Enriching the structure of the UMLS semantic network

Area of Science:

• Medical Informatics
• Ontology Engineering
• Knowledge Representation

Background:

• The Unified Medical Language System's Semantic Network (SN) uses a restrictive two-tree structure.
• Current semantic type hierarchies limit specialization by allowing only one parent.
• Proposed partitions of the SN exhibit disconnected structures, hindering semantic validity.

Purpose of the Study:

• To introduce a methodology for enhancing the SN by identifying and adding missing IS-A links.
• To transform the SN into a Directed Acyclic Graph (DAG) structure.
• To improve the connectivity and semantic validity of SN partitions.

Main Methods:

• Developed a methodology to detect and incorporate missing IS-A relationships within the SN.
• Implemented a process to convert the SN into a Directed Acyclic Graph (DAG).
• Applied the methodology to proposed SN partitions to assess connectivity.

Main Results:

• Successfully transformed the SN into a DAG, allowing semantic types multiple parents.
• The methodology successfully identified and added missing IS-A links.
• Resulting SN partitions demonstrated improved connectivity, satisfying the semantic validity principle.

Conclusions:

• The proposed methodology effectively enhances the SN's structure and connectivity.
• The DAG structure overcomes the limitations of the original two-tree hierarchy.
• This work provides a robust approach for assessing and improving semantic network validity.