Semantic Role Labeling project that reached 88.8% F1-score on English, 77.04% F1 on French and 78.68% on Spanish. This was the top-scoring homework of the class. The dataset used in this task is the UniteD-SRL dataset, which is private. I developed it in Pytorch using Transformers, BiLSTMs and pos embeddings. This was the second homework of the NLP 2022 course at Sapienza University of Rome.
- Roberto Navigli
For more details, read the Report