Assessing Emoji Use in Modern Text Processing Tools

Assessing Emoji Use in Modern Text Processing Tools Abu Awal Md Shoeb Dept. of Computer Science Rutgers University New Brunswick, NJ, USA abu.shoeb@rutgers.edu Gerard de Melo Hasso Plattner Institute / University of Potsdam Potsdam, Germany gdm@demelo.org Abstract Emojis have become ubiquitous in digital com- munication, due to their visual appeal as well as their ability to vividly convey human emo- tion, among other factors. This also leads to an increased need for systems and tools to op- erate on text containing emojis. In this study, we assess this support by considering test sets of tweets with emojis, based on which we per- form a series of experiments investigating the ability of prominent NLP and text processing tools to adequately process them. In particu- lar, we consider tokenization, part-of-speech tagging, dependency parsing, as well as senti- ment analysis. Our findings show that many systems still have notable shortcomings when operating on text containing emojis. 1 Introduction In our modern digital era, interpersonal communi- cation often takes place via online channels such as instant messaging, email, social media, etc. This entails an increasing need for tools that operate on the resulting digital data. For instance, online con- versations can be invaluable sources of insights that reveal fine-grained consumer preferences with re- gard to products, services, or businesses (Dong and de Melo, 2018). However, the shifts in modality and medium also shape the way we express ourselves, making it in- creasingly natural for us to embed emojis, images, hashtags into our conversations. In this paper, we focus specifically on emojis, which have recently become fairly ubiquitous in digital communication, with a 2017 study reporting 5 billion emojis be- ing sent daily just on Facebook Messenger (Burge, 2017). Emojis are textual elements that are encoded as characters but rendered as small digital images or icons that can be used to express an idea or emotion. Goals. Due to their increasing prominence, there is a growing need to properly handle emojis when- ever one deals with text. We consider a set of popu- lar NLP tools and empirically assess to what extent they support emojis across a set of standard tasks, encompassing tokenization, part-of-speech tagging, dependency parsing, and sentiment analysis. Although emojis can be encoded as Unicode characters, there are unique properties of emoji en- coding that merit special consideration, such as skin tone modifiers and composite emoji incorporating multiple basic emojis. Moreover, text harboring emojis may adhere to subtly different conventions than more traditional forms of text, e.g., with regard to token and sentence boundaries. Emojis can take the place of words with different parts-of-speech and assume different grammatical roles. Finally, emojis may of course also alter the semantics of the text, which in turn may, for instance, affect its sentiment polarity. Overview. For our analysis, we draw primarily on social media and study diverse forms of emoji use. We run a series of experiments on such data evaluating each NLP tool to observe its behaviour at different stages in the processing pipeline. The …

Content truncated to 500 words