This is a really cool algorithm that lets you generate relevant data for your ML needs. Note : Used for NLP based Machine Learning.
I call it genius because it addresses the woes of any ML coder ; Lack of data ! Takes a sentence as input. Uses certain libraries to translate it into many other languages. Translates those translated sentences back into the source language. This generates paraphrases of the input sentence. This method utilizes the semantic and syntactic variances in languages, and generates a much larger, clean dataset for you to use for your everyday ML needs. It really helps when you’re lacking data size consistency, as well as with data scaling. (Larger the dataset, better the results). The dataset is relevant as well as varied at the same time ! Slow, but that’s unhelpable because of the multiple API calls being made.
Dependencies :
You must have the following python libraries ;
- mtranslate
- googletrans