In this session we will be exploring sentiment classification, a natural language processing task with several applications! The goal is to discuss the steps when building a system that accurately classifies the sentiment conveyed by tweets :-)
Sentiment classification is the task of predicting the sentiment polarity of a written text passage.
Being able to automatically analyze the sentiment conveyed by such text passages
is important for various applications, for instance when assessing consumer satisfaction. Microblogging platforms like Twitter are nowadays ubiquitous and enable users to express their sentiment
on different subjects. While they are a rich source of user-generated content,
there are major challengs in analyzing their content. For instance, tweets are
very short messages where punctuation and language are used in very creative ways.
In this talk, I will present a machine learning approach for predicting the sentiment of tweets. After introducing the task and the
difficulties it poses, I will describe the steps taken for building a state-of-the-art sentiment classifier for tweets.
I will conclude the talk with empirical evaluation of the proposed method and with a discussion on the challenges and limitations of sentiment analysis.
Georgios Balikas obtained his PhD in the field of machine learning from the University of Grenoble-Alps. His research interests span the fields of machine learning, natural language processing and information retrieval. He is currently a data scientist at Kelkoo applying machine learning for product recommendations and online advertising. He is also one of the organisers of the Grenoble Data Science meetup!