Project: Smart Keyboard

• Targeted Group: Mobile devices user
• Objective : A keyboard which predicts the next word while typing
• Methodology : Training Set Data Cleaning, Analyzing, building and sampling for a predictive text model

Data cleaning Steps on the Training Dataset :

• Remove non ASCII characters, numbers and punctuation from
• Convert lowercase to uppercase
• Strip white spaces.
  

• Alogrithm use for Predictive Text is N-gram. It is used to compute P(w/h), the probability of a word w given some history h.
• In the application it Checks first for the highest-order (Trigram) n-gram or else the next lower order (Bigram);

• Input: User to enter text/word/sentence in the left handside text box
• Output: User can see the predicted text in the right side box
• Application also shows the N-Gram model currently used to predict the next word

• Current algorithm is limited to the contextual information till 3-grams
• To training dataset A larger dataset/ across regions
• To incorporate clustering to the underlying training corpus/data and predicting the cluster of the sentence typed.