Natural Language Processing Basics
For developers and aspiring data scientists who are new to working with text. You will clean and vectorise text, build recurrent and attention-based classifiers and taggers in PyTorch, and ship a sentiment service that you can compare honestly against a fine-tuned transformer.
Course Content
How to turn raw text into clean tokens and then into numbers a model can use, from bag-of-words and TF-IDF to word embeddings. 3 lessons, about 45 minutes.
Models that read text in order: plain RNNs and why their memory is short, LSTMs and GRUs that fix it, and bidirectional, stacked and attention-pooled variants. 3 lessons, about 35 minutes.
Putting sequence models to work on sentiment analysis, text classification and named entity recognition, then meeting the transformer that replaced them. 3 lessons, about 40 minutes.
Build an IMDb sentiment service end to end — data checks, a BiLSTM with attention, a transformer baseline and a deployable artefact. 1 lesson, about 15 minutes.