Utilization Of Receiver-Specific Dictionaries For Improved Auto-Completion Functionality

Open Access
Gujjula, Sai Pravallik
Area of Honors:
Computer Science
Bachelor of Science
Document Type:
Thesis Supervisors:
  • Jeremy Joseph Blum, Thesis Supervisor
  • David Scott Witwer, Honors Advisor
  • Linda Marie Null, Faculty Reader
  • Receiver Specific Dictionaries
  • Auto-complete
  • Mobile User Personalization
  • Trie-Based
  • Text Message Data
  • Mobile Keyboard
The auto-complete on mobile phones is one of the most commonly used keyboard features. It assists a user by predicting the intended choice of words based on partial input. These predictions increase productivity by reducing the number of keystrokes required to obtain an intended word. The current day auto-complete implementations focus on making the prediction algorithms efficient and precise by studying the user’s choices of input. This proves to be highly beneficial as the prediction algorithms become user specific, and predict words based on the manner in which the user communicates. However, there is a need to consider another perspective in order to make these algorithms more precise and efficient, i.e. the receiver whom the user communicates with, and the need to make the word predictions receiver specific. The importance of receiver specificity lies in a user’s tendency to communicate using different set of words with a variety of receivers. Considering the receiver specificity in the choice of words when predicting, the auto-complete feature requires an additional resource to store words specific to a set of receivers. Together with a common dictionary that contains words that are common to all users, there is a need for receiver specific dictionaries in which each consists of words that are specific just to the user and a receiver. This should ensure that the predicting algorithm will select a more receiver-associated word from the related receiver dictionary, thus, increasing precision. This thesis analyzes the implementation of an auto-complete algorithm with the receiver-specific dictionaries. It compares this implementation with the existing form of implementation that uses only one common dictionary. The results are based off a comparison on the number of keystrokes saved to obtain an intended choice of word using either of the implementations. The receiver specific implementation saw about a 25% reduction in the number of keystrokes used as compared to the baseline implementation with a single dictionary of all the previously used terms.