Skip to content

Classic ChatBots

cc

What are we trying to accomplish?

The goal of this module is to equip students with a deep understanding of how conversational AI systems are designed, built, and progressively improved—without relying on large language models. Beginning with the history and taxonomy of chatbots, students move through a deliberate progression: from rule-based systems driven by pattern matching, to data-driven systems that treat language as a structured dataset, to retrieval-based and classification-powered systems that use machine learning to understand user intent. Each lesson builds directly on the last, exposing both the capabilities and limitations of classical approaches so that students understand why each technique exists and when it breaks down. By the end of this module, students will have built multiple functional chatbot systems in Python and will be prepared to transition into generative, LLM-powered architectures.

Lessons

  1. Intro to ChatBots
  2. Regex & Rule-Based ChatBots
  3. Language as Data
  4. Deep Learning
  5. Retrieval-Based ChatBots

Module Topics

  • Chatbot taxonomy: rule-based, retrieval-based, and generative systems
  • NLU, dialogue management, and NLG components
  • Python ML environment setup with Jupyter Notebook
  • Regular expressions for pattern detection and intent inference
  • Rule-based conversation loop design and input normalization
  • Text preprocessing pipelines: tokenization, normalization, stopword removal, lemmatization
  • Vocabulary construction and corpus-based feature extraction
  • Bag of Words, N-grams, and TF-IDF vectorization
  • Information retrieval concepts and cosine similarity
  • Retrieval-based chatbot construction and response ranking
  • Intent classification with scikit-learn and PyTorch
  • Sentence embeddings and semantic retrieval
  • Classical vs. generative chatbot system design