The purpose of this thesis is to research the Hu-Tucker algorithm for 
building Optimal Alphabetic Binary Search Trees (OABST). We designed two 
implementations: one running in $O(n^2)$ and another running in $O(n\log n)$ 
time; evaluated the performance of the algorithm for practical applications; 
and the break-even point of the two implementations. We compared the 
performance of the OABST to the compression ratio of two popular 
implementations of Lempel-Ziv algorithm (gzip and compress). We 
experimentally evaluated the optimality of the OABST against the optimality of 
an Optimal Binary Tree, built using the Huffman algorithm. The experiments 
showed that the compression ratio of OABST is very close (0.09 bits per word) 
to that of OBT, however the OABST can be used during both the encoding and 
decoding processes while the OBT require additional mapping mechanism for 
the decoding phase.