• Typing on mobile devices can be slow and error-prone
• Predictive text helps users type faster and more accurately

• Accepts user input of a phrase
• Predicts the most probable next word
• Powered by cleaned SwiftKey dataset and n-gram models

• Text corpus cleaned: lowercase, punctuation removal, whitespace trim
• N-grams generated:
  • Unigrams, Bigrams, Trigrams
• Prediction Logic:
  • If trigram match: return word3
  • Else bigram match: return word2
  • Else: return most frequent unigram

```r predict_next_word(“I am”, …) # returns: “going”