Project Language Translator

Training our own Translation Model.

Model file structure: >> eng-french.txt: This is dataset for our project, contains English text and corresponding french texts. (We only take 715 sentences) >> langTraining.ipynb: This file is used to create and train the model. >> training_data.pkl: This file contains lists of characters, text of input, and target data in a binary format. >> Trained_Model: This directory contains optimizer, metrics, and weights of our trained model. >> GUI.ipynb: Gui file of our project where we load the trained model and predict output for given text. Model is trained from scratch no pre-trained dataset is taken.

Training Of Model

This model is trained for only English to French language translation. It translate only 715 sentences that are used to trained the model. The accuracy of model is approximately 70%.

Some Terms

Some Terms used in code during training that are: >> Batch Size: Batch size is the number of training steps used in one iteration of the model in training algorithm. >> Epoch: An epoch is one complete pass through the entire training dataset. In other words, it represents the number of times the model has seen and processed the entire training. >> Validation Split: The validation split is a technique used to assess the performance of the model during training and tune its hyperparameters.

Over all Time Taken During Training: It depends on the system requirements. In my case it take approximately 5 hours to complete all the training steps. Some times it produce inappropriate response.