Standardizing Free-Text Data Exemplified by Age and Data-Location Fields in the Immune Epitope Database.

Area of Science:

• Biomedical Informatics
• Data Science
• Natural Language Processing

Background:

• Unstructured biomedical data, like free text, is underutilized due to extraction challenges.
• Normalizing free text allows for structured representation using ontologies and harmonized queries.
• This paper presents a tool for free-text normalization applied to Immune Epitope Database (IEDB) data.

Purpose of the Study:

• To develop an adaptable tool for normalizing free-text biomedical data.
• To evaluate the tool's effectiveness on age and data-location fields from the IEDB.
• To enhance the usability and findability of unstructured biomedical data.

Main Methods:

• Free text entries for age and data-location from the IEDB were analyzed.
• Normalization involved three steps: character, word, and phrase normalization.
• Generalizable rules were developed and applied using the presented tool.

Main Results:

• For age data, output validity reached 99.97% (character), 98.06% (word), and 83.81% (phrase).
• For data-location data, output validity reached 99.99% (character), 98.46% (word), and 97.95% (phrase).
• The normalization process significantly reduced data variance.

Conclusions:

• A generalizable approach for normalizing free-text database fields was developed.
• The one-time effort of rule creation can be applied to ongoing data curation.
• Standardization enhances data findability, usability, search functionality, and ontology linkages.