Skip to main content

Table 5 Description of the setup of the three experiments

From: Formalizing biomedical concepts from textual definitions

 

Modeling

Feature engineering

Data

F-score

 

Corpora

Relation set

Relationship

Lexical

Semantic

Size

Without types

With types

Exp1

MeSH

SNOMED CT

InfRB

3-grams

—

424

74%

99.1%

Exp2

SemRep

SemRep

SemRep

3-grams

UMLS

1,357

51%–54%

94%

Exp3

WIKI+D4D

SNOMED CT

InfRB

3-grams

SNOMED CT

9,292

58%–70%

100%

  1. In all experiments Support Vector Machines are used.