6/29/2026
Qlean Dataset Releases Japanese 3-Speaker Business Dialogue Audio Dataset with Human-Created Transcripts
Visual Bank Co., Ltd. (Minato-ku, Tokyo; CEO: Saneyuki Nagai) announces the release of the "Japanese 3-Speaker Business Theme Dialogue Audio & Transcript Dataset" through its subsidiary amana images Inc., under its AI training data solution Qlean Dataset. The dataset pairs 13 groups of 3-speaker business dialogue audio, recorded via web conference, with high-quality human-created transcripts for use in ASR domain adaptation, multi-speaker speech recognition, and LLM development.
■ What Is a Business Theme Dialogue Audio Dataset?
A business theme dialogue audio dataset is a speech corpus of natural multi-speaker conversations in business contexts such as investment, insurance, and client meetings, used for ASR domain adaptation, multi-speaker speech recognition, and LLM business dialogue training.
■ Overview of the "Japanese 3-Speaker Business Theme Dialogue Audio & Transcript Dataset"
This dataset captures spontaneous 3-speaker business conversations among 13 groups of Japanese speakers in a web conferencing environment, covering topics such as investment and insurance. All transcripts are human-created rather than auto-generated, eliminating errors in financial terminology, missing fillers, and speaker boundary misalignment — ensuring reliable WER/CER evaluation and clean training data.
Data Type | Audio (3-speaker dialogue format) |
|---|---|
Speakers | Japanese speakers with diversity in gender and age (13 groups) |
Duration / Volume | Approx. 25 hours (63 files) / Approx. 55GB |
File Format | mp3 |
Sampling / Bitrate | 48kHz / 192kbps, stereo |
Content | 3-speaker business dialogue via web conference (investment, insurance, etc.); approx. 90 min. per session |
Transcripts | Human-created for high accuracy and quality assurance |
Usage Rights | Commercial use permitted · Research use permitted · Free for academic use (see website for details) |
Sample data available at: https://qleandataset.visual-bank.co.jp/en/lineup/ds-050
■ Frequently Asked Questions (FAQ)
Q. How does a 3-speaker dataset differ from 2-speaker data?
A. Three-speaker turn-taking patterns more closely replicate real business meeting environments, making it effective for evaluating ASR model generalization beyond 2-speaker data.
Q. Can this be used for ASR or LLM development in finance and insurance?
A. Yes. Human-created transcripts paired with investment and insurance dialogue enable domain adaptation fine-tuning (e.g., LoRA for Whisper) and SFT/evaluation data for finance-specific LLMs, without the errors common in automated transcription.
Q. Is this suitable for meeting summarization or minute-generation AI?
A. Yes. Approximately 90-minute sessions with human-created transcripts make it well-suited for summarization, minute generation, and action item extraction SFT data with long-context processing requirements.
Q. Can you accommodate custom recording requests?
A. Yes. We support custom data collection by industry (legal, healthcare, sales, etc.), speaker attributes, and dialogue scenario design.
■ Use Case Examples
Domain Adaptation Fine-Tuning for Business ASR
Paired audio and human-created transcripts enable LoRA or full fine-tuning of Whisper and ESPnet for business domains, with reliable WER/CER evaluation free from automated transcription noise.Multi-Speaker ASR Performance Evaluation
Three-speaker dialogue covering turn-taking, overlapping speech, and fillers enables stress-testing of ASR robustness against complex patterns not reproducible with 2-speaker datasets.LLM Fine-Tuning for Business Dialogue Summarization
Human-created transcripts support SFT data construction for summarization, minute generation, and action item extraction, with approximately 90-minute sessions ideal for long-context processing tasks.
About Qlean Dataset
Qlean Dataset is a commercially licensed AI training data solution provided by amanaimages Inc., a wholly owned subsidiary of Visual Bank. All datasets are rights-cleared for commercial use, giving AI developers a legally secure environment to source and deploy high-quality training data.
The platform covers audio, image, video, 3D, and text modalities — serving foundation model developers and applied AI teams alike. Through partnerships with domestic and international data holders, broadcasters, newspapers, and newswire agencies, Qlean Dataset continuously expands its AI Data Recipe lineup of industry-specific, trend-driven datasets. Existing datasets ship within 2 business days; custom recording and data collection are also available on request.
URL:https://qleandataset.visual-bank.co.jp/en
URL:https://qleandataset.visual-bank.co.jp/en/products/japanese-language-corpora
Contact
About Visual Bank Inc.
Visual Bank Group is a technology company developing data infrastructure and AI solutions that support advanced AI development. The company operates THE PEN, an AI tool for manga creators, and its subsidiary, amanaimages Inc., provides commercial digital content and AI training data solutions, including Qlean Dataset. Visual Bank is also a selected participant in GENIAC, a Japanese government initiative supporting the advancement of next generation AI technologies.
CEO: Saneyuki Nagai
Website:https://visual-bank.co.jp/en
