This project will use advances in natural language processing, specifically sentiment analysis.

In this project, we aim to improve the state-of-the-art in hate speech detection towards LGBTI+ people using advances in parameter-efficient, language-agnostic fine-tuning. Leveraging multilingual datasets will allow the language model to leverage a broader range of information.

LGBTI+ people experience online hate and abuse although LGBTI+ rights are advanced in more countries than ever. The project will contribute to making the internet safer for all. The project aligns with the UN Sustainable Development goal to promote peaceful and inclusive societies (Goal 16).

The project will be funded by Google’s exploreCSR grant.

School

Computer Science and Engineering

Research Areas

Natural language processing

The candidate will interact with the NLP research group in the school of CSE. The candidate may collaborate with external partners from Google.

A strong candidate will have experience in natural language processing and deep learning. As a result of the project, we expect the candidate to develop strong skills in multilingual language models and parameter-efficient fine-tuning techniques such as prompt tuning.

  1. Short literature review
  2. Well-documented code
  3. Research report containing experimental results and project details, suitable for submission to an NLP conference/workshop

Some multilingual datasets we will use are: