Dr Aditya Joshi | UNSW Research

My Expertise

Natural language processing

Large language models

Keywords

natural language processing

sentiment analysis

recommender systems

social media analysis

Fields of Research (FoR)

Natural language processing

Biography

I am a Lecturer (US/India Equivalent: Assistant Professor) in the School of Computer Science & Engineering at UNSW, specializing in natural language processing (NLP).

Research: As of April 2024, my papers have 2500+ citations (h-index: 23), and I have been the lead investigator in grants with a cumulative total of AUD150,000, including Google exploreCSR and Google Research scholar grants. My publications are in conferences such as ACL, EMNLP, COLING, and CONLL, and in journals such as ACM Computing Surveys and PLOS One. I have been acknowledged as an outstanding reviewer at ICML, ACL and EACL. I have presented tutorials at EMNLP, AACL, ALTA and ICON. My 2018 TEDx talk 'Detecting sarcasm, combating hate' interleaved my PhD thesis with my personal journey. I apply NLP to epidemic intelligence in my collaboration with EPIWATCH (Kirby Institute) and to cybersecurity with IFCYBER (School of CSE). I supervise undergraduate, Masters and PhD students on their research projects in NLP.

Teaching: I designed and introduced a new NLP course at UNSW in 2024. I have co-authored a textbook on NLP with Pushpak Bhattacharyya, published by Wiley in 2023.
I am an Associate of the Human Rights Institute at UNSW, and co-lead the Community of Practice for Inclusive Research with Queer, Trans & people with variations of sex characteristics. I was the organizing chair for 'Rainbow AI': a workshop to support queer students in Australia interested in HDR and research careers in AI. The workshop was funded by Google's exploreCSR grant.

Past Experience: I also have over five years of industry experience in applying NLP techniques to various domains, such as online employment marketplace, meeting analytics and epidemic intelligence. I obtained a joint PhD degree from IIT Bombay (India) and Monash University (Australia) in 2018, and a best PhD thesis award by IITB-Monash Research Academy. Before joining UNSW in 2023, I worked as a data scientist at SEEK, a market leader in online employment marketplaces, and as a machine learning engineer at Notiv, a startup that provides AI-powered meeting insights, and a senior data scientist at Fractal, a leader in analytics. In these roles, I developed and deployed recommender systems, land natural language understanding systems, using state-of-the-art tools and frameworks, such as Azure Databricks, Atlassian Suite, and PyTorch. I have also worked as a lead trainer at the Institute of Data, that offers data science certifications to industry professionals.

My Grants

Topic	Funding organisation	Name of Funding program	Co-PIs	Amount
A benchmark for sentiment and sarcasm classification for dialects of English (Lead Investigator)	Google Research	Google exploreCSR	Dr. Diptesh Kanojia (Uni of Surrey, UK)	92K AUD
Fake news detection in the context of national security policy (Lead Investigator)	UNSW Global	Global Research and Innovation Program (GRIP)	Profs. Sanjay Jha, Salil Kanhere (UNSW)	20K AUD
LGBTI+ inclusion in AI (Lead Investigator)	Google Research	Google exploreCSR	Dr. Ben Hutchinson (Google)	51K AUD

My Qualifications

PhD, Indian Institute of Technology Bombay, India and Monash University, Australia (jointly awarded). Thesis title: 'Investigations in Computational Sarcasm' [Monograph]
MTech (Computer Science and Engineering), IIT Bombay. Dissertation title: 'Adaptation of Sentiment Analysis to a New Text Form'

My Awards

Publications:

Best PhD Thesis (2018) awarded by IITB-Monash Research Academy
Best Paper at ACM FAccT 2023 in June 2023. (Collaborative paper by multiple authors at Queer in AI)
Best Paper at MoMM 2020 in December 2020. (Lead author was a PhD student at Macquarie University, Sydney, who was co-supervised by me).
Best Student Paper - Runner Up at ALTA 2019 in December 2019. (Lead author was a PhD Student at RMIT, Melbourne.)
Best Paper from IITB-Monash Research Academy consecutively in 2015 and 2014.

