We propose a revised approach to NLPre evaluation via benchmarking, motivated by the widespread use of this technique in NLP fields that reveal shortcomings in existing solutions.
The benchmarking system evaluates the submitted outcomes of NLPre systems, updating the leaderboard with results post approval, ensuring trustworthiness in tool rankings.
The NLPre-PL benchmark encompasses various factors unique to Polish, including predefined NLPre tasks and reformulated datasets, setting a performance standard for tools.
NLPre-PL serves both for integration into the benchmarking system and for conducting empirical experiments, enriching the evaluation landscape for Polish NLP tools.
Collection
[
|
...
]