engleski

Information Extraction from Free Text Medical Data

A glance at any medical record shows that free text is the basic type in a GP’s office and it is full of words, phrases, abbreviations, very often chaotic, depending of GP’s behaviour pattern. The physician, school medicine specialist, made a large database consisting records of children taking to pre-school examination. There were data as free text and coded data, too. Data were analysed by the software package SPAD.T for statistical analysis of textual data. The first results – statistics of vocabulary showed us two things: relatively big total number of words, and small number of distinct words. The correspondence analysis for tables crossing vocabulary and individual was applied, too. The most remarkable feature was that the result persisted in the same success / failure group. The next interesting feature were the clouds of words around the success / failure category. Value of this analysis was that it showed us some results not seen easily from numerical data or coded textual data (hidden information).

medical record; free text medical data; textual statistics

nije evidentirano