OpenSubtitles2018: Statistical Rescoring of Sentence Alignments in Large, Noisy Parallel Corpora

By Pierre Lison, Jörg Tiedemann, Milen Kouylekov. 2018

204 citations recorded by the source corpus.

Movie and TV subtitles are a highly valuable resource for the compilation of parallel corpora thanks to their availability in large numbers and across many languages. However, the quality of the resulting sentence alignments is often lower than for other parallel corpora. This paper presents a new major release of the OpenSubtitles collection of parallel corpora, which is extracted from a total of 3.7 million subtitles spread over 60 languages. In addition to a substantial increase in the corpus size (about 30 % compared to the previous version), this new release associates explicit quality scores to each sentence alignment. These scores are determined by a feedforward neural network based on simple language-independent features and estimated on a sample of aligned sentence pairs. Evaluation results show that the model is able predict lexical translation probabilities with a root mean square error of 0.07 (coefﬁcient of determination R 2 = 0.47). Based on the scores produced by this regression model, the parallel corpora can be ﬁltered to prune out low-quality alignments.

Subfields: Artificial Intelligence, Natural Language Processing, Evaluation & Benchmarks

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Natural Language Processing

In Evaluation & Benchmarks

Citations in the corpus

Self-training Improves Pre-training for Natural Language Understanding (2020, background): “, 2018), together with OpenSubtitles (Lison et al., 2019) and Europarl (Koehn, 2005) parallel data from English to French, Italian and Indonesian - language pairs that were shown to provide good paraphrastic sentence embeddings (Wieting et al.”
Analyzing Multi-Head Self-Attention: Specialized Heads Do the Heavy Lifting, the Rest Can Be Pruned (2019, methodology): “For English-Russian, we perform additional experiments using the publicly available OpenSubtitles2018 corpus (Lison et al., 2018) to evaluate the impact of domains on our results.”
Enhanced Direct Speech-to-Speech Translation Using Self-supervised Pre-training and Data Augmentation (2022, positive): “Parallel Text, Es-En & En-Es OpenSubtitle2018 [45] 64.”
Overcoming Catastrophic Forgetting During Domain Adaptation of Neural Machine Translation (2019, methodology): “, 2017) and OpenSubtitles18 (Lison et al., 2018) corpora.”
The FLORES Evaluation Datasets for Low-Resource Machine Translation: Nepali–English and Sinhala–English (2019, background): “3 For Sinhala–English, there are an additional 600K sentence pairs automatically aligned from from OpenSubtitles (Lison et al., 2018).”
Two New Evaluation Datasets for Low-Resource Machine Translation: Nepali-English and Sinhala-English (2019, background): “4 For Sinhala– English , there are an additional 600K sentence pairs automatically aligned from from OpenSubtitles (Lison et al., 2018).”
Context-Aware Neural Machine Translation Learns Anaphora Resolution (2018, methodology): “To this end, we construct a simple discourse-aware model, demonstrate that it achieves improvements over the discourse-agnostic baseline on an English-Russian subtitles dataset (Lison et al., 2018) and study which context information is being captured in the model.”
Cross-lingual Transfer Learning for Multilingual Task Oriented Dialog (2018, methodology): “For the Thai models, we use 10 copies of the IWSTL training data [2] as well as the OpenSubtitles data [18] for training and the IWSTL development and test data for model selection.”
DiaBLa: a corpus of bilingual spontaneous written dialogues for machine translation (2019, background): “OpenSubtitles (Lison and Tiedemann 2016; Lison et al. 2018) is a large-scale corpus of film subtitles, from a variety of domains, making for very heterogeneous content.”
FINDINGS OF THE IWSLT 2020 EVALUATION CAMPAIGN (2020, background): “• OpenSubtitles 2018 (Lison et al., 2018)”
Shaping the Narrative Arc: An Information-Theoretic Approach to Collaborative Dialogue (2019, methodology): “The first two models are based on the OpenSubtitles dataset (Lison et al., 2018).”