Research & Publications

2023

Jointly Optimizing Translations and Speech Timing to Improve Isochrony in Automatic Dubbing

A. Chronopoulou, B. Thompson, P. Mathur, Y. Virkar, S. M. Lakew, M. Federico
arXiv:2302.12979

2022

Isometric MT: Neural Machine Translation for Automatic Dubbing

S. M. Lakew, Y. Virkar, P. Mathur, M. Federico
ICASSP 2022

Sockeye 3: Fast Neural Machine Translation with PyTorch

F. Hieber, M. Denkowski, T. Domhan, B. D. Barros, C. D. Ye, X. Niu, C. Hoang, et al. (incl. S. M. Lakew)
arXiv:2207.05851

Findings of the IWSLT 2022 Evaluation Campaign

A. Anastasopoulos, L. Barrault, L. Bentivogli, M. Z. Boito, O. Bojar, R. Cattoni, et al. (incl. S. M. Lakew)
IWSLT 2022

2021

Isochrony-Aware Neural Machine Translation for Automatic Dubbing

D. Tam, S. M. Lakew, Y. Virkar, P. Mathur, M. Federico
arXiv:2112.08548

Machine Translation Verbosity Control for Automatic Dubbing

S. M. Lakew, M. Federico, Y. Wang, C. Hoang, Y. Virkar, R. Barra-Chicote, R. Maas
ICASSP 2021

Self-Learning for Zero-Shot Neural Machine Translation 🏆 Best Paper

S. M. Lakew, M. Negri, M. Turchi
LoResMT @ MT Summit 2021

2020

Multilingual and Customized Neural Machine Translation

Multilingual and Customized Neural Machine Translation Thesis

Surafel M. Lakew
PhD Thesis, University of Trento & FBK HLT-MT Lab | Advised by Marcello Federico
My doctoral thesis investigates approaches to improve the performance of neural machine translation models for language pairs without large-scale training data. This work explores zero-shot translation, verbosity control, and isochrony-aware models for automatic dubbing.

Low Resource Neural Machine Translation: A Benchmark for Five African Languages

S. M. Lakew, M. Negri, M. Turchi
arXiv:2003.14402

2019

Controlling the Output Length of Neural Machine Translation

S. M. Lakew, M. A. Di Gangi, M. Federico
IWSLT 2019

Adapting Multilingual Neural Machine Translation to Unseen Languages

S. M. Lakew, A. Karakanta, M. Federico, M. Negri, M. Turchi
IWSLT 2019

Multilingual Neural Machine Translation for Zero-Resource Languages

S. M. Lakew, M. Federico, M. Negri, M. Turchi
arXiv:1909.07342

2018

A Comparison of Transformer and Recurrent Neural Networks on Multilingual Neural Machine Translation

S. M. Lakew, M. Cettolo, M. Federico
COLING 2018

Transfer Learning in Multilingual Neural Machine Translation with Dynamic Vocabulary

S. M. Lakew, A. Erofeeva, M. Negri, M. Federico, M. Turchi
IWSLT 2018

Multilingual Neural Machine Translation for Low-Resource Languages

S. M. Lakew, M. Federico, M. Negri, M. Turchi
IJCoL 2018

Neural Machine Translation into Language Varieties

S. M. Lakew, A. Erofeeva, M. Federico
WMT 2018

Adapting Multilingual NMT to Extremely Low Resource Languages: FBK's Participation in the Basque-English Low-Resource MT Task

S. M. Lakew, M. Federico
IWSLT 2018

2017

Improving Zero-Shot Translation of Low-Resource Languages 🏆 Best Paper

S. M. Lakew, Q. F. Lotito, M. Negri, M. Turchi, M. Federico
IWSLT 2017

FBK's Multilingual Neural Machine Translation System for IWSLT 2017

S. M. Lakew, Q. F. Lotito, M. Turchi, M. Negri, M. Federico
IWSLT 2017

2015

Multi-Camera Multi-Target Tracking Using Relaxation Processes Thesis

Surafel M. Lakew
MSc Thesis, Ca' Foscari University of Venice | Advised by Marcello Pelillo & Andrea Prati
My masters thesis focuses on multi-camera, multi-target tracking framed as a game-theoretic relaxation labeling problem. Targets are detected via a probabilistic occupancy map, then consistently labeled across views using a replicated-labels extension to handle limited cross-camera visibility.

Using Dominant Sets for Data Association in Multi-Camera Tracking

A. K. Hamid, S. M. Lakew, M. Pelillo, A. Prati
ICDSC 2015

Patents

Systems, Methods, and Apparatuses for Controlling Output Length in Neural Machine Translation Patent

M. Federico, M. A. Di Gangi, S. M. Lakew
US Patent 11,295,081