Dataset Banking77 — Banking Customer Service Intentions
Annotated textual corpus containing requests from banking customers classified according to 77 types of specific intentions.
Description
Banking77 is a structured dataset for the classification of intentions in the banking sector. It contains 13,083 requests made by users, each annotated with one of 77 predefined intentions. This data set is designed to test or train models that can understand very specific customer requests in a real context.
What is this dataset for?
- Training NLP models for intent detection in banking
- Improve chatbots or voice assistants for customer service
- Conduct a detailed analysis of the needs and frustrations of users
Can it be enriched or improved?
Yes, this dataset can be completed with other languages for multilingual purposes, or adapted to other sectors (insurance, telecom, etc.). It is also possible to refine intent labels, merge similar classes, or relabel data with a finer hierarchical system.
🔎 In summary
🧠 Recommended for
- Chatbot developers
- NLP data scientists
- Fintech product teams
🔧 Compatible tools
- Scikit-learn
- SpacY
- Hugging Face Transformers
- FastText
💡 Tip
For better accuracy, try to group similar intentions into hierarchical clusters before training.
Frequently Asked Questions
Is this dataset suitable for voice banking assistants?
Yes, it is particularly suitable for bank conversational interfaces thanks to its granularity in intentions.
Can this dataset be used for sectors other than banking?
Yes, its structure can be used as a basis for creating similar datasets in other sectors such as insurance or e-commerce.
Is there a multilingual version of this dataset?
No, it is only available in English, but it can be translated or enhanced for multilingual projects.
