BPEmb: got

Gothic (got) 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

View Gothic subword embeddings in TensorFlow Projector

2-D UMAP plots

Embedding matrix plots

Training corpus sample, encoded with different BPE vocabulary sizes

Vocab size	gotwiki sample
original	𐌰𐌽𐌿𐌳𐌹𐌱𐌰𐌿𐍂𐌲𐍃 (𐌰𐌲𐌲. ''duckburg'') 𐌹𐍃𐍄 𐌱𐌰𐌿𐍂𐌲𐍃 𐌸𐌰𐍂𐌴𐌹 𐌰𐌻𐌰𐍂𐌴𐌹𐌺𐍃 𐌰𐌽𐌿𐌸𐍃 𐌾𐌰𐌷 𐍆𐍂𐌹𐌾𐍉𐌽𐌳𐍃 𐌹𐍃 𐌱𐌰𐌿𐌰𐌽𐌳. 𐍃𐍉 𐌱𐌰𐌿𐍂𐌲𐍃 𐌻𐌹𐌲𐌹𐌸 𐌹𐌽 𐌰𐌼𐌴𐍂𐌹𐌺𐌰𐌽𐌹𐍃𐌺𐌰𐌼𐌼𐌰 𐌲𐌰𐌿𐌾𐌰 "𐌺𐌰𐌻𐌹𐍃𐍉𐍄𐌰". 𐌲𐌿𐌽𐌸𐌹𐌼𐌴𐍂𐍃, 𐍅𐌹𐍃𐌿𐌲𐌿𐍄𐌰𐌽𐌴 𐌸𐌹𐌿𐌳𐌰𐌽𐍃 𐌲𐌿𐌽𐌸𐌹𐌼𐌴𐍂𐍃 (? - •𐍇𐌹𐌱•) 𐍅𐌰𐍃 𐍅𐌹𐍃𐌹𐌲𐌿𐍄𐌰𐌽𐌴 𐌸𐌹𐌿𐌳𐌰𐌽𐍃 𐍆𐍂𐌰𐌼 𐌾𐌴𐍂𐌰 •𐍇𐌹•. 𐍅𐌰𐌻𐌳𐌿𐍆𐌽𐌹 𐌸𐌰𐌹𐍂𐌷 𐍆𐍂𐌹𐌾𐌰𐌸𐍅𐌰 𐌼𐌹𐌸 𐌷𐌻𐍉𐍄𐌰𐍂𐌰 𐌹𐌹 𐌽𐌴𐍅𐍃𐍄𐍂𐌾𐍉𐍃 𐌾𐌰𐌷 𐌸𐌹𐌿𐌳𐌰𐌱 (pdf) the ethnonym apar in the turkish inscriptions of the viii. century and armenian manuscripts - mehmet tezcan
