NLP Roadmap

Natural Language Processing — teaching machines to understand and generate human language.

Topics

Foundations

• Text Preprocessing — tokenization, cleaning, normalization
• Bag of Words and TF-IDF — classical text representations
• Embeddings — dense vector representations of words

Modern NLP

• Language Models — from n-grams to GPT
• BERT and Masked Language Models — understanding text
• Text Classification — sentiment, spam, topic detection
• Named Entity Recognition — extracting entities from text
• Text Generation — autoregressive models, sampling strategies

Advanced

• Retrieval Augmented Generation — grounding LLMs with external knowledge
• Prompt Engineering — getting the most from large language models
• Fine-Tuning LLMs — adapting pretrained models to specific tasks

The evolution

Rules → Bag of Words → Word2Vec → RNNs → Transformers → LLMs
                                                  ↑
                                            we are here

In practice, modern NLP = pretrained transformers + fine-tuning or prompting.

Links

• Transformers — the architecture
• Attention Mechanism — the core innovation
• Deep Learning Roadmap