Skip to content

README.md

Latest commit

 

History

History
38 lines (25 loc) · 1.92 KB

README.md

File metadata and controls

38 lines (25 loc) · 1.92 KB

Language_Classification

A NLP project to classify text on the basis of it's language with an accuracy of 100% (on test set and new examples) and confidence higher than 99%.

Aim

The aim of this notebook is to create a (Tensorflow) model that can classify the language (English, Spanish, etc.) of a given sentence.

In this notebook, English, Italian, Spanish and French languages are classified.

Model

I have used a sequential model with following layers in order:

  1. Embedding layer with vocab size of 49314 units and embedding dimension of 128 units.
  2. Flatten layer.
  3. Dense Layer with 256 neurons (activation units).
  4. Dropout layer with a dropout factor of 0.2.
  5. Dense Layer with 256 neurons (activation units).
  6. Dropout layer with a dropout factor of 0.2.
  7. Dense Layer with 4 neurons (activation units). This is the output layer.

The optimizer used is Adam with default settings. Batch size was taken as 128.

Prediction

The predictions of the model are as follows:

The sentence ' ['TensorFlow is a free and open-source software library for dataflow and differentiable programming across a range of tasks.'] ' is in ' English ' language with 99.84663724899292 % confidence.

The sentence ' ['TensorFlow è una libreria software gratuita e open source per il flusso di dati e la programmazione differenziabili in una vasta gamma di attività.'] ' is in ' Italian ' language with 99.95958209037781 % confidence.

The sentence ' ['TensorFlow es una biblioteca de software gratuita y de código abierto para el flujo de datos y la programación diferenciable en una variedad de tareas.'] ' is in ' Spanish ' language with 99.99784231185913 % confidence.

The sentence ' ['TensorFlow est une bibliothèque logicielle gratuite et open-source pour le flux de données et la programmation différenciable sur une gamme de tâches.'] ' is in ' French ' language with 99.9789297580719 % confidence.