Y-NQ: English-Yor\`ub\'a Evaluation dataset for Open-Book Reading Comprehension and Text Generation

TL;DR

This paper introduces a bilingual English-Yorùbá dataset for evaluating reading comprehension and text generation, revealing significant performance disparities between the languages, especially with longer documents, and highlighting challenges for LLMs in low-resource languages.

Contribution

It provides a novel English-Yorùbá evaluation dataset for open-book reading comprehension and text generation, enabling assessment of model performance across high- and low-resource languages.

Findings

Yorùbá performance lags behind English on automatic metrics.

Performance drops by 2.5 times for Yorùbá with comparable document lengths.

Yorùbá performance declines sharply at 1500 words, unlike English.

Abstract

The purpose of this work is to share an English-Yor\`ub\'a evaluation dataset for open-book reading comprehension and text generation to assess the performance of models both in a high- and a low- resource language. The dataset contains 358 questions and answers on 338 English documents and 208 Yor\`ub\'a documents. The average document length is ~ 10k words for English and 430 words for Yor\`ub\'a. Experiments show a consistent disparity in performance between the two languages, with Yor\`ub\'a falling behind English for automatic metrics even if documents are much shorter for this language. For a small set of documents with comparable length, performance of Yor\`ub\'a drops by x2.5 times. When analyzing performance by length, we observe that Yor\`ub\'a decreases performance dramatically for documents that reach 1500 words while English performance is barely affected at that length. Our dataset opens the door to showcasing if English LLM reading comprehension capabilities extend to Yor\`ub\'a, which for the evaluated LLMs is not the case.