Chinese technical terminology extraction based on DC-value and information entropy.

Area of Science:

• Computer Science
• Information Science
• Computational Linguistics

Background:

• Rapid technological development in China leads to a surge in patent applications.
• Existing automatic term extraction methods often focus on English, limiting applicability to Chinese patent documents.
• Current Chinese terminology extraction methods may overlook domain correlation and termhood features, focusing primarily on frequency.

Purpose of the Study:

• To develop an effective automatic method for extracting Chinese technical terminology from patent documents.
• To address the limitations of existing methods in handling the nuances of the Chinese language and technical domain specificity.
• To enhance patent mining and analysis for technological innovation and R&D risk mitigation.

Main Methods:

• A novel Chinese technical terminology extraction method is proposed, utilizing DC-value and information entropy.
• The algorithm is designed to capture both domain correlation and the unithood feature of terminology.
• The method is applied to a corpus of Chinese patent literature.

Main Results:

• The proposed algorithm effectively extracts technical terminology from Chinese patent literature.
• Empirical results demonstrate superior performance compared to C-value, log-likelihood ratio, and mutual information methods.
• The method shows significant improvements in identifying relevant technical terms.

Conclusions:

• The developed method offers a valuable tool for in-depth mining and analysis of Chinese patent information.
• This approach supports efficient use of patent data, fostering technological innovation and mitigating R&D risks.
• The study holds both theoretical significance and practical application value for researchers and technical managers.