Papers by Mark Hepple
Igbo Diacritic Restoration using Embedding Models (N18-4)
Copied to clipboard
| Challenge: | Igbo is a low-resource language spoken by approximately 30 million people worldwide. |
| Approach: | They propose to use word embeddings to restore diacritics in Igbo by using a pre-processing task that replaces missing diacrittics on words from which they have been removed. |
| Outcome: | The embedding models performed better than n-gram models on the diacritic restoration task. |