Morpheme-based, cross-lingual indexing for medical document retrieval

Area of Science:

• Medical Informatics
• Natural Language Processing
• Computational Linguistics

Background:

• Machine-readable medical documents are abundant, but current retrieval systems struggle with complex natural language phenomena.
• Existing string-matching methods are inadequate for the medical sublanguage, especially with noun compounds.
• The widespread use of complex word forms in medical texts hinders efficient information retrieval.

Purpose of the Study:

• To develop a methodology for segmenting complex medical compounds into medically relevant morphemes.
• To create a multilingual morpheme thesaurus by structuring morphemes and their semantic relations.
• To present a tool for the compilation and management of such a thesaurus.

Main Methods:

• Developed a methodology for segmenting complex compounds into medically motivated morphemes.
• Identified morphemes derived from German, Greek, and Latin roots based on medical sublanguage patterns.
• Structured a morpheme dictionary by defining semantic relations to build a multilingual morpheme thesaurus.
• Presented a tool for thesaurus compilation and management.

Main Results:

• A methodology for segmenting complex medical compounds into constituent morphemes was successfully introduced.
• The potential for creating a structured, multilingual morpheme thesaurus from these morphemes was demonstrated.
• A tool facilitating the construction and maintenance of this specialized thesaurus was presented.

Conclusions:

• The proposed methodology enhances the processing of complex medical terminology for improved information retrieval.
• A multilingual morpheme thesaurus can significantly advance the capabilities of medical document retrieval systems.
• The developed tool supports the practical implementation and upkeep of the thesaurus.