Dataset Card for CANLI

Dataset Summary

CANLI: The Chinese Causative-Passive Homonymy Disambiguation: an Adversarial Dataset for NLI and a Probing Task

The disambiguation of causative-passive homonymy (CPH) is potentially tricky for machines, as the causative and the passive are not distinguished by the sentences syntactic structure. By transforming CPH disambiguation to a challenging natural language inference (NLI) task, we present the first Chinese Adversarial NLI challenge set (CANLI). We show that the pretrained transformer model RoBERTa, fine-tuned on an existing large-scale Chinese NLI benchmark dataset, performs poorly on CANLI. We also employ Word Sense Disambiguation as a probing task to investigate to what extent the CPH feature is captured in the models internal representation. We find that the models performance on CANLI does not correspond to its internal representation of CPH, which is the crucial linguistic ability central to the CANLI dataset.

Languages

Chinese Mandarin

Citation Information

@inproceedings{xu-markert-2022-chinese,
title = "The {C}hinese Causative-Passive Homonymy Disambiguation: an adversarial Dataset for {NLI} and a Probing Task",
author = "Xu, Shanshan  and Markert, Katja",
booktitle = "Proceedings of the Thirteenth Language Resources and Evaluation Conference",
month = jun,
year = "2022",
address = "Marseille, France",
publisher = "European Language Resources Association",
url = "https://aclanthology.org/2022.lrec-1.460",
pages = "4316--4323",

}