A Systematic Review of Emoji: Current Research and Future Perspectives

.
Abstract
A growing body of research explores emoji, which are visual symbols in computer mediated communication (CMC). In the 20 years since the first set of emoji was released, research on it has been on the increase, albeit in a variety of directions. We reviewed the extant body of research on emoji and noted the development, usage, function, and application of emoji. In this review article, we provide a systematic review of the extant body of work on emoji, reviewing how they have developed, how they are used differently, what functions they have and what research has been conducted on them in different domains. Furthermore, we summarize directions for future research on this topic.
.

.

Computer Science

Research in the field has focused on using emoji for emotional analysis of UGC data, the conversion of emoji to other expression modality, and using emoji for optimizing computer systems.

UGC=User-Generated Content.

Sentiment Analysis

With the significant growth of UGC data on the Internet, sentiment analysis which aims at changing this data into valuable asset for decision making, has become increasingly important (Al-Azani et al., 2018). As emoji are widely used in expressing emotions, they have become an effective means of sentiment analysis (Hogenboom et al., 2013; Cappallo et al., 2015). A number of studies have confirmed the effective performance of emoji in sentiment analysis (Sari et al., 2014; Cahyaningtyas et al., 2017; Felbo et al., 2017; LeCompte and Chen, 2017). Besides, emoji-based sentiment analysis is language-independent and exhibits cross-language validity (Guthier et al., 2017), for example, Al-Azani et al. (2018) found that emoji can also be used in analyzing the sentiment of Arabic tweets. However, other studies have shown that using emoji in sentiment analysis leads to higher emotional scores, and that this effect is more pronounced in positive comments (Ayvaz and Shiha, 2017).

Many studies have provided algorithms and models for emoji-based sentiment analysis, which mainly uses two kinds of techniques, sentiment lexicon, and machine learning.

The sentiment lexicon method

The sentiment lexicon approach focuses on building an emoji emotional lexicon to support text sentiment analysis. By human annotating, Petra et al. (2015) has classified 751 commonly used emoji and built an emoji lexicon based on the positivity of emoji. But because there are so many emoji, some researchers have come up with ways to build emoji dictionaries automatically.

Jiang et al. (2015) proposed an emoticon space model to automatically match emotional tags for emoji. Kimura and Katsurai (2017) assigned multi-dimensional emotional vectors to emoji by calculating the co-occurrence frequency of emoji and emotional words in WordNet-Affect. Aoki and Uchida (2011) have also automatically generated emoji vectors based on the relationship between emotional words and emoji. By using the Word2Vec clustering method, Mayank et al. (2016) divided emoji into clusters which represent different human emotions.

The machine learning method

The machine learning method refers to train sentiment classifiers based on a corpus in order to analyze the sentiments of text (Wang et al., 2012). Machine learning can be divided into supervised learning and unsupervised learning. They are different in that the former needs a human annotated corpus while the latter doesn’t.

The effectiveness of using emoji as a way of training classifiers has been proven (Hallsmar and Palm, 2016) and furthermore it has been shown that emoji outperform emoticons (Redmond et al., 2017). An example of supervised learning is the emoticon smoothed language model (ESLAM) proposed by Liu et al. (2012), which classifies twitter based on a model trained by a human annotated corpus.

A lot of research has focused on unsupervised learning (Li et al., 2018), and constructed sentiment analysis models trained automatically using emoji data sets. Chen Y. et al. (2018) trained sentiment classifiers by via bi-sense emoji embedding and attention-based long short-term memory network (LSTM) in order to analyze the sentiment of messages on Twitter. Wang et al. (2016) designed a hybrid sentimental entity recognition model (HSERM), which classifies emoji into four different emotional categories, and then categorizes the emotional data based on the model. Some research has focused on the ironic features of emoji and developed an irony detection model for emoji in order to improve the accuracy of sentiment analysis of tweets (Reyes et al., 2013; Prasad et al., 2017; Singh et al., 2019).

Modality Transitions

The visual features and Unicode basis of emoji make them anindependent expressive modality that is different from text andpictures (Cappallo et al., 2019). A lot of research focuses onconversion between emoji and other modalities such as text,picture and video.

For example, Emoji2Video offers a way to search for videosusing emoji (Cappallo et al., 2015). Later research has focusedon the shift from other modalities to emoji. Because of thecorrelation between emotional categories in text and users’ emojiselections, Hayati and Muis (2019) and Zanzotto and Santilli(2018) proposed two different ways to predict emoji based ontext. Kim et al. (2019) developed Reeboc, which can analyze chatcontent, extract different emotions or topics, and then, based onthis, recommend emoji to users. The practice of text-based emojiprediction has also been validated in other languages, such asHebrew (Liebeskind et al., 2019).

System Optimization

Emoji have played a role in improving the performance of computer hardware and software. For example, emoji can be used to achieve diverse in-car interaction design. In order to optimize the functions of the central rear-view mirror, researchers suggest that passengers emotions can be fed back to the driver through emoji and other elements, which can enhance mutual understanding between driver and back-seat passenger (Chao et al., 2019).

Furthermore, emoji can also be applied in the area of password security. Kraus et al. (2017) came up with the EmojiAuth project, exploring how the use of emoji affects the availability of mobile authentication and user experience by adding emoji into passwords. Compared with the Standard PIN (Personal Identification Number) input, a password containing emoji is easier to remember and, thus, emoji-based authentication is a practical alternative to traditional PIN authentication.
.
https://www.researchgate.net/publication/336561000_A_Systematic_Review_of_Emoji_Current_Research_and_Future_Perspectives