October 8, 2021
Title: Exploring a Sampling-based Alternative to Beam Search
Time: Oct 8, 2021 05:00 PM (17:00) Universal Time UTC
Zoom link: Join from PC, Mac, iOS or Android: https://unimelb.zoom.us/j/83119608970?pwd=dW96WEo1NzVuWHFOMTdGblVmWXBxZz09 Password: 269205
Rocketchat (for questions): https://sigtyp.inf.ethz.ch/channel/lecture-mueller
Rocketchat (for questions): https://sigtyp.inf.ethz.ch/channel/lecture-mueller
Please distribute the Zoom link as widely as you want, e.g. on local mailing lists. However, do not post it on any social media platforms to avoid Zoom-bombing. We did not put in a waiting room for convenience so if we are Zoom-bombed, we will have to restart the call with a new meeting id and that will be a hassle.
Abstract:
In this talk, I will discuss the potential of Minimum Bayes Risk (MBR) decoding — a sampling-based decoding algorithm — to replace beam search in machine translation. Beam search is the de-facto standard decoding algorithm for many language generation problems. However, recent work has found that beam search itself causes or exacerbates well-known biases in machine translation. Minimum Bayes Risk (MBR) decoding was suggested as an alternative algorithm that does not search for the highest-scoring translation but operates on a pool of samples. I will highlight that MBR does not alleviate well-known biases in machine translation, but, interestingly, increases the robustness to noise in the training data and to domain shift.
In this talk, I will discuss the potential of Minimum Bayes Risk (MBR) decoding — a sampling-based decoding algorithm — to replace beam search in machine translation. Beam search is the de-facto standard decoding algorithm for many language generation problems. However, recent work has found that beam search itself causes or exacerbates well-known biases in machine translation. Minimum Bayes Risk (MBR) decoding was suggested as an alternative algorithm that does not search for the highest-scoring translation but operates on a pool of samples. I will highlight that MBR does not alleviate well-known biases in machine translation, but, interestingly, increases the robustness to noise in the training data and to domain shift.
Bio:
Mathias is a post-doc and lecturer at the University of Zurich. His current main interests are 1) the meta-sciences of scientific integrity, methodology and reproducibility applied to machine translation, 2) decoding algorithms and 3) sign language translation. In his personal life he is a father of two and a passionate musician.
Mathias is a post-doc and lecturer at the University of Zurich. His current main interests are 1) the meta-sciences of scientific integrity, methodology and reproducibility applied to machine translation, 2) decoding algorithms and 3) sign language translation. In his personal life he is a father of two and a passionate musician.
Please find more information about him here: cl.uzh.ch/mmueller
You may watch all past talks on our Youtube channel: https://www.youtube.com/channel/UCaSWMbnmduXYlbWGEWLedww/playlists
_______________________________________________