Challenges of Low-Resource Natural Language Processing: A Focus on Sentiment Analysis and Hate Speech Detection in Amharic
Date:
This seminar is delivered at the Cambridge NLP Seminar
You can finde my presentation here
Abstract
While high-resource languages such as English have made significant progress in various natural language processing (NLP) applications, low-resource languages are struggling to keep up. Drawing from our experience and expertise at the LT (Language Technology) group, this talk will shed light on the main challenges facing NLP in low-resource languages. After providing an overview of the basics, I will showcase hate speech detection and sentiment analysis as two use cases for the Amharic language. This presentation is an expansion of my keynote speech at AFricaNLP collocated with ICLR 2023 at Kigali, Rwanda.
Bio
I’m Seid Muhie Yimam, a Technical Lead at the house of computing and data science (HCDS) and Research Associate at the Language Technology Group at the University of Hamburg, supervised by Prof. Chris Biemann. At HCDS, I lead research on data processing of textual content for digital humanities. My Ph.D. focused on integrating adaptive machine learning models into NLP applications. I’ve participated on social media NLP, mainly hate speech detection and sentiment analysis for the Ethiopian language of Amharic. I teach NLP courses and supervise Master’s projects/thesis in the group.