A tree-based corpus annotated with Cyber-Syndrome, symptoms, and acupoints

Area of Science:

• Health Informatics
• Natural Language Processing
• Digital Health

Background:

• Excessive cyberspace interaction causes health problems, including physiological and psychological disorders, termed Cyber-Syndrome.
• Existing research lacks comprehensive datasets for studying Cyber-Syndrome and its treatments.
• Developing specialized corpora is crucial for advancing computational analysis of health-related text data.

Purpose of the Study:

• To construct a gold-standard, tree-shaped corpus (CS-A) annotating Cyber-Syndrome, its clinical manifestations, and relevant acupoints.
• To define six key entities and relations for annotation within the CS-A corpus.
• To enhance the visibility of Cyber-Syndrome and its health impacts, while promoting Natural Language Processing (NLP) advancements.

Main Methods:

• Manual annotation of 448 texts for Cyber-Syndrome, clinical manifestations, and acupoints.
• Iterative refinement of annotation guidelines through three update rounds.
• Calculation of inter-annotator agreement (IAA) to ensure annotation quality.

Main Results:

• The CS-A corpus was successfully created, containing annotated data on Cyber-Syndrome and related health aspects.
• Annotation guidelines were refined, leading to a significant improvement in inter-annotator agreement.
• A final inter-annotator agreement score of 86.05% was achieved, indicating high-quality annotations.

Conclusions:

• The CS-A corpus provides a foundational resource for Cyber-Syndrome research and NLP model development.
• The corpus facilitates the study of subtle health impacts associated with cyberspace over-engagement.
• CS-A supports the advancement of NLP techniques, particularly in discontinuous and nested entity recognition.