Dataset Viewer
Auto-converted to Parquet Duplicate
id
string
kab_tfng
string
kab_latn
string
en
string
1567385
ⵜⵛⴼⵉⴹ ⴼⵍⵍⵉ ?
Tecfiḍ fell-i ?
Do you remember me?
7043947
ⴰⵔ ⵜⵉⵎⵍⵉⵍⵉⵜ!
Ar timlilit!
Good-bye!
2294332
ⵜⵙⵙⴰⵡⵍⴹ ⴰⵙ ⵉ ⵓⵎⴷⴷⴰⴽⵍ ⵉⴽ ⵉⵍⵍⴰⵏ ⴷⴳ ⴽⴰⵏⴰⴷⴰ?
Tessawleḍ-as i wemddakel-ik yellan deg Kanada?
Did you call your friend in Canada?
2294528
ⴰⵍⵍⵏ ⵉⵡ ⴷ ⵜⵉⴱⵔⴽⴰⵏⵉⵏ.
Allen-iw d tiberkanin.
I have black eyes.
2294620
ⵓⵖⴻⵖ ⴷ ⵜⴰⵢⵓⴳⴰ ⵏ ⵡⴰⵔⴽⴰⵙⵏ.
Uɣeɣ-d tayuga n warkasen.
I bought a pair of shoes.
2296647
ⴰⵙⴳⵣⴰⵡⴰⵍ ⴰ, ⵏ ⵓⵍⵜⵎⴰ.
Asegzawal-a, n weltma.
This dictionary is my sister's.
2771352
ⵓⵔ ⵜⵜⵛⵉⵖ ⴰⵔⴰ.
Ur ččiɣ ara.
I haven't eaten.
2771573
ⴰⵎⵎⵓⵙ ⵉⴳⵍⵍⵓ ⴷ ⵙ ⵡⴰⵟⵟⴰⵏⴻⵏ.
Ammus igellu-d s waṭṭanen.
Filth breeds illnesses.
2789494
ⵓⵔⵄⴰⴷ ⵖⵔⵉⵖ ⴰⵖⵎⵉⵙ ⵏ ⵡⴰⵙⵙ ⴰ.
Urεad ɣriɣ aɣmis n wass-a.
I haven't read today's newspaper yet.
2790536
ⵖⵓⵔⵙ ⵙⵉⵏ ⵉⵎⵛⴰⵛ : ⵢⵉⵡⵏ ⴷ ⴰⴱⵔⴽⴰⵏ ⵡⴰⵢⴹ ⴷ ⴰⵎⵍⵍⴰⵍ.
Ɣur-s sin yemcac : yiwen d aberkan wayeḍ d amellal.
He keeps two cats: one is black, and the other white.
2807521
ⵉⵙⵙⵔⵖ ⴰⵅⵅⴰⵎ ⵉⵙ.
Yesserɣ axxam-is.
He set his house on fire.
3675223
ⵖⵓⵔⵡⵏⵜ ⵙⴱⵄⴰ ⵡⴰⵔⵔⴰⵛ.
Ɣur-went sebεa warrac.
You have seven children.
3678725
ⵜⵓⵔⴰ, ⵓⵃⵡⴰⴷⵊⵖ ⵟⵓⵎ.
Tura, uḥwaǧeγ Ṭum.
I need Tom now.
3678730
ⵟⵓⵎ, ⵓⵃⵡⴰⴷⵊⵖ ⵜ ⵜⵓⵔⴰ.
Ṭum, uḥwaǧeγ-t tura.
I need Tom now.
3678732
ⵓⵃⵡⴰⴷⵊⵖ ⵟⵓⵎ ⵜⵓⵔⴰ.
Uḥwaǧeγ Ṭum tura.
I need Tom now.
3678735
ⵓⵃⵡⴰⴷⵊⵖ ⵜⵓⵔⴰ ⵟⵓⵎ
Uḥwaǧeγ tura Ṭum
I need Tom now.
3679112
ⵉⵙⵎ ⵉⵡ ⵢⵓⴱⴰ.
Isem-iw Yuba.
My name is Juba.
2295741
ⵉⴷⴷⵔⵖⵍ.
Yedderɣel.
He went blind.
4714129
ⵢⵢⴰⵖ ⴰⴷ ⵜⵛⵛⴹ!
Yyaɣ ad tecceḍ!
Come eat!
2771621
ⴳⵔⵔⵣⵖ.
Gerrzeɣ.
