Week 7 - Natural Language Processing, Part I

This week you will...

• get an understanding for the importance of tokenization of a text when training a neural network for texts, for example, to do a sentiment analysis. Tokenization is the process of converting the text into numeric values, with a number representing a word or a character.

• learn about embeddings, where the text tokens are mapped as vectors in a high dimensional space. With embeddings and labelled examples, these vectors can then be tuned so that words with similar meaning will have a similar direction in the vector space. This will begin the process of training a neural network to understand sentiment in text -- and you'll begin by looking at movie reviews, training a neural network on texts that are labelled 'positive' or 'negative' and determining which words in a sentence drive those meanings.

Learning Resources

467KB

241205_NLP in TensorFlow-Part-I.pdf

PDF

Open

• Week 1 and 2 of the course Natural Language Processing in TensorFlow

Until next week you should...

• watch ML with RNNs (NLP Zero to Hero - Part 4) [6 Min]

• watch LSTM for NLP (NLP Zero to Hero - Part 5) [5 Min]

• watch Training an AI to create poetry (NLP Zero to Hero - Part 6) [8 Min]

• complete this notebook to generate text in the unique speaking style of Star Wars character Master Yoda.

• watch the videos "Why human-level performance?", "Avoidable bias", and "Understanding human-level performance" to help you evaluating and improving your model

• consider a baseline model or a baseline comparison for your project task according to the instructions given here

• document the evaluation results of your baseline model and the used metric(s) in your project repository