Title Of the Talk: Detecting Hope and Peace in Social Media
Speaker: Shriphani Palakodety
Host Faculty: Dr. Vineeth N Balasubramanian
Date & Time: Tuesday, 18th Feb 2020 14:30
Venue: Room No C-LH3
Abstract:
Text analyses in the Indian subcontinent present unique challenges. Content is authored in a variety of low-resource languages and
exhibits high levels of code-switching. In this talk, we will discuss recent case studies that introduce elegant, low-supervision
methods to address the linguistic diversity in user-authored Indian web content. Our case studies include the Rohingya refugee
crisis, and the 2019 India/Pakistan conflict. We explore two new tasks - help speech detection and hope speech detection to surface
de-escalating, positive content and explore active learning pipelines to address class imbalance. The associated papers were
selected for oral presentations at AAAI 2020, and ECAI 2020
Speaker Profile:
Shriphani Palakodety, a researcher-cum-engineer at Onai (California, USA), has developed low-resource NLP methods for a variety
of computational social science problems based in the Indian subcontinent, designed some of the first neural information retrieval
techniques, and developed techniques at the intersection of Zero-Knowledge proofs and formal methods. His work in these areas has
been featured in several leading press outlets. He earned an MS from Carnegie Mellon University, and a BS from Purdue University.
He has designed and taught popular courses in deep learning, and blockchain at San Jose State University. In addition to research,
Shriphani runs one of India’s most popular data-visualization websites. His work has been featured in leading Indian press outlets
and used by several local and international geospatial analysts. On a different note, his wood sculptures have been exhibited at
leading art galleries in California
Dates:
Tuesday, 18th Feb 2020 14:30