Introduction
Welcome to the Czech Scripted Monologue Speech Dataset tailored for the BFSI (Banking, Financial Services, and Insurance) domain. This dataset empowers the development of advanced Czech speech recognition systems, natural language understanding models, and conversational AI solutions focused on the BFSI sector.
Speech Data
This dataset includes over 6,000 scripted prompt recordings in Czech, covering a wide range of realistic banking and finance-related scenarios to support robust ASR and voice AI systems.
•Participant Diversity
•
Speakers:
60 native Czech speakers.
•
Regions:
Diverse representation from various Czech Republic provinces to ensure dialect and accent coverage.
•
Demographics:
Age range of 18–70, with a male-to-female ratio of 60:40.
•Recording Details
•
Nature:
Scripted monologues and domain-specific prompt recordings.Duration:
•
Audio Format:
WAV, mono channel, 16-bit depth, recorded at 8 kHz and 16 kHz sample rates.
•Environment: Clean, echo-free, and noise-free environments.
Topic & Context Diversity
This dataset spans multiple BFSI-related themes to simulate practical customer interaction scenarios:
•Customer service interactions
•Financial transactions & balance inquiries
•Banking and insurance product queries
•Regulatory and compliance questions
•Technical help and password resets
•Promotional campaigns and service updates
Contextual Elements
To make the dataset as context-rich as possible, each prompt integrates commonly encountered real-world BFSI elements:
•
Names:
Region-specific names in multiple formats
•
Addresses:
Local address structures and pronunciations
•
Dates & Times:
Typical time expressions used in banking
•
Organization Names:
Names of banks, financial firms, and institutions
•
Currencies & Amounts:
Spoken currency formats, prices, and numeric data
•
IDs & Transaction Numbers:
For authentic service simulation
Transcription
Every audio file is paired with verbatim transcription to streamline ASR and NLP model development.
•
Content:
Exact match of each prompt
•
Format:
Clean .TXT files, mapped to audio file names
•
Accuracy:
Reviewed and validated by native Czech linguists
Metadata
Each data point is enriched with detailed metadata for advanced training and analysis:
•
Participant Metadata:
Unique ID, age, gender, state, country, dialect
•
Recording Metadata:
Transcript, recording setup, sample rate, bit depth, device, file format
Applications and Use Cases
This BFSI-focused dataset is ideal for:
•
Speech Recognition Training:
Build or fine-tune ASR models in Czech
•
Voice Synthesis Models:
Create realistic synthetic banking voices
•
Voice Assistants & IVR:
Power smart assistants and bots for finance workflows
•
Chatbot Training:
Build virtual agents for financial services
•
NER & Entity Extraction:
Train NLP models with real-world financial terms
•
Language Understanding:
Improve intent detection, sentiment analysis, and topic modeling
Secure & Ethical Data Collection
All data was collected via FutureBeeAI’s proprietary platform Yugo
•Entire workflow conducted within a secure, controlled environment
•Participants gave full consent under strict ethical protocols
•No PII (Personally Identifiable Information) is included
•Fully compliant and safe for commercial use
License
This dataset is created and owned by FutureBeeAI and is available for commercial licensing.