Natural Language Processing for Email Routing
summary
A production system used at MAIF that preprocesses and routes ~10,000 customer emails per day using natural language processing. It combines statistical methods and deep learning models — including convolutional neural networks — to classify emails, extract keywords, and improve assignment to the right handler; the code is available at https://github.com/MAIF/melusine.
AI-written summary · about
chapters (15)
- 0:00Introduction
- 2:53Email routing challenges
- 5:00Email preprocessing
- 7:30NLP pipeline
- 10:15Classification model
- 13:48Optimizing CNN
- 14:29Email auto-summary
- 16:58Production deployment
- 19:01Development best practices
- 21:52Results
- 22:13Open source launch
- 23:37Mélusine features
- 25:44Package configuration
- 27:07Open source feedback
- 30:38Mélusine demo
related talks
33mImpact of NLP in AI-powered Bank
15mMachine Learning for All
16mProduction MLOps for Spam Detection: Real-Time Multi-Modal AI
32mHarnessing Large Language Models for Enhanced Customer Experience
45mParsing Incoming Applications with NLP
32mAutomate customer ticket handling: from an idea to a production ready model
23mFighting phishing and spam with online machine learning on data streams
06mAdvancements in Natural Language Processing (NLP) and Text Analytics