Slide 1: Problem Statement

• Predict the next word based on user input
• Useful for autocomplete systems
• Built using R and NLP techniques

Slide 2: Data Used

• Blogs dataset
• News dataset
• Twitter dataset
• Combined and sampled for training

Slide 3: Approach

• Text preprocessing (lowercase, clean text)
• Tokenization into unigrams, bigrams, trigrams
• Frequency-based n-gram model
• Backoff strategy used for prediction

Slide 4: Model Logic

• Trigram match → highest priority
• Else Bigram match
• Else Unigram fallback
• Final output = most frequent next word

Slide 5: Deployment

• Built interactive Shiny app
• Hosted on shinyapps.io
• Real-time prediction interface