Hi, I’m Sahana 👋
I am a Predoctoral Researcher at Google Research, India in the Advertising Sciences team. I recently graduated from the Indian Institute of Technology, Madras, with a BTech in Electrical Engineering and an Interdisciplinary MTech in Data Science.
My research interests lie in applying machine learning and deep learning to solve NLP and multimodal problems. I am also highly interested in analyzing and incorporating interpretability in such systems. You can view my work on Google Scholar and Github.
You can contact me on Twitter, Linkedin, or by mail: sahanjich@gmail.com.
- HintedBT: Augmenting Back-Translation with Quality and Transliteration HintsIn Proceedings of the 2021 Conference on Empirical Methods in Natural Language Processing 2021
- A Framework for Rationale Extraction for Deep QA modelsarXiv preprint arXiv:2110.04620 2021
- Towards Interpreting BERT for Reading Comprehension Based QAIn Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) 2020
- Scene Graph based Image Retrieval–A case study on the CLEVR DatasetarXiv preprint arXiv:1911.00850 2019
Predoctoral Researcher @ Google Research, India
Advertising Sciences and Machine Translation
August 2020 - August 2022Teaching Assistant @ IIT Madras
Analog Systems and Lab
January 2020 - May 2020Teaching Assistant @ IIT Madras
Data Analytics Lab
August 2019 - December 2019Research Intern @ IBM Research, India
AI Horizons Network, with IIT Bombay
May 2019 - July 2019UG Student Mentor @ IIT Madras
Saathi, Student Body
August 2019 - December 2019Summer Intern @ Mastercard, India
Enterprise Architecture Team
May 2017 - August 2017CVI Coordinator @ IIT Madras
Centre for Innovation
August 2016 - May 2017Envisage Coordinator @ IIT Madras
Shaastra, IITM technical fest
August 2016 - May 2017
HintedBT - Augmenting Back-Translation with Quality and Transliteration Hints
Enabling effective usage of back-translation data for machine translation of low-resource, cross-script language pairs through hints/tags.Technologies
nlp machine translation back-translation data augmentation low-resource cross-script tensorflow
Analyzing Interpretability of Deep RCQA Systems
Defining and analyzing the interpretability of four existing deep models (BERT, BiDAF, DCN, QANet) for the task of Reading Comprehension based Question Answering, on the dataset SQuAD.Technologies
python nlp BERT interpretability integrated gradients tensorflow