Research map / Learning theory and algorithms / Sample complexity
On Learning Language-Invariant Representations for Universal Machine Translation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Require language-invariant representations without assumptions on language distributions | Universal machine translation · Many-to-many translation tasks with differing target-language marginals | maximum translation error: \frac{1}{2}\max_{k\in[K]}\max_{i\neq j}~\dtv(\mathcal{D}_{L_i, L_k}(L_k), \mathcal{D}_{L_j, L_k}(L_k)) - \frac{\epsilon} | Hurt |
| Assume parallel corpora follow a shared encoder-decoder generative process and train on a connected graph of language pairs | Universal machine translation · Unseen language pairs connected in translation graph H | translation error: 2\rho^2 \sum_{k=1}^{m-1} \epsilon_{L_k, L_{k+1}} | Helped |
| Use a translation graph with shorter paths between language pairs | Encoder-decoder generative model · Unseen language pairs | translation error: 2\rho^2 d_{L, L'}\cdot\epsilon | Helped |
| Allow randomized encoders and decoders under the stated generative assumptions | Universal machine translation · Unseen language pairs connected in translation graph H | translation error: 2\rho^2 \sum_{k=1}^{m-1} \epsilon_{L_k, L_{k+1}} | Helped |
Closest papers by meaning
- Neural Machine Translation with Joint Representation (AAAI 2020 · similarity 0.89)
- Nearest Neighbor Machine Translation (ICLR 2021 · similarity 0.88)
- Unsupervised Machine Translation Using Monolingual Corpora Only (ICLR 2018 · similarity 0.88)
- Scaling Laws for Multilingual Neural Machine Translation (ICML 2023 · similarity 0.88)
- A Paradigm Shift in Machine Translation: Boosting Translation Performance of Large Language Models (ICLR 2024 · similarity 0.88)
- Analyzing Uncertainty in Neural Machine Translation (ICML 2018 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.