1 · Text preprocessing
Tokenize, normalize, strip stopwords, stem, lemmatize, and tag — the universal preamble to any text model.
●Track 8 · live
NLP Foundations
Tokens to embeddings — turn text into vectors a model can learn.
Tokenization, stemming, lemmatization, stopwords, POS tagging, NER, OHE, Bag of Words, n-grams, TF-IDF, Word2Vec (CBOW + SkipGram), AvgWord2Vec. Two end-to-end projects: spam classification and Kindle review sentiment.
19 / 30 lessons5 chapters
2 · Feature extraction
From one-hot to BOW to n-grams — the count-vector representations that powered text classification for two decades.
3 · TF-IDF
Weight word counts by inverse document frequency — the classical text representation that still dominates search and retrieval.
4 · Word embeddings
Word2Vec CBOW + Skip-Gram, and the averaged-vector document representation — the conceptual leap from sparse counts to dense semantics.
5 · End-to-end NLP projects
Two full text classifiers: spam detection (BOW vs TF-IDF) and Kindle sentiment (AvgW2V vs TF-IDF + GridSearchCV + persistence).