Word Prediction Application

• Simple: Predicts the next word in a phrase or sentence
• Fast: Small dataset, quick to load (81 MBs)
• Easy: Requires user to just start typing

• Data: millions of lines of twitter, blogs, and news articles
• Data converted into thousands of n-grams (quadrigrams,trigrams,bigrams)
• Only n-grams that meet a certain criterion are kept
• Algorithm searches and matches top three most frequent occurences
• If less than three occurences in quadrigram dataset, will search trigram dataset then bigram dataset
• If less than three occurences in bigram data set, will fill in remaining slots with most frequent words: “the”, “to”, “and”

• User simply types in the whitespace box labeled “Type Here:”
• Top three guesses for the next word will appear on the right

• Could be implemented into a texting application to speed up texting.
• Useful for written documents that have become smeared or illegible.
• Application is small and not computationally intensive so could be used on almost any device.

NOT USEFUL for EXTREME ACCURACY

• While what is considered “good” accuracy is debatable, this application sacrifices accuracy for speed/memory.