Building a Neural Machine Translation System

Description:

While more than 50% of the data on the internet is in English, only 20% of the world speaks English. According to https://csa-research.com/ ,72.1 percent of the consumers spend most or all of their time on sites in their own language, 72.4 percent say they would be more likely to buy a product with information in their own language and 56.2 percent say that the ability to obtain information in their own language is more important than price. Machine Translation is a tool that facilitates and enhances the spread of knowledge and build better conversations.

In recent years, with the exponential growth in Deep Learning, Machine Translation system has moved from Statistical Based Models to Neural Models. LSTMs were first introduced in 1997, but it is in the 5-6 years where its potential has been utilised. This is because of increase in computation power and abundance in data. LSTM's allow processing of sequential data. This has been the building block for solving many problems like Text Summarisation, Building Chatbots, generating captions for images, Machine Translation etc. In this problem, we will use a Encoder-Decoder Architecture .The encoder reads an input sequence and generates a single output vector. The Decoder then reads this output vector to produce the output sequence. This kind of architecture is commonly used when both the input and output are sequences. Both the Encoder and Decoder consists of LSTM cells.

Outline:

1. Introduction to Machine Translation
2. Introduction to RNNs and LSTMs
3. Encoder Decoder Architecture
4. Architecture of the Model
5. Building an English to Hindi Machine Translation Model

Prerequisites:

Basic Knowledge of Neural Networks

Content URLs:

https://github.com/AiswaryaSrinivas/GoogleIO_ByteAcademy

https://www.kaggle.com/aiswaryaramachandran/english-to-hindi-neural-machine-translation

https://github.com/AiswaryaSrinivas/GoogleIO_ByteAcademy/blob/master/Machine_Translation.pdf - Link to the Presentation as PDF

Speaker Info:

Currently Senior Data Analyst at Cerner, with strong foundations in Machine Learning and Artificial Intelligence.I specialise in NLP with experience in working on Financial News Data and on Social Media Analytics. I possess good writing skills, with my articles being published in "Analytics Vidhya" publications on Medium. I hope to use analytics to improve the two pillars of any society -Healthcare and Education

Speaker Links:

https://medium.com/@aiswaryar

https://github.com/AiswaryaSrinivas/