A Chinese Named Entity Recognition Dataset for Intangible Cultural Heritage

Area of Science:

• Natural Language Processing
• Digital Humanities
• Cultural Heritage Informatics

Background:

• Named Entity Recognition (NER) is crucial for organizing knowledge in Intangible Cultural Heritage (ICH).
• Limited annotated datasets hinder advancements in ICH-specific NER.
• Existing methods struggle with the complex linguistic patterns in cultural heritage texts.

Purpose of the Study:

• To develop a specialized Chinese dataset for NER in the ICH domain.
• To propose an effective NER model tailored for ICH knowledge extraction.
• To facilitate structured knowledge representation and analysis of cultural heritage.

Main Methods:

• Creation of a novel Chinese dataset for ICH, including heritage items, inheritors, and materials.
• Development of a hybrid NER model integrating RoBERTa for feature extraction.
• Utilization of Kolmogorov-Arnold Network (KAN) for intricate pattern recognition.
• Application of Conditional Random Field (CRF) for robust sequence labeling.

Main Results:

• The proposed NER model demonstrates high effectiveness in identifying and classifying ICH entities.
• The model successfully captures complex semantic dependencies within ICH texts.
• Experimental results validate the utility of the new dataset and the proposed model.

Conclusions:

• The developed dataset and NER model significantly advance information extraction for Intangible Cultural Heritage.
• This work provides a valuable resource for digital preservation, knowledge graph construction, and cultural research.
• The findings pave the way for enhanced structured representation and analysis of cultural heritage knowledge.