Alternative source coding model for mobile text communication

This paper presents a new source coding model for the efficient transmission of mobile text messages. Unlike conventional text messages, which are sent in the form of character sequences, the proposed method uses key-code sequences that reflect the user's typing history for authoring the message. The key-code representation can be as efficient as 4 bits per key-code by utilizing existing dictionary-based text entry systems and mobile phones equipped with the standard reduced keyboard. Experiments that use an optimal dictionary show that the key-code representation can save 2.95 bits per character compared to the conventional ASCII representation. Further, computation of the entropy shows that the optimal code-word length of the ASCII representation in kth-order approximation, where probability distributions of blocks of k symbols are given, cannot be comparable to that of the key-code in (k-l)-order approximation.