Jamaican Creole English (jam) 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 Jamaican Creole English subword embeddings in TensorFlow Projector

2-D UMAP plots

Embedding matrix plots

Training corpus sample, encoded with different BPE vocabulary sizes

Vocab sizejamwiki sample
original oersiula kaaven (ursula karven, priivosli ursula karven-veres (baan ursula ganzenmüller 00 septemba 0000 ina ulm) a wah joerman akchres, raita, magl a
intanashinali, maali mechiz aalso kantiniu fi rivoerbariet mongx vierios indijinos komiuniti. far instans, di astrielian abarijinii kantiniu fi bon wa
di panish seklment ina jumieka wena seklment we arijiniet frah di 00t senchri, wen jumieka eh panish ah galang so isenshali tel 0000, di diet wen spie
1000 ▁ oers iula ▁kaa v en ▁( u r s u la ▁k ar v en , ▁pri iv os li ▁u r s u la ▁k ar v en - v er es ▁( ba an ▁u r s u la ▁g anz en m ü l l er ▁00 ▁se p t emba ▁0000 ▁ina ▁ ul m ) ▁a ▁wah ▁j oerman ▁ak chr es , ▁r ai ta , ▁ma gl ▁a
▁intan ash in ali , ▁maali ▁me ch iz ▁aalso ▁kantin iu ▁fi ▁ri v oer ba ri et ▁m ong x ▁v ieri os ▁indi jin os ▁kom iun iti . ▁f ar ▁in st ans , ▁di ▁a st ri e lian ▁a ba ri jin ii ▁kantin iu ▁fi ▁b on ▁wa
▁di ▁panish ▁sek l ment ▁ina ▁jumieka ▁wena ▁sek l ment ▁we ▁ari jin iet ▁frah ▁di ▁00 t ▁senchri , ▁wen ▁jumieka ▁eh ▁panish ▁ah ▁g al ang ▁so ▁is en sh ali ▁t el ▁0000, ▁di ▁di et ▁wen ▁sp ie
3000 ▁ oers iula ▁kaa v en ▁( ur s ula ▁k ar v en , ▁priiv os li ▁u r s ula ▁k ar v en - v er es ▁( baan ▁u r s ula ▁g anz en m ü l l er ▁00 ▁septemba ▁0000 ▁ina ▁ ul m ) ▁a ▁wah ▁joerman ▁ak chr es , ▁raita , ▁magl ▁a
▁intan ashin ali , ▁maali ▁me chiz ▁aalso ▁kantiniu ▁fi ▁riv oer ba ri et ▁mongx ▁vierios ▁indijinos ▁komiuniti . ▁f ar ▁in stans , ▁di ▁a st ri e lian ▁a ba ri jin ii ▁kantiniu ▁fi ▁b on ▁wa
▁di ▁panish ▁seklment ▁ina ▁jumieka ▁wena ▁seklment ▁we ▁arijiniet ▁frah ▁di ▁00 t ▁senchri , ▁wen ▁jumieka ▁eh ▁panish ▁ah ▁g al ang ▁so ▁is en shali ▁tel ▁0000, ▁di ▁diet ▁wen ▁sp ie
5000 ▁ oers iula ▁kaa ven ▁( ur s ula ▁kar ven , ▁priivos li ▁u r s ula ▁kar ven - ver es ▁( baan ▁u r s ula ▁g anz en m ü l ler ▁00 ▁septemba ▁0000 ▁ina ▁ ul m ) ▁a ▁wah ▁joerman ▁ak chres , ▁raita , ▁magl ▁a
▁intanashinali , ▁maali ▁mechiz ▁aalso ▁kantiniu ▁fi ▁riv oer ba riet ▁mongx ▁vierios ▁indijinos ▁komiuniti . ▁far ▁in stans , ▁di ▁ast rie lian ▁a ba ri jin ii ▁kantiniu ▁fi ▁bon ▁wa
▁di ▁panish ▁seklment ▁ina ▁jumieka ▁wena ▁seklment ▁we ▁arijiniet ▁frah ▁di ▁00 t ▁senchri , ▁wen ▁jumieka ▁eh ▁panish ▁ah ▁g al ang ▁so ▁is en shali ▁tel ▁0000, ▁di ▁diet ▁wen ▁sp ie