A rule-based appoach which outperforms almost all neural systems proposed in the shared task.
This script demonstrates how a simple rule-based entity-linking approach is able to achieve a remarkable result in SemEval 2018 Shared Task 4.
- Get SemEval 2018 Task 4 official dataset and evaluation script from here.
- Prepare a modified version of the original entity map text file where the characters are ranked by their frequency of occurrence in the training data.
- Copy the output to a text file, e.g.
result.txt - Run the official evaluation script with
-ref.out -result.txt