Leveraging Knowledge Graphs and Natural Language Processing for Automated Web Resource Labeling and Knowledge...

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• Patients and families require trusted online information, especially for neurodevelopmental disorders (NDDs).
• Existing information access is limited by a lack of shared terminology and efficient web resource labeling for NDDs.
• Neurodevelopmental disorders affect a significant portion of the population, with substantial social and economic consequences.

Purpose of the Study:

• To develop a natural language processing (NLP) pipeline for labeling web resources relevant to NDDs.
• To create a weighted knowledge graph representing these labeled resources for improved accessibility.
• To address the challenges in sharing and accessing NDD information for various stakeholders.

Main Methods:

• Assembled a dataset of NDD web resources from expert and organizational databases.
• Scraped website text to create a corpus for knowledge graph construction.
• Applied NLP techniques including named entity recognition, topic modeling, document classification, and location detection.

Main Results:

• Developed an automated resource annotation pipeline using NLP algorithms.
• Constructed a structured knowledge graph with 78,181 annotations derived from standard and free-text vocabularies.
• Created a resource search interface utilizing the knowledge graph and an ordered weighted averaging operator for query-based ranking.

Conclusions:

• An automated labeling pipeline for NDD web resources was successfully developed.
• Demonstrated the potential of AI methods like NLP and knowledge graphs for enhancing knowledge extraction and mobilization in medicine.
• Highlighted the applicability of these AI-driven approaches in other medical domains.