Papers by Joshua Otten
Script-Agnosticism and its Impact on Language Identification for Dravidian Languages (2025.naacl-long)
Copied to clipboard
| Challenge: | a recent study shows that modern systems are script-dependent in language identification (langID) many languages are written in multiple writing systems, and script diversity is common in low-resource languages. |
| Approach: | They propose to learn script-agnostic representations using different strategies . they use word-level script randomization and script exposure to a language written in multiple scripts . |
| Outcome: | The proposed methods exploit script randomization and exposure to a language written in multiple scripts to improve language identification while maintaining competitive performance on naturally occurring text. |