Search:
Match:
1 results

Analysis

This paper addresses the lack of a comprehensive benchmark for Turkish Natural Language Understanding (NLU) and Sentiment Analysis. It introduces TrGLUE, a GLUE-style benchmark, and SentiTurca, a sentiment analysis benchmark, filling a significant gap in the NLP landscape. The creation of these benchmarks, along with provided code, will facilitate research and evaluation of Turkish NLP models, including transformers and LLMs. The semi-automated data creation pipeline is also noteworthy, offering a scalable and reproducible method for dataset generation.
Reference

TrGLUE comprises Turkish-native corpora curated to mirror the domains and task formulations of GLUE-style evaluations, with labels obtained through a semi-automated pipeline that combines strong LLM-based annotation, cross-model agreement checks, and subsequent human validation.