Datasets:
Formats:
parquet
Size:
< 1K
Tags:
meeting-audio
real-world-audio
conversational-audio
speech-data
multi-speaker-audio
speech-recognition
License:
| pretty_name: Real-World Meeting Audio Dataset | |
| tags: | |
| - meeting-audio | |
| - real-world-audio | |
| - conversational-audio | |
| - speech-data | |
| - multi-speaker-audio | |
| - speech-recognition | |
| - speaker-recognition | |
| - audio-understanding | |
| - speech-processing | |
| - natural-conversation | |
| - overlapping-speech | |
| - multilingual-audio | |
| - arabic | |
| - chinese | |
| - english | |
| - french | |
| - german | |
| - italian | |
| - japanese | |
| - portuguese | |
| - spanish | |
| - wav | |
| - artificial-intelligence | |
| - machine-learning | |
| license: cc-by-4.0 | |
| # Real-World Meeting Audio Dataset | |
| ## Overview | |
| The **Real-World Meeting Audio Dataset** is a large-scale collection of real-world, multi-speaker meeting conversations featuring natural discussions, interruptions, and overlapping speech. | |
| With **50K+ hours of audio**, the dataset covers nine language categories: Arabic, Chinese, English, French, German, Italian, Japanese, Portuguese, and Spanish. | |
| The recordings capture natural conversational interactions in meeting environments, providing audio data relevant to speech processing, multilingual audio understanding, conversational AI, and machine learning research. | |
| The dataset also includes rich metadata, with speaker timestamps, diarization, transcripts, and meeting context identified as part of its metadata coverage. | |
| Audio recordings are provided in WAV format. | |
| ## Dataset Highlights | |
| - **50K+ Hours of Audio:** Large-scale real-world meeting audio collection. | |
| - **Multi-Speaker Conversations:** Recordings featuring multiple speakers. | |
| - **Natural Discussions:** Real-world conversational interactions. | |
| - **Overlapping Speech:** Includes conversations with overlapping speech and interruptions. | |
| - **Nine Language Categories:** Covers Arabic, Chinese, English, French, German, Italian, Japanese, Portuguese, and Spanish. | |
| - **Rich Metadata:** Includes speaker-related and meeting-related information. | |
| - **WAV Audio Format:** Audio recordings provided in WAV format. | |
| - **Sample Availability:** Samples are available for exploring the dataset. | |
| ## Language Coverage | |
| The dataset is organized into nine language categories: | |
| | No. | Language | | |
| |---:|---| | |
| | 1 | Arabic | | |
| | 2 | Chinese | | |
| | 3 | English | | |
| | 4 | French | | |
| | 5 | German | | |
| | 6 | Italian | | |
| | 7 | Japanese | | |
| | 8 | Portuguese | | |
| | 9 | Spanish | | |
| ## Dataset Features | |
| | Feature | Description | | |
| |---|---| | |
| | Dataset name | Real-World Meeting Audio Dataset | | |
| | Dataset type | Real-world meeting audio | | |
| | Audio format | WAV | | |
| | Audio duration | 50K+ hours | | |
| | Conversation type | Multi-speaker meetings | | |
| | Speech characteristics | Natural discussions, interruptions, overlapping speech | | |
| | Language categories | 9 | | |
| | Metadata | Rich meeting and speaker-related metadata | | |
| | Samples | Available | | |
| ## Audio Characteristics | |
| The dataset captures conversational characteristics found in real-world meetings, including: | |
| - Multi-speaker discussions | |
| - Natural conversational exchanges | |
| - Speaker interruptions | |
| - Overlapping speech | |
| - Meeting-based conversations | |
| - Real-world speaking interactions | |
| ## Metadata Coverage | |
| The dataset includes rich metadata covering: | |
| - Speaker timestamps | |
| - Speaker diarization | |
| - Transcripts | |
| - Meeting context | |
| ## Supported AI and Research Tasks | |
| The dataset may be useful for research and development involving: | |
| - Automatic speech recognition (ASR) | |
| - Multilingual speech processing | |
| - Multi-speaker audio understanding | |
| - Speaker-related audio analysis | |
| - Conversational AI | |
| - Meeting audio analysis | |
| - Speech segmentation research | |
| - Overlapping speech analysis | |
| - Audio classification | |
| - Natural language and speech technology research | |
| ## Applications | |
| Potential applications include: | |
| - AI-powered meeting analysis | |
| - Multilingual speech recognition systems | |
| - Conversational AI development | |
| - Speech processing research | |
| - Meeting transcription research | |
| - Multi-speaker conversation analysis | |
| - Audio understanding systems | |
| - Speech technology development | |
| - Machine learning experimentation | |
| - Research into natural conversational interactions | |
| ## Dataset Format | |
| The confirmed audio file format is: | |
| | Extension | Content | | |
| |---|---| | |
| | `.wav` | Audio recordings | | |
| ## Sample Availability | |
| Samples are available for users who want to explore the Real-World Meeting Audio Dataset. | |
| The sample availability provides an opportunity to review the audio content and understand the characteristics of real-world multi-speaker meeting conversations. | |
| ## Dataset Statistics | |
| | Metric | Dataset Information | | |
| |---|---| | |
| | Total audio duration | 50K+ hours | | |
| | Language categories | 9 | | |
| | Conversation type | Multi-speaker meetings | | |
| | Audio format | WAV | | |
| ## Intended Audience | |
| This dataset may be relevant to: | |
| - AI and machine learning researchers | |
| - Speech recognition developers | |
| - Audio processing researchers | |
| - Conversational AI developers | |
| - Natural language processing researchers | |
| - Multilingual AI teams | |
| - Speech technology companies | |
| - Academic institutions | |
| - Enterprise AI development teams | |
| - Audio analytics developers | |
| ## Notes and Limitations | |
| - The dataset contains real-world multi-speaker meeting conversations. | |
| - Audio recordings are provided in WAV format. | |
| - Nine language categories are represented. | |
| - Conversations include natural discussions, interruptions, and overlapping speech. | |
| - The reported audio duration is presented as 50K+ hours. | |
| - Specific language-wise duration distributions and additional file-level metadata details are not specified in this README. | |
| ## Summary | |
| The Real-World Meeting Audio Dataset is a large-scale multilingual audio collection featuring 50K+ hours of real-world meeting conversations across nine language categories. | |
| With multi-speaker discussions, natural interruptions, overlapping speech, WAV audio recordings, and rich metadata, the dataset is relevant to speech recognition, multilingual audio processing, conversational AI, and meeting audio understanding research. | |
| **Data Creation** | |
| Procured through formal agreements and generated in the ordinary course of business. | |
| **Considerations** | |
| This dataset is provided for research and educational purposes only. It contains only sample data. For access to the full dataset and enterprise licensing options, please visit our website [InfoBay.AI](https://infobay.ai/) or contact us directly. | |
| -Ph: +91 9236398619 | |
| -Email: datareq@infobay.ai |