Extraction of Complex Sentence Relationships Based on Formalized Chinese Clause Structures
Wen Jing Hao · 2024
In the field of relation extraction (RE) from text, English clause structures have recently emerged as a promising method with significant practical value. However, applying this approach to Chinese clauses presents challenges due to the absence of formal structural representation. Our research shows that Chinese clauses follow a conditional probability function of the core relation P, which can be expressed as C = (S|P, O|P, SO|P), demonstrating a formal structural characteristic of discrete complementary synthesis. Notably, the O|P structure, with a high conditional probability, accounts for 79% of cases, serving as the core of complex sentence relations in Chinese. This sharply contrasts with the structural features of English clauses. Building on this, we propose the Discrete Complementary Synthesis Formal Structure (DCSFS) clause unit model. This model is employed to analyze and verify the role of the clause unit as the fundamental encoding element in Chinese sentences. Accordingly, we develop a method for automatically extracting semantic relations from complex Chinese sentences using this formalized clause unit model. Experimental results show that our approach outperforms existing punctuation-based clause determination methods and effectively addresses the challenges of extracting complex sentence relations from Chinese texts.