OCNLI: Original Chinese Natural Language Inference

TL;DR

OCNLI presents the first large-scale, human-elicited MNLI-style NLI dataset for Chinese, built from diverse Chinese genres without translation. By employing multi-hypothesis elicitation and expert annotators, the dataset reveals substantial gaps between state-of-the-art Chinese transformers and human performance, underscoring its difficulty. The work also analyzes biases, contrasts native data with translated resources like XNLI, and demonstrates that native data offer a clear advantage for Chinese NLI. The dataset and accompanying leaderboard are publicly released to accelerate progress in Chinese NLU and model probing, with future work focusing on bias reduction and deeper linguistic analysis.

Abstract

Despite the tremendous recent progress on natural language inference (NLI), driven largely by large-scale investment in new datasets (e.g., SNLI, MNLI) and advances in modeling, most progress has been limited to English due to a lack of reliable datasets for most of the world's languages. In this paper, we present the first large-scale NLI dataset (consisting of ~56,000 annotated sentence pairs) for Chinese called the Original Chinese Natural Language Inference dataset (OCNLI). Unlike recent attempts at extending NLI to other languages, our dataset does not rely on any automatic translation or non-expert annotation. Instead, we elicit annotations from native speakers specializing in linguistics. We follow closely the annotation protocol used for MNLI, but create new strategies for eliciting diverse hypotheses. We establish several baseline results on our dataset using state-of-the-art pre-trained models for Chinese, and find even the best performing models to be far outpaced by human performance (~12% absolute performance gap), making it a challenging new resource that we hope will help to accelerate progress in Chinese NLU. To the best of our knowledge, this is the first human-elicited MNLI-style corpus for a non-English language.

Figures (1)

• Figure 1: Ablation over the number of fine-tuning examples for RoBERTa fine-tuned on OCNLI vs. XNLI.