Presentations:

Best Sprint Thesis Talk (Senior Researcher category) at RISC 2016, Research symposium organized by Department of CSE, IIT Bombay in April 2016.
Best 3-Minute Thesis Talk Awards at IITB-Monash Research Academy in 2015 and 2014.
Best Poster, IBM Research Day, Dept. of CSE, IIT Bombay in August 2015.
Invited speaker, VAIBHAV Summit organised by the Government of India, 2020.

Other:

First place in the shared task on vaccination behaviour detection at SMM4H workshop at EMNLP 2018 in October 2018.
Tata Consulting Services Research Scholar Fellowship in 2013.

My Research Activities

I have worked in several problems of natural language processing (NLP) and its applications to several fields: epidemic intelligence, cybersecurity and LGBTI inclusion. This includes research as well as industry outputs. A significant portion of my current research is making NLP models robust for dialects of English.

My Research Supervision

Supervision keywords

natural language processing

sentiment analysis

deep learning

public health surveillance

recommender systems

Areas of supervision

Broad topics in Natural Language Processing (NLP) along three 'F's:

- Fairer: New Techniques for Equity in NLP w.r.t. non-native users of English and the LGBTI+ community

- Faster: New Techniques for Algorithmic Enhancements to Transformers

-Finer: Applications of Large Language Models to: (a) Public Health, (b) Mobility, (c) Society

Currently supervising

- Kernel-based reformulation of attention in Transformers

- Prompt-based methods for sarcasm detection

- Conversation Understanding for Dialects

- Multilingual small-scale large language models

- Misinformation Detection in News Articles

-LLMs for personalized learning

-Emotion recognition in music

My Engagement

Tutorials:

`NLP for Healthcare in the Absence of a Healthcare Dataset', AACL, Suzhou, China, December 2020. (Co-speaker: Sarvnaz Karimi)
`NLP for Healthcare in the Absence of a Healthcare Dataset', ALTA, Sydney, Australia, December 2019. (Co-speaker: Sarvnaz Karimi)
`Computational Sarcasm', presented at \textbf{EMNLP} 2017, Copenhagen, Denmark, September 2017. (Co-speaker: Pushpak Bhattacharyya)

Non-conference talks:

`Language of the Queer in India' at the `Queer in AI' social at NAACL, 2021.
`Social media-based epidemic intelligence' in the panel on `NLP for social good' at the VAIBHAV summit (Vaishwik Bharatiya Vaigyanik) summit organised by the Government of India, 2020.
`Detecting Sarcasm, Combating Hate', TEDx talk at TEDxSomaiyaVidyavihar, an independently organised TEDx event, Mumbai, India, 2018.
`Detecting sarcasm using incongruity', invited speaker at WASSA workshop at EMNLP 2017, Copenhagen, Denmark, 2017.

Posters:

`Computational Sarcasm', Google NLP Summit organized by Google Zurich, September 2017.
`Sarcasm Detection' and `Drunk-Texting Prediction', Research Colloqium, XRCI Open 2016 organized by Xerox Research Center India, Bengaluru, January 2016.
`Sarcasm Technology', IBM Research Day 2015 organized by IBM Research Lab, Bengaluru, August 2015.
`Sentiment Annotation Complexity', Microsoft TechVista 2015 organized by Microsoft Research India, Bengaluru, January 2015.

Panel Discussions:

Panelist in a discussion on `Integrating ChatGPT in Education' organized by EdTech IIT Bombay, Mumbai, 2023. (Online event)
Panelist in a discussion on `Achieving better health through AI' organized by Venture Cafe Sydney in Macquarie Park, Sydney, 2019.
Panelist in a discussion on `AI in India: Today and Tomorrow' at Data Science Day organized by Web \& Coding Club at IIT Bombay, Mumbai, 2018.