| Challenge: | Current unsupervised methods for learning cross-lingual correspondences involve multiple steps, including heuristic post-hoc refinement strategies. |
| Approach: | They propose to cast the correspondence problem directly as an optimal transport problem, building on the idea that word embeddings arise from metric recovery algorithms. |
| Outcome: | The proposed method can be estimated efficiently, requires little or no tuning, and performs comparable with the state-of-the-art in various unsupervised word translation tasks. |
Similar Papers
Using Optimal Transport as Alignment Objective for fine-tuning Multilingual Contextualized Embeddings (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Recent studies suggest different methods to improve multilingual word representations in contextualized settings including techniques that align between source and target embedding spaces. |
| Approach: | They propose to use Optimal Transport as an alignment objective during fine-tuning to improve multilingual contextualized representations for downstream cross-lingual transfer. |
| Outcome: | The proposed method achieves better performance on two tasks (XNLI and XQuAD) and is competitive with existing methods. |
Hierarchical Mapping for Crosslingual Word Embedding Alignment (2020.tacl-1)
Copied to clipboard
| Challenge: | Existing strategies that map word embeddings into a crosslingual space are biased towards the choice of the pivot language. |
| Approach: | They propose to map any two languages into a different middle space by learning mappings across languages in a hierarchical way. |
| Outcome: | The proposed strategy significantly improves vocabulary induction scores in all existing benchmarks and in a new non-English–centered benchmark. |
Unbalanced Optimal Transport for Unbalanced Word Alignment (2023.acl-long)
Copied to clipboard
| Challenge: | Figure 1 illustrates the challenges of monolingual word alignment. |
| Approach: | They propose to use the family of optimal transport (OT) to achieve unbalanced word alignment that values alignment and null alignment on unsupervised datasets. |
| Outcome: | The proposed methods are competitive against the state-of-the-art methods on challenging datasets with high null alignment frequencies. |
Cross-Lingual Word Embedding Refinement by ℓ1 Norm Optimisation (2021.naacl-main)
Copied to clipboard
| Challenge: | Existing methods for building high-quality CLWEs learn mappings that minimise the l2 norm loss function but this optimisation objective has been shown to be sensitive to outliers. |
| Approach: | They propose a simple post-processing step to improve cross-lingual word embeddings using the Manhattan norm goodness-of-fit criterion. |
| Outcome: | The proposed approach outperforms four state-of-the-art baselines in bilingual lexicon induction and cross-lingual transfer tasks. |
Unsupervised Cross-lingual Transfer of Word Embedding Spaces (D18-1)
Copied to clipboard
| Challenge: | Existing methods for cross-lingual word mapping require cross-linguistic supervision, but this is not available for many low resource languages. |
| Approach: | They propose an unsupervised method that learns transformation functions over corresponding word embedding spaces using a distributed distributional matching algorithm. |
| Outcome: | The proposed method performs better on bilingual lexicon induction and cross-lingual word similarity prediction datasets than other supervised and unsupervised methods. |
Improving Cross-Lingual Word Embeddings by Meeting in the Middle (D18-1)
Copied to clipboard
| Challenge: | Cross-lingual word embeddings are becoming increasingly important in multilingual NLP. |
| Approach: | They propose to apply an additional transformation after initial alignment to align two disjoint monolingual vector spaces. |
| Outcome: | The proposed approach outperforms state-of-the-art models in monolingual and cross-lingual evaluation tasks. |
Geometry-aware domain adaptation for unsupervised alignment of word embeddings (2020.acl-main)
Copied to clipboard
| Challenge: | Existing methods for learning bilingual word embeddings have been used in natural language processing. |
| Approach: | They propose a manifold based geometric approach for learning unsupervised alignment of word embeddings between the source and target languages. |
| Outcome: | The proposed approach outperforms state-of-the-art optimal transport based approach on bilingual lexicon induction task across several language pairs. |
A Simple Approach to Learning Unsupervised Multilingual Embeddings (2020.emnlp-main)
Copied to clipboard
| Challenge: | Recent work on unsupervised cross-lingual embeddings in the bilingual setting has given the impetus to learning a shared embeddable space for several languages. |
| Approach: | They propose to solve two sub-problems together to learn a shared embedding space for several languages. |
| Outcome: | The proposed approach outperforms existing methods in bilingual lexicon induction, cross-lingual word similarity, multilingual document classification, and multilingual dependency parsing tasks. |
Massively Multilingual Document Alignment with Cross-lingual Sentence-Mover’s Distance (2020.aacl-main)
Copied to clipboard
| Challenge: | Document alignment aims to identify pairs of documents in two distinct languages that are of comparable content or translations of each other. |
| Approach: | They propose an unsupervised scoring function that leverages cross-lingual sentence embeddings to compute the semantic distance between documents in different languages. |
| Outcome: | The proposed scoring function outperforms baseline methods on high-resource language pairs, 15% on mid-resourced language pairs and 22% on low-resourcing language pairs. |
Deep Generative Model for Joint Alignment and Word Representation (N18-1)
Copied to clipboard
| Challenge: | EmbedAlign model embeds words in their complete observed context and learns by marginalisation of latent lexical alignments. |
| Approach: | They exploit translation as a distributional context and embed words as posterior probability densities, rather than point estimates, which allows them to compare words in context using a measure of overlap between distributions. |
| Outcome: | The proposed model performs on a range of lexical semantics tasks and achieves competitive results on benchmarks including natural language inference, paraphrasing, and text similarity. |