I'm fine.
5507673
ⵙⵍⵉⵖ ⴷ ⴰⴽⴽⵏ ⵏⵏⵉⴹⵏ.
Sliɣ-d akken nniḍen.
I heard otherwise.
5507689
ⵓⵔ ⵣⵎⵉⵔⵖ ⴰⵔⴰ ⴰⴷ ⵅⵎⵎⵖ ⴰⴽⴽⵏ ⵏⵏⵉⴹⵏ.
Ur zmireɣ ara ad xemmeɣ akken nniḍen.
I can't think otherwise.
5507691
ⵓⵔ ⵣⵎⵉⵔⵖ ⴰⵔⴰ ⴰⴷ ⵅⴷⵎⵖ ⴰⴽⴽⵏ ⵏⵏⵉⴹⵏ.
Ur zmireɣ ara ad xedmeɣ akken nniḍen.
I couldn't do otherwise.
5507692
ⵎⵓⵃⴰⵍ ⴰⴷ ⵜⵅⵎⵎⴹ ⴰⴽⴽⵏ ⵏⵏⵉⴹⵏ.
Muḥal ad txemmeḍ akken nniḍen.
Don't ever think otherwise.
5507698
ⵓⵔ ⵉⵢⵉ ⵉⵖⵍⵍⴱ ⵢⵉⵡⵏ ⴰⴽⴽⵏ ⵏⵏⵉⴹⵏ.
Ur iyi-iɣelleb yiwen akken nniḍen.
Nobody will convince me otherwise.
5507702
ⵢⵉⵡⵏ ⵓⵔ ⵉⵣⵎⵉⵔ ⴰⴷ ⵉⵢⵉ ⵉⵖⵍⴱ ⴰⴽⴽⵏ ⵏⵏⵉⴹⵏ.
Yiwen ur yezmir ad iyi-iɣleb akken nniḍen.
Nobody can convince me otherwise.
5507718
ⵄⵕⴹⵖ ⴰⴷ ⵖⵍⴱⵖ ⵟⵓⵎ ⴰⴽⴽⵏ ⵏⵏⵉⴹⵏ.
Ɛerḍeɣ ad ɣelbeɣ Tom akken nniḍen.
I tried to convince Tom otherwise.
5507911
ⴰⵣⵣⵍ ⵖⵉⵡⵍ, ⵎⵓⵍⴰⵛ ⴰⴷ ⴰⴽ ⵉⴷⴷⵊ ⵍⴽⴰⵕ.
Azzel ɣiwel, mulac ad ak-yeǧǧ lkaṛ.
Run fast, otherwise you will miss the bus.
5507920
ⴰⵣⵣⵍ ⵄⵊⵍ, ⵎⵓⵍⴰⵛ ⴰⴷ ⴰⴽ ⵜⴷⴷⵊ ⵜⵎⴰⵛⵉⵏⵜ.
Azzel ɛjel, mulac ad ak-teǧǧ tmacint.
Run faster, otherwise you will miss the train.
5507925
ⵄⵊⵍ, ⵎⵓⵍⴰⵛ ⵓⵔ ⵜⴻⵜⵜⵍⴰⵃⴰⵇⴹ ⴰⵔⴰ ⵍⴼⴰⴹⵓⵕ.
Ɛjel, mulac ur tettlaḥaqeḍ ara lfaḍur.
Hurry up, otherwise you'll be late for lunch.
5507938
ⵜⵓⵔⴰ ⵎⵉ ⵎⵇⵇⵔⵖ, ⵜⵜⵅⵎⵎⵉⵎⵖ ⵖⴼ ⵡⴰⵢⵏ ⵏⵏⵉⴹⵏ.
Tura mi meqqreɣ, ttxemmimeɣ ɣef wayen nniḍen.
Now that I'm a grownup, I think otherwise.
5507946
ⵏⵜⵜⴰ ⴷ ⴰⵜⵛⴰⵇⵍⴰⵍ, ⵎⵄⵏⴰ ⴷ ⴰⵇⵛⵉⵛ ⵉⵍⵀⴰⵏ ⴷⴳ ⵡⴰⵢⵏ ⵏⵏⵉⴹⵏ.
Netta d ačaqlal, meɛna d aqcic yelhan deg wayen nniḍen.
He is noisy, but otherwise a very nice boy.
5508028
ⴰⵣⴰⵍ ⵏ ⵓⵎⵍⵢⴰⵕ ⵉ ⵉⵀⴷⴷⵕⵏ ⵜⴰⴳⵍⵉⵣⵉⵜ.
Azal n umelyaṛ i iheddṛen taglizit.
English is spoken by about a billion people.
2799850
ⵉⵡⵡⴰ?
Yewwa?
Is it baked?
2296134
ⴼⴽ ⵉⵢⵉ ⴷ ⴰⵎⴷⵢⴰ.
Efk-iyi-d amedya.
Give me an example.
9472533
ⵎⵥⵥⵉⵢⵖ.
Meẓẓiyeɣ.
I am young.
6707613
ⴰⴽⴷⴷⴰⴼ!
Akeddav!
Liar!
6715863
ⵙⴰⵎⵉ ⵓⵔ ⵉⵏⵖⵉ ⴰⵔⴰ ⵉⵎⴰⵏ ⵉⵙ.
Sami ur yenɣi ara iman-is.
Sami didn't kill himself.
6715871
ⴳⵣⵎⵖ ⵜ ⵖⴼ ⵙⵉⵏ ⵉⵎⵎⵓⵔⵏ.
Gezmeɣ-t ɣef sin immuren.
I've cut it into two pieces.
6715871
ⴳⵣⵎⵖ ⵜ ⵖⴼ ⵙⵉⵏ ⵉⵎⵎⵓⵔⵏ.
Gezmeɣ-t ɣef sin immuren.
I cut it into two pieces.
6715880
ⵙⴰⵎⵉ ⵉⵎⵎⵓⴳⵔ ⴷ ⵍⴰⵢⵍⴰ.
Sami yemmuger-d layla.
Sami met Layla.
6715886
ⵙⴰⵎⵉ ⵢⵙⵃⴱⵉⴱⵉⵔ ⵖⴼ ⵍⴰⵢⵍⴰ.
Sami yseḥbibir ɣef Layla.
Sami cared about Layla.
6715888
ⵙⴰⵎⵉ ⵢⵓⴽⵔ ⵜⵉⵖⴰⵡⵙⵉⵡⵉⵏ ⵙⴳ ⵓⵅⵅⴰⵎ ⵏ ⵍⴰⵢⵍⴰ.
Sami yuker tiɣawsiwin seg uxxam n Layla.
Sami stole items from Layla's home.
6715889
ⵙⴰⵎⵉ ⵉⵏⵏⴰ ⴷ ⵖⴼ ⵍⴷⵊⴰⵍ ⵉⵙ ⵉ ⵉⵎⵎⵓⵜ ⴱⴰⴱⴰⵙ
Sami yenna-d ɣef lǧal-is i yemmut baba-s
Sami said it was his fault that his father was dead.
6715894
ⵙⴰⵎⵉ ⵉⵙⵡⴰ ⵉⵔⵏⴰ ⵉⴽⵉⵢⴼ.
Sami yeswa yerna ikeyyef.
Sami drank and smoked.
6715898
ⵙⴰⵎⵉ ⵓⵔ ⵉⵡⴰⵍⴰ ⴰⵔⴰ ⵢⴰⵔⴽ ⵍⴰⵢⵍⴰ ⵜⴻⵜⵜⵔⵓ.
Sami ur iwala ara yark Layla tettru.
Sami didn't see Layla cry at all.
6715904
ⵙⴰⵎⵉ ⵓⵔ ⴷ ⵉⵔⵓ ⵢ ⴰⵔⴰ ⵢⴰⵔⴽ ⴷⵉ ⵍⴷⵊⴰⵏⴰⵣⴰ.
Sami ur-d iru y-ara yark di lǧanaza.
Sami didn't cry at the funeral.
6715924
ⵙⴰⵎⵉ ⵢⵓⴹⵏ.
Sami yuḍen.
Sami got sick.
6715928
ⵙⴰⵎⵉ ⵉⵜⵇⵍⵍⴰⵇ ⴰⵟⴰⵙ ⵉⵔⵏⴰ ⵢⵓⴳⴰⴷ.
Sami yetqellaq aṭas yerna yugad.
Sami was very nervous and scared.
6715930
ⵙⴰⵎⵉ ⵉⵜⵇⵍⵍⴰⵇ ⴰⵟⴰⵙ.
Sami yetqellaq aṭas.
Sami was extremely nervous.
6715931
ⵙⴰⵎⵉ ⵉⵙⵙⴰⵡⵍ ⵉ ⴱⵓⴳⴰⵟⵓ.
