German Street Dictionary Compression

While at ScanSoft, I asked the manager of the embedded speech team what I can help out with, and he gave me something related to compression. To tackle the task at hand, I read about Huffman coding and other algorithms such as arithmetic coding and sliding window method. After doing some research, I decided to write a length limited Huffman coding program customized to ScanSoft’s pronunciation dictionary file. The method was able to perform better than the 6-bits method that was used at ScanSoft. The 6-bits method performs compression by using only 6 bits to represent each character.

compression code - 227 lines of code

decompression code - 309 lines of code