eliyahabba Goto Github PK

👋 Hey there! I'm Eliya, a passionate NLP researcher and data scientist based in Jerusalem, Israel. I thrive on challenges and dynamic teamwork, especially in the exciting fields of NLP and data science.

This repository contains the code for an annotation interface designed for assessing judicial attitudes toward victims of sexual violence in the Israeli court system. The study focuses on analyzing court statements to evaluate how rape myths resonate in the criminal justice system's response to sex crimes, particularly in the judicial assessment of victim credibility.

For an in-depth understanding of the motivation and details behind this annotation interface, please refer to the Data section in our paper "The Perfect Victim: Computational Analysis of Judicial Attitudes towards Victims of Sexual Violence" presented at ICAIL 2023.

This project explores correlations between attention layers of Audio Speech Recognition (ASR) and Natural Language Processing (NLP) models, aiming to enhance ASR performance and uncover hidden language models in ASR. Utilizing BERT for NLP and Wav2vec2 for ASR, we define metrics for correlation, address alignment challenges, and investigate relationships across the English language. Our user-friendly interface visually presents aligned attention matrices, aiding in model analysis and investigating correlations.

The paper concludes with potential linguistic and semantic relations between these models, along with methods and metrics applicable for further research.