Optimizing User Interactions with Intelligent Chatbots
summary
Shelly is an NLP-powered chatbot for Shell lubricants deployed in 11 countries and handling 750,000 interactions that returns product facts and personalized recommendations; the stack uses React and Go, Dockerized services with CI/CD, and a mix of static and on‑the‑fly translation to support four languages. Training uses tokens, intents and entities; testing includes Cypress unit scenarios, end‑to‑end and performance benchmarks. Limitations noted: maintenance overhead, security risks, edge‑case language failures and lack of human empathy.
AI-written summary · about
chapters (11)
- 0:00Introduction to Shell's tech ecosystem
- 3:12What are chatbots and NLP
- 4:59Advantages of chatbots
- 6:46Choosing NLP solutions
- 8:09Training NLP models
- 11:23Handling multilingual support
- 14:41Testing chatbot systems
- 16:24Performance and security testing
- 17:47Drawbacks of chatbots
- 20:17Error handling in chatbots
- 22:46Live demo of Shell's lubricant chatbot
related talks
13mExploring AI-Powered Chatbots and Their Transformative Role
33mYour e-commerce bot
46mChatbots, shall we shift into second gear?
47mTesting Conversational AI
48mUnlocking LLM Potential: From Exploration to Integration
55mEnterprise Chatbots (using messaging and AI to build novel user interfaces for work)
35mDesigning Experiences for Chatbots
35mChat Bot Building Story - The Good, the Bad and the Unexpected