Introduction
The Bengali BFSI Chat Dataset is a comprehensive collection of over 12,000 text-based chat conversations between customers and call center agents. Focused on Banking, Financial Services, and Insurance (BFSI) interactions, this dataset captures real-world service dialogues, complete with domain-specific language, customer intents, and varied conversational flows.
Participant & Chat Overview
•
Participants:
200 native Bengali speakers from the FutureBeeAI Crowd Community
•
Conversation Length:
300–700 words per chat
•
Turns per Chat:
50–150 dialogue turns across both participants
•
Chat Types:
Inbound and outbound
•
Sentiment Coverage:
Includes positive, neutral, and negative interaction outcomes
Topic Diversity
This dataset reflects the wide range of customer interactions typically encountered in the BFSI sector:
•Inbound Chats (Customer-Initiated)
•Account opening and management
•Transaction-related queries
•Loan inquiries and applications
•Insurance questions and requests
•Outbound Chats (Agent-Initiated)
•Product and service promotions
•Cross-selling and upselling efforts
•Loan follow-ups and reminders
•Customer retention and loyalty program outreach
•Insurance policy renewals and verifications
This topic spread ensures applicability across customer service automation, intent classification, and domain-specific model training.
Language Nuance & Cultural Relevance
Conversations capture natural Bengali as spoken in BFSI contexts, incorporating:
•
Names & Branding:
Realistic Bengali personal and business names
•
Local Contextual Elements:
Emails, phone numbers, addresses, time/date references, and currency in Bengali format
•
Colloquial Speech & Slang:
Regional idioms, informal expressions, and domain-specific jargon
•
Numerical Expressions:
Use of Bengali numerals, amounts, dates, and measurements as per local conventions
This linguistic richness enables the training of models that can understand real-world customer queries in culturally relevant contexts.
Conversational Structure & Flow
The dataset reflects structured dialogue flow and interaction dynamics seen in BFSI customer service environments:
•Complex problem-solving discussions
•Follow-ups and routine status checks
•Typical Chat Components:
•Closing messages and follow-up
•Feedback and escalation (where applicable)
This detailed structure is ideal for training NLP systems capable of handling complete customer journeys.
Data Format & Structure
Available in JSON, CSV, and TXT formats, each record includes:
•Full chat history with labeled turns
•Topic tags or metadata fields
The dataset is structured for easy integration with modern NLP frameworks.
Applications
This dataset supports a broad range of commercial and research-focused use cases in Bengali-language NLP:
•Chatbot and Smart Assistant Development for BFSI use cases
•Sentiment Analysis and Intent Detection
•NER and Entity Linking for financial domain terms
•Text Generation and Prediction
•Customer Experience Automation
•Text Recognition & Document Parsing for BFSI content
Ethical Collection & Data Security
•
Informed Consent:
All participants contributed with full consent
•
Privacy Compliant:
No personally identifiable information (PII) is shared
•
Data Security:
Collected, processed, and stored within FutureBeeAI’s secure platform
•
Ethical Standards:
Fully aligned with global responsible AI and data governance principles
Customization & Expansion
This dataset is continuously updated and can be tailored to meet your project-specific requirements:
•
Annotation Services:
Add sentiment labels, intents, named entities, or any custom annotation
•
Domain Customization:
Collect chats on specific BFSI subtopics like mortgages, mobile banking, or insurance claims
•
Multilingual Expansion:
FutureBeeAI supports custom chat collection in Bengali and other languages
Licensing
This dataset is developed and owned by FutureBeeAI and is available for commercial use under flexible licensing terms suited for enterprises, startups, and academic researchers.
