Tilde at WMT 2020: News Task Systems

2020-10-29 · WMT (EMNLP) 2020 11 · Rihards Krišlauks, Mārcis Pinnis

This paper describes Tilde's submission to the WMT2020 shared task on news translation for both directions of the English-Polish language pair in both the constrained and the unconstrained tracks. We follow our submissions from the previous years and build our baseline systems to be morphologically motivated sub-word unit-based Transformer base models that we train using the Marian machine translation toolkit. Additionally, we experiment with different parallel and monolingual data selection schemes, as well as sampled back-translation. Our final models are ensembles of Transformer base and Transformer big models that feature right-to-left re-ranking.

📄 PDF Abstract BibTeX arXiv:2010.15423

Code (0)

등록된 구현이 없습니다.

Tasks

Machine TranslationRe-RankingTranslation

Methods 이 논문이 사용한 방법론

Linear Layer A Linear Layer is a projection $\mathbf{XW + b}$.

Absolute Position Encodings Absolute Position Encodings are a type of position embeddings for [Transformer-based models] where positional encodings are…

Position-Wise Feed-Forward Layer 설명 없음

Attention 설명 없음

Residual Connection 설명 없음

Multi-Head Attention 설명 없음

Layer Normalization Unlike batch normalization, Layer Normalization directly estimates the normalization statistics from the summed inputs…

BPE Byte Pair Encoding, or BPE, is a subword segmentation algorithm that encodes rare and unknown words as sequences of subword units. The intuition is that various word…

Tilde's Machine Translation Systems for WMT 2018

2018-10-01 · WS 2018 10 · M{\=a}rcis Pinnis, Mat{\=\i}ss Rikters, Rihards Kri{\v{s}}lauks

The paper describes the development process of the Tilde{'}s NMT systems that were submitted for the WMT 2018 shared task on news translation. We describe the data filtering and pre-processing workflows, the NMT system t…

Machine Translation NMT Translation

Tilde's Machine Translation Systems for WMT 2019

2019-08-01 · WS 2019 8 · Marcis Pinnis, Rihards Kri{\v{s}}lauks, Mat{\=\i}ss Rikters

The paper describes the development process of Tilde{'}s NMT systems for the WMT 2019 shared task on news translation. We trained systems for the English-Lithuanian and Lithuanian-English translation directions in constr…

Machine Translation NMT Translation

Introducing RONEC - the Romanian Named Entity Corpus

2020-05-01 · LREC 2020 5 · Stefan Daniel Dumitrescu, Andrei-Marius Avram

We present RONEC - the Named Entity Corpus for the Romanian language. The corpus contains over 26000 entities in {\textasciitilde}5000 annotated sentences, belonging to 16 distinct classes. The sentences have been extrac…

named-entity-recognition Named Entity Recognition Named Entity Recognition (NER)

Tilde's Parallel Corpus Filtering Methods for WMT 2018

2018-10-01 · WS 2018 10 · M{\=a}rcis Pinnis

The paper describes parallel corpus filtering methods that allow reducing noise of noisy {``}parallel{''} corpora from a level where the corpora are not usable for neural machine translation training (i.e., the resulting…

Machine Translation Translation Transliteration Word Alignment

CAMB at CWI Shared Task 2018: Complex Word Identification with Ensemble-Based Voting

2018-06-01 · WS 2018 6 · Sian Gooding, Ekaterina Kochmar

This paper presents the winning systems we submitted to the Complex Word Identification Shared Task 2018. We describe our best performing systems{'} implementations and discuss our key findings from this research. Our be…

Binary Classification Complex Word Identification General Classification Lexical Simplification+2