💬Core AI
NLP
Natural Language Processing with transformers
Master Natural Language Processing from text preprocessing to building transformer-based models. Covers tokenization, embeddings, sentiment analysis, NER, and HuggingFace pipelines.
56 lessons
12 hrs
Intermediate
What you will learn
Build real-world projects from scratch
Write clean, production-ready code
Understand core concepts deeply with diagrams
Follow industry best practices
Get hands-on with code in every lesson
Access lifetime updates as the tech evolves
Curriculum
Module 1Text Preprocessing Fundamentals4 lessons
Module 2Regular Expressions for Text Extraction4 lessons
Regex Fundamentals — Literal Characters and Character Classes13 min
Extracting Real Structured Data — Emails, Phone Numbers, and Dates15 min
Capturing Groups and Lookaheads — Precise, Structured Extraction14 min
Combining Everything Into a Real Extraction Pipeline14 min
Module 3Bag-of-Words and TF-IDF3 lessons
Bag-of-Words — Turning Text Into Word Counts13 min
TF-IDF — Weighting Words by How Informative They Actually Are15 min
TF-IDF vs Embeddings — An Honest, Measured Comparison14 min
Module 4N-grams and Statistical Language Models3 lessons
N-grams — Capturing Short Sequences Instead of Single Words12 min
Building a Statistical Language Model From Bigram Counts15 min
Where This Approach Structurally Breaks Down13 min
Module 5Part-of-Speech Tagging2 lessons
What Part-of-Speech Tagging Actually Assigns12 min
A Real, Context-Aware Tagger — Fixing the Measured Failure14 min
Module 6Named Entity Recognition3 lessons
Why Regex Cannot Find Names, Places, and Organizations12 min
Building a Real NER Pipeline with spaCy14 min
Evaluating NER Correctly — Why Entity-Level F1, Not Token Accuracy14 min
Module 7Coreference Resolution3 lessons
Why Finding Entities Isn't Enough — The Pronoun Resolution Gap12 min
Building a Real Coreference Resolution Pipeline15 min
Where Coreference Resolution Still Struggles — Honest Limitations13 min
Module 8Dependency Parsing2 lessons
Why POS Tags Alone Cannot Answer 'Who Did What'12 min
Extracting Subject-Verb-Object Relationships for Real Use14 min
Module 9Text Classification Pipelines End to End2 lessons
Assembling a Complete Classification Pipeline15 min
Is TF-IDF Still Competitive? A Direct Measurement14 min
Module 10Using Hugging Face Pipelines Practically3 lessons
The Pipeline Abstraction — One Line, a Full Transformer Underneath12 min
Zero-Shot Classification — Classifying Without Any Training Data13 min
Translation Pipelines and Practical Model Selection Tradeoffs13 min
Module 11Fine-Tuning a Pretrained Transformer, the Practical Workflow2 lessons
From DL Module 32's Hand-Written Loop to Hugging Face's Trainer14 min
Practical Fine-Tuning Decisions — Learning Rate, Freezing, and Data Size14 min
Module 12Topic Modeling3 lessons
Discovering Topics With No Labels at All13 min
Choosing the Right Number of Topics — LDA's Real Practical Challenge13 min
BERTopic — Fixing LDA's Limitations With Embeddings14 min
Module 13Text Summarization — Extractive vs Abstractive2 lessons
Extractive Summarization — Selecting the Most Important Existing Sentences13 min
Abstractive Summarization — Generating New Text With a Pretrained Transformer14 min
Module 14Semantic Search and Sentence Embeddings2 lessons
Sentence Embeddings — Compressing Meaning Into One Vector12 min
Building a Real Semantic Search Engine15 min
Module 15NLP-Specific Evaluation Metrics3 lessons
BLEU — Measuring Translation Quality Without One Correct Answer14 min
ROUGE — Measuring Summary Quality by Recall, Not Precision13 min
Perplexity — Measuring How Well a Language Model Predicts Real Text13 min
Module 16Handling Multilingual and Low-Resource Text3 lessons
Language Detection — Why It Must Come Before Everything Else12 min
Cross-Lingual Embeddings — Semantic Search Across Languages13 min
Translate-Then-Process vs a Multilingual Model — A Measured Comparison13 min
Module 17Bias and Fairness in NLP Systems3 lessons
Measuring Bias Directly Inside a Real Embedding Model13 min
Measuring Bias in a Real Classifier's Actual Predictions13 min
A Concrete Mitigation Technique — And Its Honest Limits13 min
Module 18Building a Production NLP Pipeline2 lessons
Assembling Every Piece Into One Real Pipeline15 min
Serving the Pipeline With FastAPI and Basic Monitoring14 min
Module 19Prompting vs Fine-Tuning — Choosing the Right Tool2 lessons
Zero-Shot and Few-Shot Prompting — Classification With No Training Step13 min
Measuring the Real Cost-Accuracy Tradeoff Directly14 min
Module 20Data Augmentation for Text2 lessons
Synonym Replacement and Back-Translation — Two Ways to Multiply Small Datasets14 min
Measuring Whether Augmentation Actually Helps13 min
Module 21Evaluating Generated Text for Factual Consistency3 lessons
Why ROUGE and BLEU Cannot Catch a Hallucinated Fact13 min
Building a Concrete Factual Consistency Check14 min
A Complete Factual Consistency Pipeline for Real Summaries14 min
56 total lessons across 21 modulesPreview Module 1 free