Abstract
This paper describes the submission of our multilingual NLP model performance evaluation system for the SUMEval 2022 shared task, a system for predict the performance of a model on a set of target languages. The system is based on the LITMUS model (Srinivasan et al., 2022), with the addition of 3 new features and model ensembling. Experimental results show that our system obtains a significant improvement than the baseline on both the test set and the surprised test set. Our system has achieved a 11% MAE reduction on the test set and is the best-performing submission on the surprise test set with 17% MAE reduction compared to the baseline.
Syeda Sabrina Akter
CS PhD Student
I am Syeda Sabrina Akter. I am pursuing my PhD in Computer Science at George Mason University. Previously I studied CS (BSc) at the Bangladesh University of Professionals, Bangladesh. My academic interest involves learning different aspects of multilingual natural language processing and speech translation. Currently, I am working on a project of documenting low resource endangered languages with the help of pre-trained ASR models of high resource languages. Outside academics, I like reading and traveling.
Antonios Anastasopoulos
Assistant Professor
I work on multilingual models, machine translation, speech recognition, and NLP for under-served languages.