Shiny Application
Next Word Predict is a Shiny application that
predicts the next word in a sentence based on text entered by a
user.
The application uses an n-gram language model to
generate predictions. An n-gram is a contiguous sequence of
n words from a body of text.
The predictive model was trained using a large corpus of:
- Blog posts
- News articles
- Twitter messages
N-grams were extracted from this corpus and used to construct the
language model.
Various natural language processing and text mining techniques were
explored to improve both prediction accuracy and
application performance.
The Predictive Text Model
The predictive text model was built using a sample of approximately
800,000 lines from the original text corpus.
Data Preparation Steps
- Converting text to lowercase
- Removing non-ASCII characters
- Removing URLs and email addresses
- Removing Twitter handles and hashtags
- Removing ordinal numbers
- Removing profane words
- Removing punctuation and excess whitespace
After cleaning, the text was tokenized into
n-grams.
Prediction Algorithm
As the user enters text, the algorithm:
- Attempts to match the longest possible n-gram (4-gram)
- Backs off to shorter n-grams if needed
- Selects the most frequent matching n-gram
- Predicts the next word
This uses a simple back-off strategy, prioritizing
longer and more informative word sequences.
Application User Interface
The predicted next word appears once the application detects that the
user has completed typing one or more words.
Interface Features
- Short delay while predictions are generated
- Slider allows up to three predictions
- Predictions shown in order of likelihood
Click the image below for a larger view:
