Hosted by the Digital Footprints Lab, University of Bristol

Abstract: Extracting information from free text sources is a complex and exciting data science challenge. Textual data generated by humans is rich, complex, with its meaning often contextual and packed with nuance. This project analysed data from Reddit, an online social news website and discussion forum community, using Natural Language Processing (NLP) techniques. Using data from subreddits related to mental health and antidepressant use, sentiment analysis and topic modelling was used to explore themes discussed in these subreddits. In this talk I will discuss the construction of a pipeline for sentiment analysis and topic modelling, the complexities involved in model output interpretation, as well as the challenges of social media data analytics more broadly.

Registration link