Sami issawel i bugaṭu.
Sami hired a lawyer.
6715932
ⵙⴰⵎⵉ ⵉⵙⵙⴰⵡⵍ ⵉ ⵉⵎⵎⴰⵙ.
Sami issawel i yemma-s.
Sami called his mother.
6715934
ⵙⴰⵎⵉ ⵉⵙⵙⴽⵔⴰ ⴰⵅⵅⴰⵎ ⵉ ⵍⴰⵢⵍⴰ.
Sami yessekra axxam i Layla.
Sami rented the house for Layla.
6715936
ⵉⵣⵎⵔ ⴷ ⵙⴰⵎⵉ ⵉ ⵉⵏⵖⴰⵏ ⵍⴰⵢⵍⴰ
Izmer d Sami i yenɣan Layla
Sami could have murdered Layla.
6715937
ⵙⴰⵎⵉ ⵉⵇⵔⵃⵉⵜ ⵓⵄⵔⵓⵔ ⵉⵙ.
Sami iqerḥit weɛrur-is.
Sami had a bad back.
6715939
ⵙⴰⵎⵉ ⴷ ⵜⵉⵎⵏⵖⵉⵡⵜ ⵉ ⵉⵏⵖⴰ ⵉⵎⴰⵏ ⵉⵙ.
Sami d timenɣiwt i yenɣa iman-is.
Sami's death was a suicide.
6715941
ⵙⴰⵎⵉ ⵉⵥⵥⴰ ⵜⵉⵊⴷⴷⵊⵉⴳⵉⵏ.
Sami yeẓẓa tijeǧǧigin.
Sami planted flowers.
6715943
ⵙⴰⵎⵉ ⵉⵙⵄⴰ ⴰⴽⴽⵏ ⵙⵜⵜⴰ ⵉⵙⴳⴳⴰⵙⵏ.
Sami yesɛa akken setta iseggasen.
Sami was about six.
6715944
ⴰⵅⵏⴼⵓⵛ ⵏ ⵙⴰⵎⵉ ⵉⵍⵍⴰ ⵉⵍⴷⵉ.
Axenfuc n Sami yella yeldi.
Sami's mouth was open.
6715945
ⵙⴰⵎⵉ ⵢⵓⵎⵏ ⵉⵎⵎⴰⵙ.
Sami yumen yemma-s.
Sami trusted his mom.
6715947
ⵙⴰⵎⵉ ⵉⵔⵓⵃ ⵙ ⵓⵙⵓ.
Sami iruḥ s usu.
Sami went to bed.
6715950
ⵙⴰⵎⵉ ⵉⵍⵍⴰ ⴷⵉ ⵜⴰⵀⵉⵜⵉ.
Sami yella di Tahiti.
Sami was in Tahiti.
6715951
ⵙⴰⵎⵉ ⵢⵓⴽⵔ ⵜⴰⵎⴷⴰⴽⵍⵜ ⵏ ⴼⴰⵔⵉⴷ.
Sami yuker tamdakelt n Farid.
Sami stole Farid's girlfriend.
6715954
ⵙⴰⵎⵉ ⵓⵔ ⵢⵓⵎⵉⵏ ⴰⵔⴰ ⵍⴰⵢⵍⴰ ⵎⴰⵣⴰⵍ ⵜⴷⴷⵔ.
Sami ur yumin ara Layla mazal tedder.
Sami couldn't believe Layla was alive.
6715956
ⵙⴰⵎⵉ ⴷ ⴼⴰⵔⵉⴷ ⵍⵍⴰⵏ ⴰⴽⴽⵏ ⴷⵉ ⵍⵃⴱⵙ.
Sami d Farid llan akken di lḥebs.
Sami and Farid shared the same prison cell.
6715962
ⵙⴰⵎⵉ ⵉⵙⵡⴰ ⵔⴷ ⴱⵓⵍⵍ.
Sami yeswa Red Bull.
Sami drank Red Bull.
6715963
ⵙⴰⵎⵉ ⵉⵍⵍⴰ ⵉⵜⵜⵉⴷⵉⵔ ⴰⴽⴽⴷ ⵅⴰⵍⵜⵉⵙ.
Sami yella yettidir akked xalti-s.
Sami lived with his aunt.
7000984
ⴰⵍⵍⴰ ⵢⵉⵡⵏ ⵓⵎⵙⵉⵎⵖⵔ ⵉⵜⴷⴷⵓⵏ.
