UAlacant machine translation quality estimation at WMT 2018: a simple approach using phrase tables and feed-forward neural networks

Abstract

We describe the Universitat d'Alacant submissions to the word-and sentence-level machine translation (MT) quality estimation (QE) shared task at WMT 2018. Our approach to word-level MT QE builds on previous work to mark the words in the machine-translated sentence as OK or BAD, and is extended to determine if a word or sequence of words need to be inserted in the gap after each word. Our sentence-level submission simply uses the edit operations predicted by the word-level approach to approximate TER. The method presented ranked first in the sub-task of identifying insertions in gaps for three out of the six datasets, and second in the rest of them.

Findings of the WMT 2018 Shared Task on Quality Estimation

“…Fluency is computed by taking the inverse of cross-entropy, according to an in-domain language model. Both measures are combined ID Participating team CMU-LTI Carnegie Melon University, US (Hu et al, 2018) JU-USAAR Jadavpur University, India & University of Saarland, Germany (Basu et al, 2018) MQE Vicomtech, Spain (Etchegoyhen et al, 2018) QEbrain Alibaba Group Inc, US RTM Referential Translation Machines, Turkey (Biçici, 2018) SHEF University of Sheffield, UK (Ive et al, 2018b) TSKQE University of Hamburg (Duma and Menzel, 2018) UAlacant University of Alacant, Spain (Sánchez-Martíínez et al, 2018) UNQE Jiangxi Normal University, China UTartu University of Tartu, Estonia (Yankovskaya et al, 2018) via simple arithmetic means on rescaled values, i.e., no machine learning is used. Since it is unsupervised, the method can only be meaningfully evaluated on the ranking task.…”

Neural Machine Translation for Low Resource Bantu Languages in East and Southern Africa

“…Their BLEU score was 16.5 for Shona-English translation. As for English-Swahili, we benchmarked with [24] work. We used similar data set from [5].…”