Authors: Sayed Hesam Alavian, Ali Satvaty, Sadra Sabouri, Ehsaneddin Asgari, Hossein Sameti Venue: Document-grounded Dialogue and Conversational Question Answering Workshop (DialDoc) at ACL, 2022

Summary

The team’s entry for the DialDoc-22 (MultiDoc2Dial) shared task, and the substance of Sadra’s BSc thesis. Docalog is a three-stage pipeline: first retrieve the document the answer should come from, then predict the answer span inside it, then pick the most likely span out of everything predicted.

Why this one matters

This is the paper Sadra has a soft spot for. It was his first, and the first time he followed an idea the whole way from a thesis topic to something with his name on it in workshop proceedings. It is what pulled him into NLP, and into research and publishing at all — everything else he has published came after it.

Paper

Bachelor of Electrical Engineering and Computer Science — Sharif University of Technology — this was the bachelor’s thesis. Sharif University of Technology — Speech and Language Processing Lab — where it was done, pre-ChatGPT. naab: A Ready-to-Use Plug-and-Play Corpus for Farsi — another paper from the same lab, with Hossein Sameti and Ehsaneddin Asgari as shared co-authors.