Universal Dependency Parsing on SpokeBench Discourse v1 (test)
79ID AccuracyBiLingua Parser
Evaluation results
| Method | Links | |||||
|---|---|---|---|---|---|---|
| BiLingua ParserType=LLM-based pipeline2026.02 | 79 | 79.5 | 55.8 | 60 | 70.6 | |
| DECAPBackbone=GPT-4.12026.02 | 63.5 | 83 | 52.5 | 57 | 66.7 | |
| Bilingual Traditional ParserBackbone=Stanza2026.02 | 10.7 | 10.7 | 10.7 | 10.7 | 25.9 |