FAR-Berkeley-Parser contains grammar models and resources derived from training the Berkeley Parser on the Faroese Parsed Historical Corpus (FarPaHC). The repository includes the combined Faroese Parsed Historical Corpus file (FARPAHC-Singlefile.txt), as well as the grammar file and related outputs generated using the Berkeley Parser’s GrammarTrainer.java. Additional files for grammar merging, smoothing, splitting, and text-based outputs are also provided.
The grammar’s accuracy was evaluated using the Berkeley Parser’s GrammarTester.java script, which produced high F1 scores of 94.73% (current) and 98.14% (average). While these scores suggest strong performance, potential anomalies in the process might have affected the results, warranting further investigation.
Release: 2022
Contact: annika@hi.is
