Bot or Not? The Illusion of Intelligence
summary
A practical report of what actually works: Python libraries SpaCy, NLTK, sklearn, Keras get wired into a retrieval QA pipeline that ingests a file by SHA256, loads a malware knowledge base, and answers about nine or ten intents. A linear SVM outperforms an overcomplicated Keras model on this dataset. Logging and HipChat integration let the dataset grow from real queries. Remaining gaps are brittle yes/no handling and domain entity extraction for things like kernel32.dll.
AI-written summary · about
chapters (8)
- 0:00Introduction
- 4:38What defines a chatbot
- 8:50Types of chatbots
- 14:33Bot development kits
- 21:43Building a chatbot in Python
- 29:56Introducing Malbot
- 36:35Malbot architecture and scaling
- 39:26Q&A
related talks
26mNLP for security log analysis : Learning to crawl before you run
55mDemystifying Natural Language Processing across several industry verticals
43mGetting started with AI for free
39mKeynote 3: How did you know? Explaining Black Box Model Predictions in Python
49mKeynote 4: Here Come The Robots - Python and Machine Learning
26mA Practical Introduction to Productionizing NLP Models
44mApplied Machine Learning for Data Exfil and Other Fun Topics
34mTechnical: Harvesting Trust: Methods for Validating Chatbots in