Akan (ak) 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 |
Training corpus sample, encoded with different BPE vocabulary sizes
| Vocab size | akwiki sample |
|---|---|
| original | kraman yare kunkum nipa bεyε mpem aduonu asia kosi mpem aduonum anum afe biara wͻ wiase nyinaa. owu yi ͻha mu nkyekyem bεboro aduokrͻn anum si wͻ asia akwan a wͻfa so siw yare yi kwan bi ne nsu a emu tew a wͻde di dwuma ne nsu mu nwaw a wokum wͻn. wͻ mmeae a yare yi abu so no, aduru praziquantel yε n ahometew ne amannehu maa britain fo no faa adwen sɛɛ wɔbefi ghana asaase so. afe 0000 mu no, wɔhyehyɛɛ abatoɔ a edi kan wɔ abibiman afrika mu. nkrumah |
| 1000 | ▁kraman ▁yare ▁k un kum ▁nipa ▁b ε y ε ▁mpem ▁aduonu ▁asia ▁kosi ▁mpem ▁aduon um ▁anum ▁afe ▁biara ▁w ͻ ▁wiase ▁nyinaa . ▁owu ▁yi ▁ͻ ha ▁mu ▁nkyekyem ▁b ε boro ▁adu ok r ͻ n ▁anum ▁si ▁w ͻ ▁asia ▁akwan ▁a ▁w ͻ fa ▁so ▁siw ▁yare ▁yi ▁kwan ▁bi ▁ne ▁nsu ▁a ▁emu ▁te w ▁a ▁w ͻ de ▁di ▁dwuma ▁ne ▁nsu ▁mu ▁nwa w ▁a ▁wo kum ▁w ͻ n . ▁w ͻ ▁mmeae ▁a ▁yare ▁yi ▁abu ▁so ▁no , ▁aduru ▁p ra z i qu an te l ▁y ε ▁n ▁aho me te w ▁ne ▁aman ne hu ▁maa ▁britain ▁fo ▁no ▁faa ▁adwen ▁sɛ ɛ ▁wɔ be fi ▁ghana ▁as aase ▁so . ▁afe ▁0000 ▁mu ▁no , ▁wɔ h yehyɛ ɛ ▁aba to ɔ ▁a ▁edi ▁kan ▁wɔ ▁abibiman ▁afrika ▁mu . ▁nkrumah |
| 3000 | ▁kraman ▁yare ▁kun kum ▁nipa ▁b ε y ε ▁mpem ▁aduonu ▁asia ▁kosi ▁mpem ▁aduonum ▁anum ▁afe ▁biara ▁w ͻ ▁wiase ▁nyinaa . ▁owu ▁yi ▁ͻ ha ▁mu ▁nkyekyem ▁b ε boro ▁aduokr ͻ n ▁anum ▁si ▁w ͻ ▁asia ▁akwan ▁a ▁w ͻ fa ▁so ▁siw ▁yare ▁yi ▁kwan ▁bi ▁ne ▁nsu ▁a ▁emu ▁tew ▁a ▁w ͻ de ▁di ▁dwuma ▁ne ▁nsu ▁mu ▁nwaw ▁a ▁wo kum ▁w ͻ n . ▁w ͻ ▁mmeae ▁a ▁yare ▁yi ▁abu ▁so ▁no , ▁aduru ▁pra ziqu antel ▁y ε ▁n ▁aho metew ▁ne ▁amanne hu ▁maa ▁britain ▁fo ▁no ▁faa ▁adwen ▁sɛɛ ▁wɔ be fi ▁ghana ▁asaase ▁so . ▁afe ▁0000 ▁mu ▁no , ▁wɔ hyehyɛ ɛ ▁abatoɔ ▁a ▁edi ▁kan ▁wɔ ▁abibiman ▁afrika ▁mu . ▁nkrumah |
| 5000 | ▁kraman ▁yare ▁kun kum ▁nipa ▁b ε y ε ▁mpem ▁aduonu ▁asia ▁kosi ▁mpem ▁aduonum ▁anum ▁afe ▁biara ▁w ͻ ▁wiase ▁nyinaa . ▁owu ▁yi ▁ͻ ha ▁mu ▁nkyekyem ▁b ε boro ▁aduokr ͻ n ▁anum ▁si ▁w ͻ ▁asia ▁akwan ▁a ▁w ͻ fa ▁so ▁siw ▁yare ▁yi ▁kwan ▁bi ▁ne ▁nsu ▁a ▁emu ▁tew ▁a ▁w ͻ de ▁di ▁dwuma ▁ne ▁nsu ▁mu ▁nwaw ▁a ▁wo kum ▁w ͻ n . ▁w ͻ ▁mmeae ▁a ▁yare ▁yi ▁abu ▁so ▁no , ▁aduru ▁praziquantel ▁y ε ▁n ▁ahometew ▁ne ▁amanne hu ▁maa ▁britain ▁fo ▁no ▁faa ▁adwen ▁sɛɛ ▁wɔbe fi ▁ghana ▁asaase ▁so . ▁afe ▁0000 ▁mu ▁no , ▁wɔ hyehyɛɛ ▁abatoɔ ▁a ▁edi ▁kan ▁wɔ ▁abibiman ▁afrika ▁mu . ▁nkrumah |