Scientific data processing · BEGINNER LESSON
Normalisation
Normalisation is a process that turns different forms of information into a common form so they can be compared or linked. In the sources, it means either mapping variant medical terms to one standard database identifier or rescaling numbers to a shared scale.
What normalisation measures or produces
The process does not measure a biological or physical property by itself. Instead, it produces a standard identifier in medical concept normalisation, or a comparable numeric scale in multiple-criteria decision analysis (MCDA, a method that compares several measured criteria). In both cases, the goal is to make different representations of the same underlying item easier to compare, combine, or retrieve.
[1] Transformer-based relation extraction and concept normalization using an annotated clinical trials corpus [2] Entropy and Normalization in MCDA: A Data-Driven ... - PMC [3] pmc.ncbi.nlm.nih.gov › articles › PMC12534389Generating detectors from anomaly samples via negative ...
How medical concept normalisation works
Medical concept normalisation (MCN, also called biomedical entity linking) takes a term written in free text and maps it to a controlled vocabulary entry such as a Unified Medical Language System (UMLS) concept unique identifier, or CUI. This helps group variant forms like 25(OH)D, 25OHD, and 25-hidroxi-vitamina-D under the same concept. The sources also note that some cases are unclear, marked with (?) , and some entities need more than one CUI when the concept is redundant or combines multiple ideas.
How numeric normalisation works in MCDA and preprocessing
In MCDA, normalisation rescales criterion values that may use different units, ranges, or magnitudes so they can be compared fairly. The sources describe vector normalisation, which scales values relative to the Euclidean norm (a length measure for a set of numbers), linear max normalisation, which uses the largest or smallest observed value, max–min normalisation, which maps values into the [0,1] interval, and sum-based normalisation, which rescales by the total sum. A similar preprocessing step is also used in data pipelines such as network intrusion detection, where normalisation can place feature values into a set interval like [0,1].
[2] Entropy and Normalization in MCDA: A Data-Driven ... - PMC [3] pmc.ncbi.nlm.nih.gov › articles › PMC12534389Generating detectors from anomaly samples via negative ...
Why normalisation is useful
For medical text, normalisation improves interoperability and supports better cohort definition across clinical trials and electronic health records (EHRs, digital patient records). For numeric comparison, it keeps a variable with large raw numbers from dominating an aggregation just because of its scale. In MCDA, this makes rankings more interpretable and comparable.
[1] Transformer-based relation extraction and concept normalization using an annotated clinical trials corpus [2] Entropy and Normalization in MCDA: A Data-Driven ... - PMC
Mechanism and limitations
Normalisation is not neutral in its effects. The MCDA source says different schemes can change preference values and rankings, especially when data are skewed or contain extreme values. Vector, max-based, max–min, and sum-based methods differ in sensitivity to outliers. The medical source also notes limits: annotation schemes may not fit other teams or tasks, and only part of the corpus was used to compute agreement scores. So normalisation depends on the chosen standard and on the structure of the data.
[1] Transformer-based relation extraction and concept normalization using an annotated clinical trials corpus [2] Entropy and Normalization in MCDA: A Data-Driven ... - PMC
SEE THE IDEA
Two common paths of normalisation
Variant terms in notes or numbers with different units and scales
Either map to a standard identifier or rescale numerically
One UMLS CUI for a concept, or comparable values such as [0,1]
Search, retrieval, aggregation, comparison, or ranking
Input data are transformed by a chosen normalisation rule into a standard form used for analysis or retrieval.
The same word can describe different processes in different fields, but both make heterogeneous data comparable.
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.