Predicting the Next Word with Text Mining

This is the final project to complete the Data Science Specialization offered by the Johns Hopkins University - Coursera.

The project consists of a Text Mining analysis that manipulates a text corpora formed with corpus from three social media sources:

• News
• Blogs
• Twitter

• Retrieving and Sampling the Data
  • The raw data is huge in size once unziped.
  • The dataset are sampled to produce a manageable corpus of text.
  • The text corpora is cleaned using the tm and stringr packages among others.
• Geneation of N-grams
  • N-grams are generated with RWeka package.
  • A dataframe with these N-grams is generated.
• Generation of the Model
  • With the N-grams dataframe a model using the e1071 package is created.
  • The search algorithm is Naive Bayes.

As shinyapps.io has restrictions with the sizes of user's datasets, even for commercial licences, I deployed the app in the shiny server running in my computer, this can manipulate the 3GB dataset of the Naive Bayes model needed to use the app.

Please visit this link which hosts the app.

Be patient as it could take a few seconds to load the dataset.