Python Natural Language Processing

Techniques used to generate final vectors and probability prediction stage

In this section, we will see how we will generate the final vector. We will also use some heuristics to generate output efficiently. So, we will talk about those heuristics as well.

As we have already seen, to generate the word vector we need to update input as well as the output vector. Suppose we have a million words in our vocabulary, so the process of updating the input and output vectors will take a lot of time and it will be inefficient. We have to solve this challenge. So we use some of the optimized ways to perform the same operations and those techniques are given as follows:

• Hierarchical softmax
• Negative sampling

So, let's start to understand these techniques. ...