To test the performance of the NER and/or NEL, it is necessary to create a so-called gold standard (ground truth).

In speech processing, ground truth or goldstandard refers to a (manually) corrected, ideal data set that is used for testing and training.

In the SEMOHI project seven gold standard articles were produced, which were manually corrected and tagged in accordance with TEI to evaluate our workflows for automatic name recognition and linking. The following graphic shows the manually identified names of places in the gold standard article “Ski tours in the Ötztal Alps”