Use Deep Learning to Generate Fantasy Names: Build a Language Model from Scratch
To truly grasp the intricacies of Language Models (LM) and become familiar with their underlying principles, there is no other way than rolling up our sleeves and starting to write code. In this article, I present the creation of a Recurrent Neural Network (RNN) built entirely from scratch, without the aid of any deep learning library.
Tensorflow, Keras, Pytorch make building deep and complex neural networks effortless. Undoubtedly, this is a great advantage for Machine Learning practitioners, however, this approach has the massive downside of leaving the functioning of those networks unclear as they happen “under the hood”.
This is why today we will perform the inspiring exercise of building a Language Model using only the Numpy Python library!
The main breakthrough in AI application within the NLP field is undeniably represented by Recurrent Neural Networks (RNN).
RNNs are a class of artificial neural networks particularly well suited for NLP tasks and text generation. The reason for their efficacy lies in their ability of capturing sequential dependencies in data. Human language deeply relies on considering the context and linking the first words in a sentence to the last ones. Consider these sentences:
0 Comments