Entity Linking in 100 Languages
This work addresses entity linking for low-resource languages and rare entities, though it is incremental with improvements in feature representation and training techniques.
The authors tackled multilingual entity linking across 100+ languages and 20 million entities by proposing a new formulation and training a dual encoder model, which outperformed state-of-the-art results in cross-lingual linking tasks.
We propose a new formulation for multilingual entity linking, where language-specific mentions resolve to a language-agnostic Knowledge Base. We train a dual encoder in this new setting, building on prior work with improved feature representation, negative mining, and an auxiliary entity-pairing task, to obtain a single entity retrieval model that covers 100+ languages and 20 million entities. The model outperforms state-of-the-art results from a far more limited cross-lingual linking task. Rare entities and low-resource languages pose challenges at this large-scale, so we advocate for an increased focus on zero- and few-shot evaluation. To this end, we provide Mewsli-9, a large new multilingual dataset (http://goo.gle/mewsli-dataset) matched to our setting, and show how frequency-based analysis provided key insights for our model and training enhancements.