Text Algorithm

A Shiny web application where users input a word or sentence, and algorithm predicts the next 3 most likely words.

• Trained on blogs, news, and Twitter data
• Data was cleaned to create a corpora (e.g., converted to lower case, remove punctuation, profanity, and white spaces)
• Built a n-gram language model
• Uses pre-computed frequency tables for speed

Goal: provide fast and accurate text prediction similar to smartphone keyboards.

The model uses a Backoff N-Gram Algorithm. Here is how it works:

1. N-grams are generated: Bigrams, Trigrams, Quadgrams

2. Prediction logic:

   • Match quadgram
   • If none → backoff to trigram
   • If none → backoff to bigram
   • If none → return most common starting words

3. Top 3 predicted words are outputted

Key strengths:

• Clean UI: Simple and User-Friendly
• Fast Predictions
• Practical NLP Application

Future Improvements:

• Use a larger training dataset to make more contextual predictions
• Be able to tackle more advanced language models

Check it out here: https://arunparmar.shinyapps.io/FinalProjectPredictionAlgorithm/