Grasping other person’s emotions or sentiments through texts is challenging even for humans sometimes. I bet you also have an experience that once you were reading someone’s message multiple times to figure out, “Is he/she upset? Or am I too sensitive?” Despite the difficulties, figuring out emotion or sentiment in texts is crucial nowadays, due to our daily habit of using emails, mobile messengers, social media, chats. An interesting question is, “can we teach machines to understand sentiment and emotions inside our texts?”
We introduce how to build a sentiment/emotion analysis system by training deep learning models with a huge amount of tweets. We exploit the emotional knowledge inside those texts by using emojis and hashtags as weakly supevised label. Our system ranked Top 3 in SemEval 2018: Affect in Tweets, a well-known competition in the NLP research community.
paper: https://arxiv.org/abs/1804.08280
Buzzwords: Natural Language Processing, Sentiment Analysis, Representation Learning
Level: Intermediate: Target audiences with intermediate experience in python programming
Requirements to Audiences: Nil
Language: English
Speaker: Ji Ho Park (South Korea)
Speaker Bio: Ji Ho is from Korea but lived in HK for almost 6 years. He graduated from Hong Kong University of Science and Technology (HKUST) with a Master of Philosophy(MPhil). His research area is in natural language processing and machine learning, especially on sentiment/emotion analysis and automatic abusive language detection. Now he is currently working at Oyalabs, a HK baby-tech startup as a machine learning engineer (NLP).
GitHub: github.com/jihopark
LinkedIn: https://www.linkedin.com/in/parkjiho/
Website/Blog: medium.com/@parkjiho