Next Word Prediction

Inspiration

I always have wondered how our smartphone keyboard is able to predict text. Being an smartphone user for a long time, I was able to understand that it used some sort of pattern or sequence to determine the next word. Being a software engineer made me curious to figure out how it predicts. So this is my venture into Machine Learning.

What it does

This project contains a Jupyter Notebook which contains code to read data from a text file, train and generate a model using LSTM and asks input from the user upon which the suggestion/next word is provided.

How we built it

Initially we started with small data set to figure out how the prediction/suggestion works while entering a minimum of 3 words. Then regenerated the model to train on sequences of length 1 input to 1(target label).

Challenges we ran into

Data. With LSTM to work properly it needed more data.

Accomplishments that we're proud of

We are proud that we have done something new to improve our knowledge. Also proud to have stepped into the world of Machine Learning.

What we learned

The selection of data is important as is the preprocessing of the data itself. Model can be tuned to our expectation.

What's next for Next Word Prediction

Character level prediction like Microsoft Swiftkey does.

Built With

• deeplearning
• jupyter-notebook
• keras
• lstm
• machine-learning
• nltk
• python