This repository contains the code used by the D2KLab team for SemEval Task 2: MultiCoNER. The goal of this task is to detect semantically ambiguous and complex entities in short and low-context settings.
Each folder represents the year we have participated and the experiements we have conducted:
We evaluated our fine-tuned language model on the test sets of the MultiCoNER datasets. The results are summarized in the tables below:
Method | BN | DE | EN | ES | FA | HI | KO | NL | RU | TR | ZH | MIX | MULTI | Avg. |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Official Baseline | 39.41 | 63.74 | 61.36 | 57.84 | 52.24 | 48.22 | 55.25 | 62.01 | 59.59 | 46.25 | 63.4 | 58.14 | 48.22 | 47.78 |
D2KLab | 52.57 | 67.09 | 74.57 | 62.77 | 55.91 | 52.78 | 64.96 | 66.7 | 68.21 | 56.57 | 54.92 | 77.6 | 52.78 | 68.08 |
Method | BN | DE | EN | ES | FA | FR | HI | IT | PT | SV | UK | ZH | Avg. |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Official Baseline | 1.07 | 64.61 | 36.97 | 49.07 | 41.28 | 41.39 | 2.89 | 43.13 | 39.85 | 69.22 | 62.08 | 48.46 | 41.67 |
D2KLab | 61.43 | 67.09 | 61.29 | 63.17 | 54.2 | 64.09 | 63.29 | 64.77 | 60.79 | 62.98 | 64.14 | 54.92 | 61.84 |