1 |
Improving Word Translation via Two-Stage Contrastive Learning ...
|
|
|
|
Abstract:
Word translation or bilingual lexicon induction (BLI) is a key cross-lingual task, aiming to bridge the lexical gap between different languages. In this work, we propose a robust and effective two-stage contrastive learning framework for the BLI task. At Stage C1, we propose to refine standard cross-lingual linear maps between static word embeddings (WEs) via a contrastive learning objective; we also show how to integrate it into the self-learning procedure for even more refined cross-lingual maps. In Stage C2, we conduct BLI-oriented contrastive fine-tuning of mBERT, unlocking its word translation capability. We also show that static WEs induced from the `C2-tuned' mBERT complement static WEs from Stage C1. Comprehensive experiments on standard BLI datasets for diverse languages and different experimental setups demonstrate substantial gains achieved by our framework. While the BLI method from Stage C1 already yields substantial gains over all state-of-the-art BLI methods in our comparison, even stronger ... : ACL 2022 Main ...
|
|
Keyword:
Artificial Intelligence cs.AI; Computation and Language cs.CL; FOS Computer and information sciences; Information Retrieval cs.IR; Machine Learning cs.LG
|
|
URL: https://arxiv.org/abs/2203.08307 https://dx.doi.org/10.48550/arxiv.2203.08307
|
|
BASE
|
|
Hide details
|
|
2 |
Plan-then-Generate: Controlled Data-to-Text Generation via Planning ...
|
|
|
|
BASE
|
|
Show details
|
|
3 |
Prix-LM: Pretraining for Multilingual Knowledge Base Construction ...
|
|
|
|
BASE
|
|
Show details
|
|
4 |
Learning Domain-Specialised Representations for Cross-Lingual Biomedical Entity Linking ...
|
|
|
|
BASE
|
|
Show details
|
|
5 |
MirrorWiC: On Eliciting Word-in-Context Representations from Pretrained Language Models ...
|
|
|
|
BASE
|
|
Show details
|
|
6 |
Visually Grounded Reasoning across Languages and Cultures ...
|
|
|
|
BASE
|
|
Show details
|
|
7 |
Fast, Effective, and Self-Supervised: Transforming Masked Language Models into Universal Lexical and Sentence Encoders ...
|
|
|
|
BASE
|
|
Show details
|
|
9 |
Measurement of $W^{\pm}$-boson and $Z$-boson production cross-sections in $pp$ collisions at $\sqrt{s}=2.76$ TeV with the ATLAS detector
|
|
|
|
BASE
|
|
Show details
|
|
10 |
A framework for streamlined statistical prediction using topic models ...
|
|
|
|
BASE
|
|
Show details
|
|
11 |
Fraud detection in telephone conversations for financial services using linguistic features ...
|
|
|
|
BASE
|
|
Show details
|
|
12 |
Which Other Race Effect?: Cross linguistic and Cross modal Asymmetries in Perceptual Narrowing ...
|
|
|
|
BASE
|
|
Show details
|
|
14 |
Unseen Word Representation by Aligning Heterogeneous Lexical Semantic Spaces ...
|
|
|
|
BASE
|
|
Show details
|
|
15 |
Factors affecting the motivation of EFL instructors living in South Korea to learn Korean
|
|
|
|
BASE
|
|
Show details
|
|
16 |
Research data supporting "Vancouver Welcomes You! Minimalist Location Metonymy Resolution" ...
|
|
|
|
BASE
|
|
Show details
|
|
17 |
Learning Rare Word Representations using Semantic Bridging ...
|
|
|
|
BASE
|
|
Show details
|
|
19 |
Observation of $B^0_s\to\bar{D}^0 K^0_S$ and evidence for $B^0_s\to\bar{D}^{*0} K^0_S$ decays
|
|
|
|
BASE
|
|
Show details
|
|
20 |
Translation of Book I of the "Chronicle of the Notable and Royal University of the City of Mexico and New Spain", with critical notes ...
|
|
|
|
BASE
|
|
Show details
|
|
|
|