Alert button

Score Combination for Improved Parallel Corpus Filtering for Low Resource Conditions

Nov 16, 2020
Muhammad N. ElNokrashy, Amr Hendy, Mohamed Abdelghaffar, Mohamed Afify, Ahmed Tawfik, Hany Hassan Awadalla

Figure 1 for Score Combination for Improved Parallel Corpus Filtering for Low Resource Conditions
Figure 2 for Score Combination for Improved Parallel Corpus Filtering for Low Resource Conditions
Figure 3 for Score Combination for Improved Parallel Corpus Filtering for Low Resource Conditions
Figure 4 for Score Combination for Improved Parallel Corpus Filtering for Low Resource Conditions

Share this with someone who'll enjoy it:

This paper describes our submission to the WMT20 sentence filtering task. We combine scores from (1) a custom LASER built for each source language, (2) a classifier built to distinguish positive and negative pairs by semantic alignment, and (3) the original scores included in the task devkit. For the mBART finetuning setup, provided by the organizers, our method shows 7% and 5% relative improvement over baseline, in sacreBLEU score on the test set for Pashto and Khmer respectively.

* Accepted at WMT20 (EMNLP 2020 Fifth Conference on Machine Translation)  
View paper onarxiv icon

Share this with someone who'll enjoy it: