Update README.md

README.md

@@ -45,7 +45,7 @@ python ./spert.py eval --config configs/example_eval.conf
 python ./spert.py predict --config configs/example_predict.conf
 ```
 ## Reproduction of Experimental Results
-- The final models were trained on the combined train+dev datasets.
+- The final models were trained on the combined train+dev datasets (e.g. 'conll04_train_dev.json').
 - Reproduction of SciERC results: To add a feature, the sampling of negative symmetric relations needed to be changed in commit [7b27b7d](https://github.com/lavis-nlp/spert/commit/7b27b7d258d0b4bb44103b9d0f9e19f2ce08611f). This lead to a slight improvement of experimental results for SciERC. Please use commit [3f4ab22](https://github.com/lavis-nlp/spert/commit/3f4ab22857f9ca0d96b582084a2a0ceb3e9826f9) if you want an exact reproduction of the ECAI 2020 paper results.