Final Submission

Swarnika Shakya

Slide 1 — Title

Next-Word Prediction Engine
 Turning text data into real-time intelligence

  • Built using large-scale text datasets (blogs, news, Twitter)

  • Predicts the next word based on user input

  • Designed for speed, simplicity, and usability

Slide 2 — Problem & Opportunity

Typing assistance and text prediction are core to:

  • Search engines

  • Messaging platforms

  • Content creation tools

Problem:
 Users lose time typing and correcting text manually

Opportunity:
 Provide fast, context-aware word predictions to improve efficiency and user experience

Slide 3 — Algorithm Approach

The model uses an N-gram language model:

  • Trigrams (3-word sequences) for highest accuracy

  • Bigrams as fallback

  • Unigrams as final fallback

How it works:

  1. User enters a phrase

  2. Model searches for matching word sequences

  3. Returns the most probable next word

Optimization:

  • Sampling used for performance

  • Frequency-based lookup for fast response

  • Backoff strategy ensures prediction even with unseen phrases

Slide 4 — The Application

Features:

  • Text input box for user phrase

  • Instant next-word prediction

  • Lightweight and responsive

How to use:

  1. Enter a phrase (e.g., “I am going to”)

  2. Click submit or wait briefly

  3. Receive predicted next word

Design Focus:

  • Simplicity

  • Speed

  • Accessibility via browser (shinyapps.io)

Slide 5 — Results & Experience

Performance:

  • Generates predictions in real time

  • Works across informal and formal text inputs

User Experience:

  • Fast and intuitive

  • Minimal learning curve

  • Useful for drafting and idea generation

Future Improvements:

  • Better accuracy using smoothing techniques

  • Multi-word predictions

  • Context-aware personalization