Table 1 Results of the nearest neighbors (NN) experiments with different word embedding models
From: MedLexSp – a medical lexicon for Spanish medical natural language processing
Word embedding model | OOVs | % of NN | OOVs mapped to CUI | % of OOVs mapped to CUI |
|---|---|---|---|---|
SBCWE, uncased, | 740 | 74.0% | 48 | 6.49% |
SkipGram, d=100 | ||||
SBCWE, uncased, | 742 | 74.2% | 51 | 6.88% |
CBOW, d=100 | ||||
SBCWE, uncased, | 762 | 76.2% | 45 | 5.91% |
SkipGram, d=50 | ||||
SBCWE, uncased, | 732 | 73.2% | 47 | 6.42% |
CBOW, d=50 | ||||
SBCWE, uncased, | 752 | 75.2% | 46 | 6.12% |
SkipGram, d=300 | ||||
SBCWE, uncased, | 741 | 74.1% | 50 | 6.75% |
CBOW, d=300 | ||||
COVID-19 corpus, uncased, | 677 | 67.7% | 56 | 8.27% |
SkipGram, d=100, min=5 | ||||
COVID-19 corpus, uncased, | 690 | 69.0% | 78 | 11.30% |
SkipGram, d=100, min=3 |