Learning Emoji Embeddings using Emoji Co-occurrence Network Graph

Abstract

Usage of emoji in social media platforms has seen a rapid increase over the last few years. Majority of the social media posts are laden with emoji and users often use more than one emoji in a single social media post to express their emotions and to emphasize certain words in a message. Utilizing the emoji cooccurrence can be helpful to understand how emoji are used in social media posts and their meanings in the context of social media posts. In this paper, we investigate whether emoji cooccurrences can be used as a feature to learn emoji embeddings which can be used in many downstream applications such sentiment analysis and emotion identification in social media text. We utilize 147 million tweets which have emojis in them and build an emoji cooccurrence network. Then, we train a network embedding model to embed emojis into a low dimensional vector space. We evaluate our embeddings using sentiment analysis and emoji similarity experiments, and experimental results show that our embeddings outperform the current state-of-the-art results for sentiment analysis tasks.

Emojis as Contextual Indicants in Location-Based Social Media Posts

“…Accordingly, they are utilized for research in various fields. Due to their function as emotional signals [6,7] adding a sentimental information to a written text, sentiment analysis or emotion identification is one application field within computer science [7,9,[12][13][14][15][16][17], which, among other things, led to the result that the utilization of emojis helps to improve sentiment scores [18,19]. Other research questions related to emojis are of semantic origin and explore their meaning and use [14,[20][21][22].…”

Multimodal Emotion Classification

“…Researchers have started using Emojinet [31] to learn embeddings for sentiment analysis task and have achieved better accuracies than the previous state of the art emoji embeddings [32]. There have been many approaches which use emoji as a feature to classify sentiment on social media using emoji as a feature, Illendula et al [13] has used emoji co-occurrence has a feature to learn sentiment features of emojis. Also, emojis have also been a very important feature to classify emotional content; previous research has always manually specified which emotional category each hashtag or emoji belongs to [22].…”

“…There have been many word embedding models to learn word representations including GloVe, skip-gram and CBOW. Similarly, there are different approaches to learn emoji embeddings that use co-occurrence feature [13], skip-gram word embedding architecture [4], and semantic knowledge of emojis from EmojiNet [32]. Since EmojiNet gives us access to the emoji sense forms, we make use of the embedding model developed by Wijeratne et al [32] to learn emoji embeddings.…”

“…This further helps in reducing or eliminating the chances "Sad" emotional category in our dataset of misunderstanding, often associated with plain text messages 2 . Recent study proved that co-occurrence helps users to express their sentiment more effectively 3 [13].…”

Which Emoji Talks Best for My Picture?

“…Recent research by Seyednezhad et al [39] and Fede et al [13] has shown that emoji cooccurrence is one of the important features which helps us to understand the context of use of multiple emojis. Illendula et al [22] have worked on learning emoji representations using emoji co-occurrence network graph and state-of-the-art network embedding model, these embeddings out-performed the previous state-of-the-art accuracies for sentiment analysis task.…”