Mapping query terms to data and schema using content based similarity search in clinical information systems

Area of Science:

• Clinical Informatics
• Database Management
• Natural Language Processing

Background:

• Mapping user queries to clinical database schemas (Entity Relationship and Entity Attribute Value models) is challenging.
• Existing methods lack efficiency and accuracy in handling diverse clinical data, including medication names.

Purpose of the Study:

• To develop and evaluate a similarity-based Top-K algorithm for accurate query-to-schema mapping in clinical information systems.
• To extend the Entity Attribute Value mapping to specifically address medication names.
• To detail the pre-processing steps, including Natural Language Processing (NLP) tasks, required for the mapping algorithm.

Main Methods:

• A similarity-based Top-K algorithm was employed for mapping query terms to Entity Relationship (ER) and Entity Attribute Value (EAV) models.
• Natural Language Processing (NLP) techniques were utilized for pre-processing and resource preparation.
• An extension to the EAV mapping was developed for medication names.

Main Results:

• The developed mapping algorithm achieved an accuracy rate exceeding 84% on an example clinical information system.
• The pre-processing steps, including NLP, were detailed and shown to be effective.
• The extended EAV mapping successfully handled medication names.

Conclusions:

• The Top-K algorithm provides a highly accurate solution for mapping clinical queries to database schemas.
• The integration of NLP and extended EAV mapping improves data accessibility and usability in clinical settings.
• The developed methodology will be integrated into the Clinical Data Analytics Language (CliniDAL) to automate the mapping process.