How well do layout-aware/visually-rich/Document AI models perform on classification requiring content understanding in multiple languages?
Here's our 📑 accepted to #EMNLP2023 findings w/ Siddharth, Nishant, Bonan Min, Srikar Yogarshi Vyas at Amazon Science
📑 arxiv.org/pdf/2310.16356…
🧵
Contrastive Training Improves Zero-shot Classification of Semi-structured Documents
with Yogarshi Vyas Miguel Ballesteros and others from Amazon Science
Summary: A contrastive pretraining objective to boost zero-shot classification of visually rich docs
Paper: arxiv.org/abs/2210.05613
Big thank you to my co-authors Neha Anna John, Ling Liu, Yogarshi Vyas, Jie Ma, Yoshinari Fujinuma, Miguel Ballesteros, Vittorio Castelli, and Dan Roth for their help!
Check out our paper for more details!
🚨Internship Alert🚨
We have an NLP research intern opening for our team (AWS Comprehend) this upcoming fall, and we are currently *very* actively looking for candidates for this position.
Drop either me or Yogarshi Vyas a note if you applied, or if you have further questions.
I would finally like to thank my team at Amazon Web Services Shuai, Yogarshi (@yogarshi ), Neha, Yassine (@benajibayassine ) and Miguel (@migballesteros) for the great mentorship and support during my internship!
Yogarshi Vyas English.
Others have also suggested Tesseract so will give it a try!
Thanks Yogarshi