A Simple Word Prediction

The text prediction has become an increasing challenge in everyday tasks and those that are not.

Here is a simple application for the prediction of a given word phrase, using the knowledge acquired through Data Science Specialization offered by Johns Hopkins University.

We will talk about:

• The algorithm
• The Shiny application
• How can you use this app?

The base of the algorithm is based in the Markov Property:

The prediction of the future is based on exactly n-moments of the past

Based in this Markov Model, you can think in predict the next word of “I live in New York”. The word with the highest chance given this word, will be the best prediction given a specific number of words, for example,

P( best word | I live in New York ) = P( best word | New York )

So, based in this concepts,

• search for the combination of word that the sample text shows.
• Calculate the probabilities fo the sentences.
• The algorithm evaluate the best option for your phrase.

The application is based in Shiny (a web application framework for R), with a simple interface, basic analysis of the corpus and a detail explication of the model.

1. In the Predict! tab you can find the main reason for the app, a word predictor.
2. About the Corpus, you will see basic statistics of the documents used to train the model.
   
3. At the Model tab you can see a very specific explication of the model.
4. Finally, in the More tab, you can finf info about me and this presentation.