SART

SART 是鞑靼语的相似性、类比性和相关性三个数据集的集合。 这三个子集是： * 相似度数据集 - 202 对单词以及单词之间相似度的平均人类分数（0 到 10 级）。例如，“йорт, бина, 7.69”。 * 相关性数据集 - 252 对单词以及单词之间相关度的平均人类分数。例如，“урам，балалар，5.38”。 * 类比数据集 - 一组 A:B::C:D 形式的分析问题，意味着 A 到 B 就像 C 到 D，并且 D 是要预测的。例如，“Әнкара Төркия Париж Франция”。包含 34 个类别，共 30 144 个问题。SART is a collection of three datasets focused on similarity, analogy, and correlation in the Tatar language. The three subsets are as follows: * Similarity Dataset: Consists of 202 word pairs paired with average human-rated similarity scores on a scale of 0 to 10. An example entry is "йорт, бина, 7.69". * Correlation Dataset: Consists of 252 word pairs paired with average human-rated correlation scores. An example entry is "урам, балалар, 5.38". * Analogy Dataset: Comprises a set of analogy problems formatted as A:B::C:D, where the relationship between A and B is analogous to that between C and D, with D as the target to be predicted. An example entry is "Әнкара Төркия Париж Франция". This dataset includes 34 categories and a total of 30,144 problems.

搜集汇总

背景与挑战

背景概述

SART是一个鞑靼语的自然语言处理数据集，包含相似度、相关性和类比三个子集，用于评估词嵌入模型。具体包括202对单词的相似度评分、252对单词的相关性评分以及30,144个类比问题。