Ofir Arviv, Ruixiang Cui, Daniel Hershcovich
This paper describes the HUJI-KU system submission to the shared task on Cross-Framework Meaning Representation Parsing (MRP) at the 2020 Conference for Computational Language Learning (CoNLL), employing TUPA and the HIT-SCIR parser, which were, respectively, the baseline system and winning system in the 2019 MRP shared task. Both are transition-based parsers using BERT contextualized embeddings. We generalized TUPA to support the newly-added MRP frameworks and languages, and experimented with multitask learning with the HIT-SCIR parser. We reached 4th place in both the cross-framework and cross-lingual tracks.
| Task | Dataset | Metric | Value | Model |
|---|---|---|---|---|
| Semantic Parsing | EDS (english, MRP 2020) | F1 | 80 | HUJI-KU |
| Semantic Parsing | DRG (german, MRP 2020) | F1 | 62 | HUJI-KU |
| Semantic Parsing | DRG (english, MRP 2020) | F1 | 63 | HUJI-KU |
| Semantic Parsing | PTG (czech, MRP 2020) | F1 | 58 | HUJI-KU |
| Semantic Parsing | PTG (english, MRP 2020) | F1 | 54 | HUJI-KU |
| Semantic Parsing | AMR (english, MRP 2020) | F1 | 52 | HUJI-KU |
| Semantic Parsing | UCCA (german, MRP 2020) | F1 | 75 | HUJI-KU |
| Semantic Parsing | UCCA (english, MRP 2020) | F1 | 73 | HUJI-KU |
| Semantic Parsing | AMR (chinese, MRP 2020) | F1 | 45 | HUJI-KU |