The training is based on the idea that a translated sentence should be mapped to the same location in the vector space as the original sentence.
However, it requires that both sentences are fed into the network, which causes a massive computational overhead: Finding the most similar pair in a collection of 10, 000 sentences requires about 50 million inference computations (~65 hours) with BERT.
Ranked #6 on Semantic Textual Similarity on STS Benchmark (Spearman Correlation metric)
We present an approach based on multilingual sentence embeddings to automatically extract parallel sentences from the content of Wikipedia articles in 85 languages, including several dialects or low-resource languages.
We introduce an architecture to learn joint multilingual sentence representations for 93 languages, belonging to more than 30 different families and written in 28 different scripts.
CROSS-LINGUAL BITEXT MINING CROSS-LINGUAL DOCUMENT CLASSIFICATION CROSS-LINGUAL NATURAL LANGUAGE INFERENCE CROSS-LINGUAL TRANSFER DOCUMENT CLASSIFICATION JOINT MULTILINGUAL SENTENCE REPRESENTATIONS PARALLEL CORPUS MINING
Machine translation is highly sensitive to the size and quality of the training data, which has led to an increasing interest in collecting and filtering large parallel corpora.
Although much effort has recently been devoted to training high-quality sentence embeddings, we still have a poor understanding of what they are capturing.
For both variants, we investigate and report the relationship between model complexity, resource consumption, the availability of transfer task training data, and task performance.
Ranked #1 on Text Classification on TREC-6
The analysis sheds light on the relative strengths of different sentence embedding methods with respect to these low level prediction tasks, and on the effect of the encoded vector's dimensionality on the resulting representations.