Meghana Bhange

A research project that I worked on with the Trustworthy Information Systems Lab at ETS Montreal. We researched how model adversaries can leverage the information provided by counterfactual explanations to build high-fidelity and high-accuracy model extraction attacks.

Named entity recognition based on Spacy3 trained on the LitBank dataset. This project uses Roberta XLM as a base model and fine-tuned literature data to understand the terms generally used in literature. The pre-trained model released with the project can also be used to perform NER tasks on any literary text.

This work adds two common approaches, fine-tuning large transformer models and sample efficient methods like ULMFiT. Prior work demonstrated the efficacy of classical ML methods for polarity detection. We fine-tuned general-purpose languages representation models, such as those of the BERT family, which are benchmarked along with classical machine learning and ensemble methods. We showed that NB-SVM beats RoBERTa by 6.2%. The best-performing model is a majority-vote ensemble which achieves an F1 of 0.707.

Simple streamlit application to be paired with textbooks so that one can easily extract the keyphrases from the text they are reading and get detailed information from Wikipedia to understand the relevant context.

Artificial Insanity is a multiplayer game for the Toptal Hackathon, which uses Stable Diffusion to generate image responses for a text prompt. In each round, a player chooses prompt, and other players generate a response using an integrated Stable Diffusion interface.

I benchmarked performance in terms of quality and latency for DALLE and Stable Diffusion. Also, I deployed the final model on FastAPI to make it easier to integrate with the rest of the back end. The solution won the second prize in the Hackathon.