Skip to main content
. 2019 Dec 23;19(Suppl 7):274. doi: 10.1186/s12911-019-0981-y

Table 7.

Results for Swedish

Source Feature Algorithm Precision Recall F1-score
EHR W Glove 76.03 74.42 75.22
word2vec 75.91 75.44 75.68
FastText 76.35 74.90 75.62
L Glove 76.04 72.10 74.02
word2vec 74.44 74.49 74.46
FastText 75.25 72.99 74.10
GenMed W Glove 73.14 71.15 72.13
word2vec 74.83 68.55 71.56
FastText 74.09 69.65 71.80
L Glove 74.76 67.67 71.03
word2vec 74.05 68.89 71.38
FastText 75.48 69.51 72.37
Gen W Glove 73.79 67.60 70.56
word2vec 72.50 70.33 71.40
FastText 76.18 68.49 72.13
L Glove 73.15 70.05 71.57
word2vec 72.47 68.76 70.56
FastText 73.96 68.01 70.86
EHR W,L word2vec 74.64 77.49 76.04
EHR W+L word2vec 74.43 76.26 75.34
EHR W-L word2vec 73.45 50.95 60.17

Embeddings of single base-units on top and, below, with combined base-units