Amharic Speech Emotion Dataset (ASED)

Amharic Speech Emotion Dataset (ASED) 是首个针对阿姆哈拉语的语音情感识别数据集，由西北大学信息科学与技术学院创建。该数据集包含来自65名母语为阿姆哈拉语的志愿者的2,474个声音样本，覆盖四种方言（Gojjam, Wollo, Shewa, Gonder）和五种不同情感（中性、恐惧、快乐、悲伤、愤怒）。数据集中的每个样本长度为两到四秒，由八位评委进行情感标注，标注一致性高（Fleiss kappa = 0.8）。ASED数据集旨在推动阿姆哈拉语情感识别的研究，并探索其他模型在该语言上的应用。

Amharic Speech Emotion Dataset (ASED) is the first speech emotion recognition dataset dedicated to Amharic, developed by the School of Information Science and Technology, Northwest University. This dataset contains 2,474 audio samples from 65 native Amharic speakers, covering four dialects (Gojjam, Wollo, Shewa, Gonder) and five distinct emotions: neutral, fear, happiness, sadness, and anger. Each sample in the dataset has a duration of 2 to 4 seconds, and was annotated by eight raters with high inter-annotator consistency (Fleiss' kappa = 0.8). The ASED dataset aims to advance research on Amharic speech emotion recognition and explore the application of other models to this language.