Word Predictor App Presentation

• Predict next word given the previous words
• Proof-of-concept Shiny web application
• Prediction model could be further developed and used in many different business cases

• https://arttu.shinyapps.io/word_predictor
• Write input sentence and the app will predict a list of words
• Slider controls the number of suggested words
• Word cloud on the right hand side visualises the frequencies
• Table on the left hand side shows predictions and contains a search box for fnding specific words

• Data is from a corpus called HC Corpora
• Consists of text files collected from publicly available sources by a web crawler
• English language files that were gathered from Twitter and different blogs and news sources
• Should give a rather good mix of general language used today
• Predicting based on previous two words and giving five suggestions, the app shows right word 76% of time

       5suggestions 3suggestions 1suggestions
2-gram         0.76         0.73         0.65
1-gram         0.72         0.68         0.58

• Based on the stupid backoff -algorithm from the Markov family of probalistic models:
  1. Take the input and use the same text transformations as for the training data and return last two words.
  2. Search for two first input words in the 3-grams training data and if matched, predict the third word. If no match, then next step.
  3. Search with only the last input word in the first word of 2-grams training data. If matched, predict the second word. If no match, then next step.
  4. Predict the most common words in the 1-gram data.
• Despite its name, it actually performs quite well given very large data.

• Could us different data sets for domain specific application or another language
• Try out more advanced models like Markov chains
• UI could be improved