Performance Evaluation of Efficient Hybrid Compression Methods For Devanagiri-Encoded Hindi Text Using Lossless Algorithms

Abstract: This research paper provides a comprehensive performance analysis of five standard lossless compression algorithms-LZMA, Zstd, Brotli, Bzip2, and LZ4HC-alongside sixty hybrid combinations on UTF-8 encoded Hindi text datasets, written in the Devanagari script, across varying sizes-small, medium, and large. The evaluation covers compression ratio, compression speed, decompression speed, and a derived weighted normalized efficiency score. Hybrid algorithms-especially those combining Zstd with LZ4HC or Brotli-demonstrate superior performance across all individual parameters. Independent Zstd achieves the highest efficiency score for the medium-sized dataset, illustrating how a well-balanced algorithm can excel in overall efficiency despite not leading in any single metric. However, in terms of compression ratio, speed, and decompression speed, hybrid methods consistently outperform standalone approaches. Notably, the hybrid combination Zstd + LZ4HC achieves the highest efficiency, with scores of 0.6764 for small files and 0.8597 for large files. Independent algorithms such as Zstd and Bzip2 perform well but are consistently surpassed by strategic hybrid pairings. The analysis underscores the significance of combining fast and high-ratio compression techniques, particularly for language-specific data. However, not all hybrids yielded significant improvements, with some introducing computational overhead. This comparative analysis highlights the potential of hybrid compression in applications such as cloud storage, mobile messaging, and NLP-based Hindi text processing.