Analysis of the Efficiency of Huffman and LZW Compression Algorithms on Vietnamese Text Data

This paper analyzes and compares the effectiveness of two representative lossless compression algorithms, Huffman coding and LZW, on Vietnamese text data encoded in UTF-8. A web-based text compression application was developed in JavaScript, implementing byte-level canonical Huffman coding and LZW with variable code width from 9 to 16 bits. The two algorithms were evaluated on a corpus of approximately 360 KB covering four text categories (encyclopedic, literary, legal–administrative, and scientific–technological) collected from Vietnamese Wikipedia and Wikisource, through three experiments: by text category, by data size, and by the influence of Vietnamese diacritics. The results show that LZW achieves compression ratios of 29.3–48.2% of the original size and becomes increasingly effective as data grows, clearly outperforming Huffman coding (64.3–68.6%), which is lower-bounded by the symbol entropy of the data (about 5.2 bits/byte). In contrast, Huffman compresses roughly four times faster (about 80 MB/s versus 20 MB/s). The experiments also indicate that the multi-byte UTF-8 encoding of Vietnamese script significantly reduces the effectiveness of byte-level Huffman coding while having almost no effect on LZW. Overall, these findings indicate that dictionary-based coding is better suited to Vietnamese UTF-8 text, whereas byte-level statistical coding is best reserved for speed-critical settings or as an entropy-coding stage after another transform; the compression tool and the four-category corpus built here are offered as reusable resources.