Assessing Text Similarity From Compressed Representations Using Dictionary-Based Compression Algorithms
Year: 2026
Type: article
Abstract: This paper presents an investigation into the impact of text compression on the performance of text similarity detection. Two experimental scenarios were designed and tested on text datasets: 1) comparison without compression, 2) comparison using a pre-compressed text collection. Also, the effect of different compression ratios on similarity rates in the same text collection is investigated. A custom compression algorithm, specifically designed to preserve word integrity during the compression process, was employed. The results obtained with this algorithm were also compared with other algorithms of the same nature found in the literature. Text similarity was evaluated using the Cosine Similarity Algorithm in compressed form of texts. For each algorithm, five trials were conducted to calculate average values, and in these trials, compression time, and similarity calculation time were measured. The experimental results indicate that using compressed text collections offers a processing-time advantage compared to uncompressed text. Moreover, the similarity ratios obtained from compressed texts were found to be very close to those derived from uncompressed (plain) texts.
Source: IEEE Access
Authors: Resat Kazanoglu, Halil Nusret Bulus
Institution: Tekirdağ Namık Kemal University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: gold
Estimated APC paid: $2,075
OpenAlex ID: W7168158918