Scholarlink

<p>When synonyms coexist in a synset (synonymous set), they often compete with one another in language use. This study examines Chinese synonym competition over a century using the Google Books Ngram Corpus (GBNC) and Chinese Open Wordnet (COW) as the data sources. It divides the timeline into twenty-year intervals and uses XGBoost to predict the future winner based on three statistical features (relative growth, linear extrapolation and central moments) and four linguistic features (stroke, radical, word age and categorial variation) in order to explore which features contribute to winner prediction. Multiple experiments, including models using all features, a single feature, statistical-only features, and linguistic-only features, as well as ablation experiments, consistently show that statistical features have significantly stronger predictive power than linguistic features, with central moments being the most predictive feature, while linguistic features play an auxiliary role. In addition, the average performance of any single feature is worse than that of the all-feature model. The findings indicate that Chinese synonym competition shows a frequency-dominant mechanism supplemented by linguistic factors. This study provides quantitative evidence for understanding Chinese lexical change based on century-long language data and machine learning methods.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

The dynamic competition of Chinese synonyms over a century

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月6日

查看全文
  • Shan Wang,
  • Yichi Wang,
  • Haodi Zhang

摘要

When synonyms coexist in a synset (synonymous set), they often compete with one another in language use. This study examines Chinese synonym competition over a century using the Google Books Ngram Corpus (GBNC) and Chinese Open Wordnet (COW) as the data sources. It divides the timeline into twenty-year intervals and uses XGBoost to predict the future winner based on three statistical features (relative growth, linear extrapolation and central moments) and four linguistic features (stroke, radical, word age and categorial variation) in order to explore which features contribute to winner prediction. Multiple experiments, including models using all features, a single feature, statistical-only features, and linguistic-only features, as well as ablation experiments, consistently show that statistical features have significantly stronger predictive power than linguistic features, with central moments being the most predictive feature, while linguistic features play an auxiliary role. In addition, the average performance of any single feature is worse than that of the all-feature model. The findings indicate that Chinese synonym competition shows a frequency-dominant mechanism supplemented by linguistic factors. This study provides quantitative evidence for understanding Chinese lexical change based on century-long language data and machine learning methods.