Why Overfitting Isn't Always Bad: Retrofitting Cross-Lingual Word Embeddings to Dictionaries
This work addresses a limitation in CLWE evaluation for downstream tasks, showing that overfitting can be beneficial, but it is incremental as it builds on existing CLWE methods.
The paper tackles the problem that cross-lingual word embeddings (CLWE) methods underfit training dictionaries to generalize on bilingual lexicon induction (BLI), which can hinder performance on downstream tasks. By retrofitting CLWE to the training dictionary to overfit it, they often improve accuracy on two downstream tasks, despite lowering BLI test accuracy.
Cross-lingual word embeddings (CLWE) are often evaluated on bilingual lexicon induction (BLI). Recent CLWE methods use linear projections, which underfit the training dictionary, to generalize on BLI. However, underfitting can hinder generalization to other downstream tasks that rely on words from the training dictionary. We address this limitation by retrofitting CLWE to the training dictionary, which pulls training translation pairs closer in the embedding space and overfits the training dictionary. This simple post-processing step often improves accuracy on two downstream tasks, despite lowering BLI test accuracy. We also retrofit to both the training dictionary and a synthetic dictionary induced from CLWE, which sometimes generalizes even better on downstream tasks. Our results confirm the importance of fully exploiting training dictionary in downstream tasks and explains why BLI is a flawed CLWE evaluation.