--- language: - en - de - es - fr - es - zh license: - cc-by-nc-sa-3.0 multilinguality: - multilingual task_categories: - automatic-speech-recognition - text-to-speech - text-to-audio pretty_name: talkbank_4_stt dataset_info: features: - name: audio struct: - name: bytes dtype: binary - name: path dtype: string - name: transcript dtype: string - name: language_code dtype: string - name: subset dtype: string - name: full_language dtype: string - name: switch_id dtype: string - name: segment_id dtype: string - name: transcript_filename dtype: string - name: audio_len_sec dtype: int64 - name: orig_file_start dtype: float64 - name: orig_file_end dtype: float64 - name: channel dtype: int64 splits: - name: de_switch_train num_bytes: 370364837 num_examples: 217 - name: de_switch_test num_bytes: 66355623 num_examples: 37 - name: de_segment_train num_bytes: 132668099 num_examples: 10355 - name: de_segment_test num_bytes: 24165594 num_examples: 1847 - name: ja_switch_train num_bytes: 615411743 num_examples: 273 - name: ja_switch_test num_bytes: 128442494 num_examples: 52 - name: ja_segment_train num_bytes: 109828966 num_examples: 14620 - name: ja_segment_test num_bytes: 23579006 num_examples: 3153 - name: zh_switch_train num_bytes: 435453051 num_examples: 314 - name: zh_switch_test num_bytes: 81031328 num_examples: 62 - name: zh_segment_train num_bytes: 13593465 num_examples: 1943 - name: zh_segment_test num_bytes: 2209388 num_examples: 288 - name: en_switch_train num_bytes: 733609978 num_examples: 369 - name: en_switch_test num_bytes: 118207856 num_examples: 68 - name: en_segment_train num_bytes: 345948745 num_examples: 30736 - name: en_segment_test num_bytes: 52235458 num_examples: 4166 - name: es_switch_train num_bytes: 1447559659 num_examples: 488 - name: es_switch_test num_bytes: 225061634 num_examples: 89 - name: es_segment_train num_bytes: 662759917 num_examples: 55213 - name: es_segment_test num_bytes: 95103746 num_examples: 7743 - name: fr_switch_train num_bytes: 426622783 num_examples: 93 - name: fr_switch_test num_bytes: 61436995 num_examples: 17 - name: fr_segment_train num_bytes: 223181171 num_examples: 18639 - name: fr_segment_test num_bytes: 33991480 num_examples: 3002 download_size: 13250672382 dataset_size: 6428823016 configs: - config_name: default data_files: - split: de_switch_train path: data/de_switch_train-* - split: de_switch_test path: data/de_switch_test-* - split: de_segment_train path: data/de_segment_train-* - split: de_segment_test path: data/de_segment_test-* - split: ja_switch_train path: data/ja_switch_train-* - split: ja_switch_test path: data/ja_switch_test-* - split: ja_segment_train path: data/ja_segment_train-* - split: ja_segment_test path: data/ja_segment_test-* - split: zh_switch_train path: data/zh_switch_train-* - split: zh_switch_test path: data/zh_switch_test-* - split: zh_segment_train path: data/zh_segment_train-* - split: zh_segment_test path: data/zh_segment_test-* - split: en_switch_train path: data/en_switch_train-* - split: en_switch_test path: data/en_switch_test-* - split: en_segment_train path: data/en_segment_train-* - split: en_segment_test path: data/en_segment_test-* - split: es_switch_train path: data/es_switch_train-* - split: es_switch_test path: data/es_switch_test-* - split: es_segment_train path: data/es_segment_train-* - split: es_segment_test path: data/es_segment_test-* - split: fr_switch_train path: data/fr_switch_train-* - split: fr_switch_test path: data/fr_switch_test-* - split: fr_segment_train path: data/fr_segment_train-* - split: fr_segment_test path: data/fr_segment_test-* --- # Dataset Card ## Dataset Description This dataset is a benchmark based on the TalkBank[1] corpus—a large multilingual repository of conversational speech that captures real-world, unstructured interactions. We use CA-Bank [2], which focuses on phone conversations between adults, which include natural speech phenomena such as laughter, pauses, and interjections. To ensure the dataset is highly accurate and suitable for benchmarking conversational ASR systems, we employ extensive set of pre-processing. ## Preprocessing Steps We apply the following preprocessing steps to ensure the dataset’s quality: - Manual filtering of conversations - Speaker-channel alignment - Timestamp alignment using voice activity detection (VAD) - Discarding segments based on Word Error Rate (WER) thresholds ## Paper and Code Repository For a comprehensive explanation of the preprocessing pipeline and dataset details, refer to our paper [ASR Benchmarking: The Need for a More Representative Conversational Dataset](https://arxiv.org/abs/2409.12042) and explore our [GitHub repository](https://github.com/Diabolocom-Research/ConversationalDataset) for code and additional resources. ## Segmentation Types: Speaker Switch vs Annotation We offer two types of segmentation for this dataset: - **Annotation-based Segmentation**: Segments are derived directly from the annotations provided in the original TalkBank corpus. - **Speaker Switch Segmentation**: We consolidate consecutive segments from the same speaker into a single, larger audio segment, providing an alternative structure for analysis. ## Citations While using this dataset please cite: ``` @article{maheshwari2024asr, title={ASR Benchmarking: Need for a More Representative Conversational Dataset}, author={Maheshwari, Gaurav and Ivanov, Dmitry and Johannet, Th{\'e}o and Haddad, Kevin El}, journal={arXiv preprint arXiv:2409.12042}, year={2024} } ``` In addition, please acknowledge the TalkBank dataset:: ``` @article{macwhinney2010transcribing, title={Transcribing, searching and data sharing: The CLAN software and the TalkBank data repository}, author={MacWhinney, Brian and Wagner, Johannes}, journal={Gesprachsforschung: Online-Zeitschrift zur verbalen Interaktion}, volume={11}, pages={154}, year={2010}, publisher={NIH Public Access} } ``` ## Licensing Information This dataset is released under the [CC BY-NC-SA 3.0](https://creativecommons.org/licenses/by-nc-sa/3.0). ## References [1]: MacWhinney, Brian. "TalkBank: Building an open unified multimodal database of communicative interaction." (2004). [2]: MacWhinney, Brian, and Johannes Wagner. "Transcribing, searching and data sharing: The CLAN software and the TalkBank data repository." Gesprachsforschung: Online-Zeitschrift zur verbalen Interaktion 11 (2010): 154.