SNLI Indo : Dataset Recognizing Textual Entailment (RTE)

Published: 11 October 2023| Version 1 | DOI: 10.17632/k4tjhzs2gd.1


SNLI Indo is derived from the SNLI corpus, where each premise (P) and hypothesis (H) sentence is translated directly from English to Indonesian using the Google Cloud Translation API. This translation process is applied to both premise and hypothesis sentences, starting from index 0 to the last index for each data file (train, val, and test). This ensures that the same number of sentences as the original SNLI dataset, 570k sentence pairs, is obtained. A filtering process is carried out to remove incomplete sentence pairs and those with a gold label '-'. As a result, 569,030 sentence pairs are obtained.



Institut Teknologi Sepuluh Nopember


Data Science, Natural Language Processing