Puzzling Together a TeacherBot: ML, NLP, Active Learning & MicroServ
summary
A TeacherBot named Emil maps KNIME forum questions to tutorial links using KNIME workflows and microservices, trained on ~5,000 forum posts and ~400 resources collapsed to 20 classes. Text processing uses chi-square keyword extraction and autoencoding features; classification uses a random forest with an active learning loop that labels the top 10% most uncertain by probability gap and expands labels via k‑NN (k=1). Microservices run through a core remote workflow node exchanging JSON; accuracy barely moves but link relevance improves.
AI-written summary · about
chapters (9)
- 0:00Start
- 11:07Goals
- 14:13Similarity Search Approach
- 15:23Ontology
- 18:09Active Learning
- 19:17The Active Learning Cycle
- 19:35Active Learning Cycle
- 24:14Create Micro Services
- 26:41What Did We Learn during this Project
related talks
31mMachine Learning-Enabled Chatbots
32mChatBot Learning Assistant with Large Language Models
59mMaking the most of artificial and human intelligence for data science
48mUnlocking LLM Potential: From Exploration to Integration
56mMy AI Learning Journey
48mBuilding an AI/ML powered text search system
19mBigBlueBot: teaching strategies for successful human-agent interactions
35mExtensionBot: A Domain Specific Gen AI Tool for the Cooperative Extension Service