Interlingue (ie) subword embeddings

Vocab size vocab model 25 dim 50 dim 100 dim 200 dim 300 dim
1000 vocab model txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix
3000 vocab model txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix
5000 vocab model txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix
txt | bin
bokeh | umap | matrix

Embedding matrix plots

Training corpus sample, encoded with different BPE vocabulary sizes

Vocab sizeiewiki sample
original li inf es guvernat per un congress queli eveni chascun du annus. it have un directoratu quel incontra se trivez in li annu e un secretariatu e un chef
desde li annu 0000 zelmerlöw partiprendet in diversi talent e casting-shows. in li annus 0000 e 0000 il partiprendet in quelc sved versiones de musica
zrinjevac (croat con plen nómine ''trg nikole šubića zrinskog'') es un parc e plazza in li cité capital de croatia zagreb. li zrinjevac apartene al do
1000 ▁li ▁inf ▁es ▁guvern at ▁per ▁un ▁con g ress ▁que li ▁e ven i ▁ch asc un ▁du ▁annus . ▁it ▁have ▁un ▁direc tor atu ▁quel ▁in con tra ▁se ▁t r iv ez ▁in ▁li ▁annu ▁e ▁un ▁sec ret ari atu ▁e ▁un ▁chef
▁desde ▁li ▁annu ▁0000 ▁z el mer l ö w ▁parti pr en det ▁in ▁d ivers i ▁t al ent ▁e ▁c ast ing - s ho w s . ▁in ▁li ▁annus ▁0000 ▁e ▁0000 ▁il ▁parti pr en det ▁in ▁quelc ▁s ve d ▁v ers ion es ▁de ▁m us ica
▁z r in je v ac ▁( c ro at ▁con ▁pl en ▁nómine ▁'' t r g ▁n ik ol e ▁ š ub ić a ▁z r in s k og '') ▁es ▁un ▁par c ▁e ▁p la z z a ▁in ▁li ▁cité ▁capital ▁de ▁croatia ▁z ag re b . ▁li ▁z r in je v ac ▁apartene ▁al ▁d o
3000 ▁li ▁inf ▁es ▁guvern at ▁per ▁un ▁congress ▁que li ▁eveni ▁chascun ▁du ▁annus . ▁it ▁have ▁un ▁directoratu ▁quel ▁in con tra ▁se ▁tr iv ez ▁in ▁li ▁annu ▁e ▁un ▁sec ret ari atu ▁e ▁un ▁chef
▁desde ▁li ▁annu ▁0000 ▁z el mer l ö w ▁partiprendet ▁in ▁diversi ▁tal ent ▁e ▁c ast ing - s ho w s . ▁in ▁li ▁annus ▁0000 ▁e ▁0000 ▁il ▁partiprendet ▁in ▁quelc ▁sved ▁vers iones ▁de ▁musica
▁z rin je v ac ▁( croat ▁con ▁plen ▁nómine ▁'' t r g ▁nik ole ▁š ub ić a ▁z rin sk og '') ▁es ▁un ▁par c ▁e ▁pla z za ▁in ▁li ▁cité ▁capital ▁de ▁croatia ▁zagreb . ▁li ▁z rin je v ac ▁apartene ▁al ▁do
5000 ▁li ▁inf ▁es ▁guvern at ▁per ▁un ▁congress ▁que li ▁eveni ▁chascun ▁du ▁annus . ▁it ▁have ▁un ▁directoratu ▁quel ▁incon tra ▁se ▁tr iv ez ▁in ▁li ▁annu ▁e ▁un ▁secret ari atu ▁e ▁un ▁chef
▁desde ▁li ▁annu ▁0000 ▁z el mer l ö w ▁partiprendet ▁in ▁diversi ▁tal ent ▁e ▁cast ing - s ho ws . ▁in ▁li ▁annus ▁0000 ▁e ▁0000 ▁il ▁partiprendet ▁in ▁quelc ▁sved ▁vers iones ▁de ▁musica
▁z rin jev ac ▁( croat ▁con ▁plen ▁nómine ▁'' tr g ▁nik ole ▁š ub ić a ▁z rin sk og '') ▁es ▁un ▁parc ▁e ▁pla z za ▁in ▁li ▁cité ▁capital ▁de ▁croatia ▁zagreb . ▁li ▁z rin jev ac ▁apartene ▁al ▁do