Assistant Research Scientist -Part time summer

New York University

New York University

New York, NY, USA
Posted 6+ months ago

New York University: NYU - NY: Courant Institute of Mathematical Sciences: Computer Science - NYU Courant Institute

Location

New York

Open Date

Jun 23, 2021

Description

This research assistantship requires the selected candidate to work on analysis of learned NLG evaluation metrics (transformer models trained using regression to predict quality of a generated text such as dialog response or machine translation) in presence of quality drifts (i.e. shift in data distribution due to changes in NLG models over time, such as transition from seq-2-seq models of text generation to using GPT based models for the same). Additionally, the assistant will build experiments around approaches that can help mitigate the impact of quality drift and improve robustness of NLG evaluation metrics. This will be a short duration research project.

Responsibilities:

  • You will be working with dialog response evaluation and machine translation evaluation datasets.

  • You will be reading and staying up to date with the latest research in the field of building evaluation metrics for NLG evaluation.

  • You will run experiments to reproduce published research results, combining published research to obtain results in novel settings, and drive research towards building better unbiased and robust evaluation metrics.

  • You will be required to attend weekly meetings with the supervising professor and provide weekly updates on research experiments.

  • You will be required to perform critical thinking on existing research, and empirical results obtained from experiments carried out to excel in this role.


Qualifications

Skills and Qualifications:

  • MS in Computer Science.

  • Experience and coursework in Machine Learning, Deep Learning and Natural Language Processing.

  • Knowledge of large scale transformer models for language.

  • Experience in python based machine learning and deep learning tech stack, such as numpy, pytorch, HuggingFace and scikit-learn

  • Analytical approach coupled with solid communication skills and a sense of ownership.

Application Instructions

Please provide a cover letter and CV/List of Publications.