PPR-SSM

PPR-SSM applies Personalized PageRank and semantic similarity measures to link biomedical named entities to concepts in domain-specific ontologies.

Key Features:

• Personalized PageRank (PPR): Uses Personalized PageRank to generate and rank candidate concepts by propagating importance through ontology relationships.
• Semantic Similarity Measures (SSM): Computes semantic similarity measures to weight edges between candidate concepts and evaluate semantic proximity.
• Domain-Specific Ontologies: Operates on domain-specific ontologies, including chemical compounds, phenotypes, and gene-product localization and processes.
• Graph-Based Methodology: Constructs a graph of candidate concepts and does not require training data.

Scientific Applications:

• Entity Linking: Links named entities in biomedical literature to knowledge-base concepts, facilitating document retrieval and relation extraction.
• Improving Accuracy: Demonstrates superior performance over existing entity linking methods across four gold standards, achieving a 0.1385 improvement in chemical compound linking accuracy compared to methods without semantic similarity measures.
• Drug Design and Development: Applies to dynamic domains such as drug design and development where new concepts frequently emerge.

Methodology:

Constructs a graph from domain-specific ontologies, applies Personalized PageRank to identify and rank candidate concepts for entities mentioned in documents, and applies semantic similarity measures to weight and assess relationships between those candidates.