2026-10-14T21:59:11Z https://www.scidb.cn/oai

10.57760/sciencedb.j00104.00118 2026-09-10T10:58:15Z

2026-09-10 Amdo-Chinese Speech Translation Dataset doi:10.57760/sciencedb.j00104.00118 en The Amdo Tibetan-Chinese AST dataset (AMDO-CHS AST) is organized into two main folders, where 'f' designates female voice data, and 'm' designates male voice data. The audio data encompasses both broadcast and recorded segments, with the radio audio delivered by a female professional announcer. Within the recording data, female voice recordings were conducted by 3 individuals, while male voice recordings involved 5 contributors. The dataset comprises 3136 pieces of female voice audio data (including the announcer) and 3555 pieces of male voice audio data. Each main folder is further subdivided into two sub-folders. The 'wav' sub-folder contains audio data in *.wav format, totaling 6691 audio files. The 'text' sub-folder includes Tibetan text data (t.txt) and Chinese text data (c.txt), amounting to 12182 text files. The naming convention for each data piece follows the format: language type (Tibetan/Chinese)_recorder's initials (broadcast audio represented by 'ad')_text chapter number_sentence number in the chapter. For instance, "t_ztj_3_436.wav" denotes Tibetan data, recorded by an individual with the initials 'ztj', originating from Chapter 3, Sentence 436 of the corresponding text. The content of this data is "t_ztj_3_436:ཤུགས་ཆེ་བའི་སྐུལ་སློང་ནུས་ པ་མངོན་ཡོད།", corresponding to Chinese data "c_ztj_3_436:显示出强劲的动员作用". The entire data set comprises 11 hours of Amdo Tibetan audio data and corresponding Tibetan and Chinese text content.  Speech translation; Amdo Tibetan; Dataset 丽平 REN Zengzhuoma PUBLIC https://creativecommons.org/licenses/by/4.0/ dataset Science Data Bank