--- dataset_info: - config_name: multispeaker-ach features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 63849387.25 num_examples: 4830 - name: dev num_bytes: 1338106.0 num_examples: 101 - name: test num_bytes: 1349215.0 num_examples: 96 download_size: 66406423 dataset_size: 66536708.25 - config_name: multispeaker-eng features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 56568803.5 num_examples: 4804 - name: dev num_bytes: 1183267.0 num_examples: 101 - name: test num_bytes: 1256266.0 num_examples: 96 download_size: 58835221 dataset_size: 59008336.5 - config_name: multispeaker-lgg features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 86777842.25 num_examples: 4814 - name: dev num_bytes: 1774879.0 num_examples: 101 - name: test num_bytes: 1841950.0 num_examples: 96 download_size: 90251520 dataset_size: 90394671.25 - config_name: multispeaker-lug features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 66308098.75 num_examples: 5002 - name: dev num_bytes: 1354874.0 num_examples: 103 - name: test num_bytes: 1414079.0 num_examples: 99 download_size: 68947353 dataset_size: 69077051.75 - config_name: multispeaker-nyn features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 78167853.5 num_examples: 4884 - name: dev num_bytes: 1630828.0 num_examples: 103 - name: test num_bytes: 1679432.0 num_examples: 99 download_size: 81242972 dataset_size: 81478113.5 - config_name: multispeaker-teo features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 71561845.0 num_examples: 4904 - name: dev num_bytes: 1445165.0 num_examples: 101 - name: test num_bytes: 1472470.0 num_examples: 98 download_size: 73751780 dataset_size: 74479480.0 - config_name: studio-ach features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 203557155.625 num_examples: 4803 - name: dev num_bytes: 4236003.0 num_examples: 101 - name: test num_bytes: 4163362.0 num_examples: 96 download_size: 207458547 dataset_size: 211956520.625 - config_name: studio-eng features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 75859866.875 num_examples: 2425 - name: dev num_bytes: 1593216.0 num_examples: 50 - name: test num_bytes: 1343598.0 num_examples: 43 download_size: 78248223 dataset_size: 78796680.875 - config_name: studio-lgg features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 345723041.75 num_examples: 1594 - name: dev num_bytes: 6802603.0 num_examples: 32 - name: test num_bytes: 5878605.0 num_examples: 28 download_size: 357905379 dataset_size: 358404249.75 - config_name: studio-lug features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 64679184.625 num_examples: 2395 - name: dev num_bytes: 1382546.0 num_examples: 50 - name: test num_bytes: 1183029.0 num_examples: 43 download_size: 66576859 dataset_size: 67244759.625 - config_name: studio-nyn features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 178272124.25 num_examples: 3326 - name: dev num_bytes: 3764345.0 num_examples: 72 - name: test num_bytes: 3550903.0 num_examples: 62 download_size: 184867186 dataset_size: 185587372.25 - config_name: studio-teo features: - name: id dtype: int64 - name: text dtype: string - name: audio dtype: audio - name: audio_language dtype: string - name: is_studio dtype: bool - name: speaker_id dtype: int64 - name: sample_rate dtype: int64 splits: - name: train num_bytes: 265092538.875 num_examples: 4649 - name: dev num_bytes: 5706033.0 num_examples: 99 - name: test num_bytes: 5371919.0 num_examples: 92 download_size: 273498778 dataset_size: 276170490.875 - config_name: text-all features: - name: id dtype: int64 - name: teo_text dtype: string - name: swa_text dtype: string - name: eng_text dtype: string - name: nyn_text dtype: string - name: ibo_text dtype: string - name: ach_text dtype: string - name: lgg_text dtype: string - name: lug_text dtype: string splits: - name: train num_bytes: 11763775 num_examples: 23947 - name: dev num_bytes: 242587 num_examples: 496 - name: test num_bytes: 253968 num_examples: 500 download_size: 7228279 dataset_size: 12260330 - config_name: text-hard features: - name: id dtype: int64 - name: eng_text dtype: string - name: xog_text dtype: string - name: ttj_text dtype: string splits: - name: train num_bytes: 93291 num_examples: 491 download_size: 67805 dataset_size: 93291 configs: - config_name: multispeaker-ach data_files: - split: train path: multispeaker-ach/train-* - split: dev path: multispeaker-ach/dev-* - split: test path: multispeaker-ach/test-* - config_name: multispeaker-eng data_files: - split: train path: multispeaker-eng/train-* - split: dev path: multispeaker-eng/dev-* - split: test path: multispeaker-eng/test-* - config_name: multispeaker-lgg data_files: - split: train path: multispeaker-lgg/train-* - split: dev path: multispeaker-lgg/dev-* - split: test path: multispeaker-lgg/test-* - config_name: multispeaker-lug data_files: - split: train path: multispeaker-lug/train-* - split: dev path: multispeaker-lug/dev-* - split: test path: multispeaker-lug/test-* - config_name: multispeaker-nyn data_files: - split: train path: multispeaker-nyn/train-* - split: dev path: multispeaker-nyn/dev-* - split: test path: multispeaker-nyn/test-* - config_name: multispeaker-teo data_files: - split: train path: multispeaker-teo/train-* - split: dev path: multispeaker-teo/dev-* - split: test path: multispeaker-teo/test-* - config_name: studio-ach data_files: - split: train path: studio-ach/train-* - split: dev path: studio-ach/dev-* - split: test path: studio-ach/test-* - config_name: studio-eng data_files: - split: train path: studio-eng/train-* - split: dev path: studio-eng/dev-* - split: test path: studio-eng/test-* - config_name: studio-lgg data_files: - split: train path: studio-lgg/train-* - split: dev path: studio-lgg/dev-* - split: test path: studio-lgg/test-* - config_name: studio-lug data_files: - split: train path: studio-lug/train-* - split: dev path: studio-lug/dev-* - split: test path: studio-lug/test-* - config_name: studio-nyn data_files: - split: train path: studio-nyn/train-* - split: dev path: studio-nyn/dev-* - split: test path: studio-nyn/test-* - config_name: studio-teo data_files: - split: train path: studio-teo/train-* - split: dev path: studio-teo/dev-* - split: test path: studio-teo/test-* - config_name: text-all data_files: - split: train path: text-all/train-* - split: dev path: text-all/dev-* - split: test path: text-all/test-* - config_name: text-hard data_files: - split: train path: text-hard/train-* license: cc-by-sa-4.0 task_categories: - automatic-speech-recognition - text-to-speech - translation language: - ach - eng - lgg - lg - nyn - teo - ttj - sw - xog pretty_name: Sunbird African Language Technology (SALT) dataset --- # Sunbird African Language Technology (SALT) dataset SALT is a multi-way parallel text and speech corpus of Engish and six languages widely spoken in Uganda and East Africa: Luganda, Lugbara, Acholi, Runyankole, Ateso and Swahili. The core of the dataset is a set of 25,000 sentences covering a range of topics of local relevance, such as agriculture, health and society. Each sentence is translated into all languages, to support machine translation, and speech recordings are made for approximately 5,000 of the sentences both by a variety of speakers in natural settings (suitable for ASR) and by professionals in a studio setting (suitable for text-to-speech). ## Subsets | Subset name | Contents | | --------- | ------- | | text-all | Text translations of each sentence. | | text-hard | Text translations of each sentence. | | multispeaker-`{lang}` | Speech recordings of each sentence, by a variety of speakers in natural settings. | | studio-`{lang}` | Speech recordings in a studio setting, suitable for text-to-speech. | The sentence IDs map across subsets, so that for example the text of a sentence in Acholi can be mapped to the studio recording of that concept being expressed in Swahili. The subsets can therefore be combined to support the training and evaluation of several further tasks, such as speech-to-text translation and speech-to-speech translation. ## Language support | ISO 639-3 | Language | Translated text | Multispeaker speech | Studio speech | | --- | --- | --- | --- | --- | | eng | English (Ugandan accent) | Yes | Yes | Yes | | lug | Luganda | Yes | Yes | Yes | | ach | Acholi | Yes | Yes | Yes | | lgg | Lugbara | Yes | Yes | Yes | | teo | Ateso | Yes | Yes | Yes | | nyn | Runyankole | Yes | Yes | Yes | | swa | Swahili | Yes | No | Yes | | ibo | Igbo | Yes | No | No | ## Helper utilities Code for convenient experimentation with multilingual models can be found at [https://github.com/SunbirdAI/salt](https://github.com/SunbirdAI/salt). See example notebooks [here](https://github.com/SunbirdAI/salt/tree/main/notebooks). ## Collaborators This dataset was collected in practical collaboration between Sunbird AI and the Makerere University AI Lab (Ugandan languages) and KenCorpus, Maseno University (Swahili). ## Reference [Machine Translation For African Languages: Community Creation Of Datasets And Models In Uganda](https://openreview.net/pdf?id=BK-z5qzEU-9). Benjamin Akera, Jonathan Mukiibi, Lydia Sanyu Naggayi, Claire Babirye, Isaac Owomugisha, Solomon Nsumba, Joyce Nakatumba-Nabende, Engineer Bainomugisha, Ernest Mwebaze, John Quinn. 3rd Workshop on African Natural Language Processing, 2022.