Natural Language Processing with Sequence Models

Description

In this course, you will learn :

• Create word embeddings, then use them to train a neural network to perform sentiment analysis on tweets.
• Using a Gated Recurrent Unit (GRU) language model, generate a synthetic Shakespeare text.
• Using named entity recognition (NER) and LSTMs with linear layers, train a recurrent neural network to extract important information from text.
• Use a Siamese network to compare questions in a text and identify duplicates: questions that are worded differently but have the same meaning.

Syllabus :

1.  Neural Networks for Sentiment Analysis

• Trax: Neural Networks
• Why we recommend Trax1
• Trax: Layers
• Dense and ReLU Layers
• Serial Layer
• Other Layers 
• Training

 

2. Recurrent Neural Networks for Language Modeling

• Traditional Language models
• Recurrent Neural Networks
• Applications of RNNs
• Math in Simple RNNs
• Cost Function for RNNs
• Implementation Note 
• Gated Recurrent Units
• Deep and Bi-directional RNNs 

 

3. LSTMs and Named Entity Recognition

• RNNs and Vanishing Gradients
• Introduction to LSTMs
• LSTM Architecture
• Introduction to Named Entity Recognition
• Training NERs: Data Processing
• Computing Accuracy

 

4. Siamese Networks

• Siamese Networks
• Architecture
• Cost Function
• Triplets
• Computing The Cost 
• One Shot Learning
• Training / Testing

Show more