Average shortest-path length in word-adjacency networks: Chinese versus English

Area of Science:

• Linguistics
• Computational Linguistics
• Network Science

Background:

• Complex networks are valuable for analyzing natural language structures.
• Previous research suggests punctuation marks exhibit word-like properties in Zipfian analysis and stylometry.

Purpose of the Study:

• To analyze the topology of growing word-adjacency networks in Chinese and English literary works.
• To investigate the impact of including punctuation marks alongside words in network analysis.
• To compare the functional dependence of average shortest path length on network size across languages and epochs.

Main Methods:

• Construction of word-adjacency networks from Chinese and English literary texts.
• Inclusion of punctuation marks as nodes in the networks.
• Analysis of the average shortest path length L(N) as a function of network size N.
• Approximation of empirical results using a growing network model.

Main Results:

• Satisfactory agreement was found between empirical results and the growing network model.
• Including punctuation marks leads to asymptotically similar L(N) behavior for both Chinese and English networks.
• Neglecting punctuation results in significantly larger L(N) for Chinese compared to English.

Conclusions:

• Punctuation marks carry significant linguistic and structural information in natural language networks.
• The inclusion of punctuation provides a more unified understanding of complex network topology across different languages.
• Word-adjacency network analysis, especially with punctuation, offers insights into linguistic evolution and stylistic features.