et al. (2016) explored whether emoji renderings or differences across platforms gave rise to diverse interpreta- tions of emoji. For the same emoji, the sender and the receiver may have different interpreta- tions of its meaning. This misinterpretation oc- curs when joint perceptual experience of sender and receiver lacks or the platforms’ rendering style differs. Some efforts have been devoted to study- ing emoji through its distributed representation. ∗corresponding author. Barbieri et al. (2016a,b) trained emoji embed- dings with a skip-gram model through millions of tweets, and explored the similarity and relatedness among these embeddings in various languages. Their results suggested that the overall seman- tic of emoji was preserved across languages, but some emojis were interpretated differently due to users’ socio-geographical differences. Eisner et al. (2016) trained emoji embeddings with their short descriptions and demonstrated that emoji embed- dings trained through this way were beneficial to sentiment analysis task. We believe that the key to better classify emo- jis is understanding their meaning, since people intend a particular meaning when they send an emoji. People view the same characters during the exchange of plain text. Unlike plain text, emoji is not definite enough and doesn’t have a gen- eral acknowledgement of how we should use it. It is common for different readers to have differ- ent interpretations of the same emoji, which nat- urally results in different ways of using emoji. Na’aman et …

Content truncated to 500 words