COPD Corpus Entity Annotation Automatic Entity Normalisation Annotation Format Download and Visualisation

Entity normalisation

HYPHEN Method

We have applied an automatic normalisation method (HYPHEN) to link our annotated named entities to UMLS concept identifiers (CUIs), wherever possible.

HYPHEN is a hybrid method that it employs a pipeline of different techniques to generate variations of the original NE mention (mostly based on systematic syntactic and semantic variations of the original mention) and tries to match generated variants against existing variants listed in the target terminological resources. HYPHEN employs the following six individual techniques:

1. Acronym and abbreviation expansion and context-sensitive disambiguation (e.g. Type 2 DM -> Type 2 diabetes mellitus)
2. Conversion of plural forms to singular (e.g., alveolar septa -> alveolar septum)
3. Generation of English equivalents of Neoclassical compounds (e.g. elevated blood leukocyte counts -> elevated white blood cell counts)
4. Generation of Neoclassical equivalents of English terms (e.g., tpleural inflammation -> pleuritis)
5. Generation of syntactic variants (e.g., supplemental oxygen -> oxygen supplementation)
6. Generation of synonyms (e.g., worsening pulmonary function -> deterioration of lung function)

The processes are applied in the order shown above, based on the results of experiments to determine the optimal ordering. The output of each process is passed as input to the next technique in the sequence. This can increase the accuracy of normalisation, since multiple transformations are sometimes necessary to allow mapping to the terminological resource, e.g. ,hypertensive eyes -> (singular) -> hypertensive eye -> (syntactic) -> eye hypertension -> (Neoclassical) -> ocular hypertension [UMLS: C0028840]. The pipeline is terminated as soon as one of the techniques generates a variant that can be matched in the terminological resource.

Thompson, P. and Ananiadou, S. (2018). HYPHEN: A flexible, hybrid method to map phenotype concept mentions to terminological resources. Terminology, 24(1), 91-121.

Statistics of applying HYPHEN to the annotations of the COPD Corpus

In Table 1, we report on the number of NE annotations in the COPD corpus to which HYPHEN is able to assign UMLS CUIs. As illustrated in the table, HYPHEN is able to normalise a large percentage of the entities belonging to the majority of categories in our corpus to UMLS concept identifiers.

Table 1. Number of NEs in the COPD corpus normalised to UMLS concepts by HYPHEN

Category	Total NEs	# NE normalised	% NEs normalised
Problem	2556	2151	83.15
MedicalCondition	5119	4969	97.07
RiskFactor	1211	942	77.79
SignOrSymptom	2065	1140	55.21
IndividualBehaviour	194	124	63.92
TestOrMeasureResult	685	259	37.81
Treatment	4337	3775	87.04
TestOrMeasure	3576	2609	72.96
AnatomicalConcept	2616	2372	90.67
Drug	2593	2368	91.32
Protein	820	727	87.66
Quality	1153	1015	88.03

Some examples from the corpus of successful normalisations by HYPHEN are shown in Table 2.

Table 2. Sample normalisation results

Entity annotation	Semantic Category	Mapped UMLS concept
increased PVR	Problem	increased pulmonary vascular resistance (C1867423)
lung failure	MedicalCondition	pulmonary failure (C0948755)
left atrial	AnatomicalConcept	left atrium (C0225860)/
arm training	Treatment	upper limb training (C0556501)
spirometric test	TestOrMeasure	spirometry test (C0037981)
genetic predisposition	RiskFactor	genetic susceptibility to disease (C1455997)

Visualisation of concept IDs in brat

When visualising the annotated corpus, information about NEs that have been automatically linked to concept IDs by HYPHEN can be viewed by hovering the mouse over the relevant entity. Should normalisation have been carried out for the term, then the following information will be displayed:

• CUI assigned to the entity. This appears right aligned underneath the horizontal line. This is preceded by UMLSCUI
• A preferred/heading term for the concept. This is preceded by UMLS in bold.

Note that some entitites are normalised to more than one UMLS concept. In this case, information for each of the CUIs to which the entity is mapped are shown.

For example, Figure 1 shows that the IndividualBehaviour annotation never smokers has been linked to the UMLS concept Never smoked tobacco with CUI C0425293, while in Figure 2, the Problem annotation AE-COPD has been linked to the UMLS concept Acute exacerbation of chronic obstructive airways disease with CUI C0340044.

Figure 1: Display of concept ID information an IndividualBehaviour annotation

Figure 2: Display of concept ID information for a Problem annotation