1000	▁𐌰𐌽 𐌿𐌳 𐌹𐌱 𐌰𐌿𐍂𐌲𐍃 ▁( 𐌰𐌲𐌲 . ▁'' d uc k b ur g '') ▁𐌹𐍃𐍄 ▁𐌱𐌰𐌿𐍂𐌲𐍃 ▁𐌸𐌰𐍂𐌴𐌹 ▁𐌰𐌻 𐌰𐍂𐌴𐌹𐌺 𐍃 ▁𐌰𐌽 𐌿 𐌸𐍃 ▁𐌾𐌰𐌷 ▁𐍆𐍂𐌹 𐌾𐍉𐌽 𐌳 𐍃 ▁𐌹𐍃 ▁𐌱𐌰𐌿 𐌰𐌽𐌳 . ▁𐍃𐍉 ▁𐌱𐌰𐌿𐍂𐌲𐍃 ▁𐌻𐌹𐌲 𐌹𐌸 ▁𐌹𐌽 ▁𐌰𐌼 𐌴 𐍂𐌹 𐌺 𐌰𐌽 𐌹𐍃𐌺𐌰𐌼𐌼𐌰 ▁𐌲𐌰𐌿 𐌾𐌰 ▁" 𐌺 𐌰𐌻 𐌹𐍃 𐍉 𐍄𐌰 " . ▁𐌲 𐌿𐌽 𐌸 𐌹𐌼 𐌴𐍂 𐍃 , ▁𐍅𐌹𐍃 𐌿𐌲 𐌿𐍄𐌰𐌽𐌴 ▁𐌸𐌹𐌿𐌳𐌰𐌽𐍃 ▁𐌲 𐌿𐌽 𐌸 𐌹𐌼 𐌴𐍂 𐍃 ▁( ? ▁- ▁• 𐍇 𐌹𐌱 • ) ▁𐍅𐌰𐍃 ▁𐍅𐌹𐍃 𐌹𐌲 𐌿𐍄𐌰𐌽𐌴 ▁𐌸𐌹𐌿𐌳𐌰𐌽𐍃 ▁𐍆𐍂𐌰𐌼 ▁𐌾𐌴𐍂𐌰 ▁• 𐍇 𐌹 •. ▁𐍅 𐌰𐌻𐌳 𐌿 𐍆 𐌽 𐌹 ▁𐌸𐌰𐌹𐍂𐌷 ▁𐍆𐍂𐌹 𐌾𐌰 𐌸 𐍅𐌰 ▁𐌼𐌹𐌸 ▁𐌷𐌻 𐍉 𐍄𐌰 𐍂𐌰 ▁𐌹 𐌹 ▁𐌽𐌴 𐍅 𐍃𐍄 𐍂 𐌾𐍉𐍃 ▁𐌾𐌰𐌷 ▁𐌸𐌹𐌿𐌳𐌰 𐌱 ▁( p d f ) ▁the ▁ et h n on y m ▁a p ar ▁in ▁the ▁t ur k ish ▁in s c ri p tion s ▁of ▁the ▁v i i i . ▁c ent ur y ▁and ▁a r m en i an ▁m an us c ri p t s ▁- ▁me h m et ▁t e z c an
3000	▁𐌰𐌽 𐌿𐌳 𐌹𐌱 𐌰𐌿𐍂𐌲𐍃 ▁( 𐌰𐌲𐌲 . ▁'' duc k b ur g '') ▁𐌹𐍃𐍄 ▁𐌱𐌰𐌿𐍂𐌲𐍃 ▁𐌸𐌰𐍂𐌴𐌹 ▁𐌰𐌻 𐌰𐍂𐌴𐌹𐌺𐍃 ▁𐌰𐌽 𐌿𐌸𐍃 ▁𐌾𐌰𐌷 ▁𐍆𐍂𐌹𐌾𐍉𐌽 𐌳𐍃 ▁𐌹𐍃 ▁𐌱𐌰𐌿𐌰𐌽𐌳 . ▁𐍃𐍉 ▁𐌱𐌰𐌿𐍂𐌲𐍃 ▁𐌻𐌹𐌲𐌹𐌸 ▁𐌹𐌽 ▁𐌰𐌼𐌴𐍂𐌹 𐌺 𐌰𐌽 𐌹𐍃𐌺𐌰𐌼𐌼𐌰 ▁𐌲𐌰𐌿𐌾𐌰 ▁" 𐌺 𐌰𐌻𐌹𐍃 𐍉𐍄𐌰 ". ▁𐌲𐌿𐌽𐌸 𐌹𐌼 𐌴𐍂𐍃 , ▁𐍅𐌹𐍃𐌿𐌲𐌿𐍄𐌰𐌽𐌴 ▁𐌸𐌹𐌿𐌳𐌰𐌽𐍃 ▁𐌲𐌿𐌽𐌸 𐌹𐌼 𐌴𐍂𐍃 ▁(? ▁- ▁• 𐍇 𐌹𐌱 •) ▁𐍅𐌰𐍃 ▁𐍅𐌹𐍃𐌹𐌲𐌿𐍄𐌰𐌽𐌴 ▁𐌸𐌹𐌿𐌳𐌰𐌽𐍃 ▁𐍆𐍂𐌰𐌼 ▁𐌾𐌴𐍂𐌰 ▁• 𐍇 𐌹 •. ▁𐍅𐌰𐌻𐌳𐌿𐍆𐌽𐌹 ▁𐌸𐌰𐌹𐍂𐌷 ▁𐍆𐍂𐌹𐌾𐌰 𐌸 𐍅𐌰 ▁𐌼𐌹𐌸 ▁𐌷𐌻 𐍉𐍄𐌰 𐍂𐌰 ▁𐌹 𐌹 ▁𐌽𐌴 𐍅𐍃𐍄 𐍂 𐌾𐍉𐍃 ▁𐌾𐌰𐌷 ▁𐌸𐌹𐌿𐌳𐌰 𐌱 ▁( p d f ) ▁the ▁ et hn on y m ▁ap ar ▁in ▁the ▁t ur k ish ▁in s cri p tion s ▁of ▁the ▁v ii i . ▁c ent ur y ▁and ▁ar men ian ▁man us cript s ▁- ▁me h met ▁te z c an
5000	▁𐌰𐌽 𐌿𐌳 𐌹𐌱 𐌰𐌿𐍂𐌲𐍃 ▁( 𐌰𐌲𐌲 . ▁'' duc k b ur g '') ▁𐌹𐍃𐍄 ▁𐌱𐌰𐌿𐍂𐌲𐍃 ▁𐌸𐌰𐍂𐌴𐌹 ▁𐌰𐌻 𐌰𐍂𐌴𐌹𐌺𐍃 ▁𐌰𐌽 𐌿𐌸𐍃 ▁𐌾𐌰𐌷 ▁𐍆𐍂𐌹𐌾𐍉𐌽 𐌳𐍃 ▁𐌹𐍃 ▁𐌱𐌰𐌿𐌰𐌽𐌳 . ▁𐍃𐍉 ▁𐌱𐌰𐌿𐍂𐌲𐍃 ▁𐌻𐌹𐌲𐌹𐌸 ▁𐌹𐌽 ▁𐌰𐌼𐌴𐍂𐌹 𐌺𐌰𐌽 𐌹𐍃𐌺𐌰𐌼𐌼𐌰 ▁𐌲𐌰𐌿𐌾𐌰 ▁" 𐌺 𐌰𐌻𐌹𐍃 𐍉𐍄𐌰 ". ▁𐌲𐌿𐌽𐌸𐌹𐌼𐌴𐍂𐍃 , ▁𐍅𐌹𐍃𐌿𐌲𐌿𐍄𐌰𐌽𐌴 ▁𐌸𐌹𐌿𐌳𐌰𐌽𐍃 ▁𐌲𐌿𐌽𐌸𐌹𐌼𐌴𐍂𐍃 ▁(? ▁- ▁• 𐍇 𐌹𐌱 •) ▁𐍅𐌰𐍃 ▁𐍅𐌹𐍃𐌹𐌲𐌿𐍄𐌰𐌽𐌴 ▁𐌸𐌹𐌿𐌳𐌰𐌽𐍃 ▁𐍆𐍂𐌰𐌼 ▁𐌾𐌴𐍂𐌰 ▁• 𐍇 𐌹 •. ▁𐍅𐌰𐌻𐌳𐌿𐍆𐌽𐌹 ▁𐌸𐌰𐌹𐍂𐌷 ▁𐍆𐍂𐌹𐌾𐌰 𐌸 𐍅𐌰 ▁𐌼𐌹𐌸 ▁𐌷𐌻 𐍉𐍄𐌰 𐍂𐌰 ▁𐌹𐌹 ▁𐌽𐌴𐍅𐍃𐍄 𐍂𐌾𐍉𐍃 ▁𐌾𐌰𐌷 ▁𐌸𐌹𐌿𐌳𐌰 𐌱 ▁( p d f ) ▁the ▁et hn on y m ▁ap ar ▁in ▁the ▁t urk ish ▁in scri p tion s ▁of ▁the ▁vii i . ▁cent ury ▁and ▁ar men ian ▁man us cript s ▁- ▁me h met ▁te z c an