Alla yiwen umsimɣer iteddun.
Only one speaker is working.
7014554
ⴱⴷⴻⴷ !
Bded !
Get up!
7014556
ⵟⵟⴼ ⵉⵜ ⵉⴷ.
Ṭṭef-it-id.
Grab him.
7014560
ⵜⴻⵜⵜⴷⴷⵓ.
Tetteddu.
She walks.
7014562
ⵉⴳⵔⵔⵣ
Igerrez
Well done!
7014565
ⵓⵍⴰⵛ ⴰⵖⵉⵍⵉⴼ
Ulac aɣilif
No problem.
7014560
ⵜⴻⵜⵜⴷⴷⵓ.
Tetteddu.
It works.
7014571
ⵜⴻⵜⵜⵔⵓ ⵉⵎⵎⴰ.
Tettru yemma.
Mama cried.
7014572
ⴷⴷⵎ ⵉⵜ.
Ddem-it.
Take it.
7014574
ⵉⵏⵉ ⵢⵉ ⴷ.
Ini-yi-d.
Tell me.
2294723
ⵓⵏⴼⵏ ⵉⵢⵉ ⴰⴷ ⵔⵓⵃⵖ.
Unfen-iyi ad ruḥeɣ.
They let me go.
7100797
ⴷⴷⵊⴰⵏ ⵉⵢⵉ ⴰⴷ ⵔⵓⵃⵖ.
Ǧǧan-iyi ad ruḥeɣ.
They let me go.
7024394
ⵜⵉⴽⵜⵉ ⵉⵏⵓ ⵜⵎⵅⴰⵍⴰⴼ ⵖⴼ ⵜⵉⵏ ⵉⵏⴽ.
Tikti-inu temxalaf ɣef tin-inek.
I think different from you.
7024422
ⵓⵔⴷⵊⵉⵏ ⴰⴷ ⵏⴰⵃⴱⵙ ⴰⵍⵎⴰⴷ.
Werǧin ad naḥbes almad.
We never stop learning.
7026877
ⴷⴷⵊ ⴰⵇⵊⵓⵏ ⴷⴳ ⴱⵕⵕⴰ.
Eǧǧ aqjun deg beṛṛa.
Leave the dog outside.
7026879
ⵜⵣⵎⵕⴹ ⴷⴰⵖⵏ ⴰⴷ ⵜⵎⵙⵍⴰⵢⴹ ⵜⴰⴼⵔⴰⵏⵚⵉⵙⵜ?
Tzemreḍ daɣen ad tmeslayeḍ Tafranṣist?
Can you also speak French?
7028101
ⵉⵍⵍⴰ ⵡⴰⵙⵙ ⵉⴷⴳ ⵉⵎⵛⵜⵜⵛⴰⵡ ⵢⵉⵡⵏ ⴷ ⵓⵡⴰⵖⵣⵏ?
Yella wass ideg yemceččaw yiwen d uwaɣzen?
Has anyone ever faced an ogre?
7027275
ⵉⵍⵍⴰ ⵡⴰⵙ ⵉⴷⴳ ⵜⵏⵏⵓⵖⴹ ⴷ ⵓⵡⴰⵖⵣⵏ?
Yella was ideg tennuɣeḍ d uwaɣzen?
Have you ever fought with an ogre?
7028506
ⵡⵉⵏ ⵉⵇⵇⵙ ⵓⵣⵔⵎ, ⵉⵜⵜⴰⴳⵡⴰⴷ ⴰⵙⵖⵡⵏ.
Win yeqqes uzrem, yeţţagwad aseɣwen.
Whoever has been bitten by a snake fears the slightest stick.
7028589
ⵃⵓⵙⵖ ⵛⵡⵉⵢⴰ ⴽⴰⵏ ⵉ ⵎⵉ ⵓⵔ ⵣⵎⵉⵔⵖ ⴰⵔⴰ ⴰⴷ ⴷ ⴰⵙⵖ ⴰⵙⵙ ⴰ.
Ḥuseɣ cwiya kan i mi ur zmireɣ ara ad d-aseɣ ass-a.
I feel bad that I can't come today.
7028592
ⵜⴻⵜⵜⴰⵕⵓⴹ ⵜⴰⴱⵔⴰⵜ?
Tettaruḍ tabrat?
Are you writing a letter?
7028594
ⴷ ⵜⴰⴱⵔⴰⵜ ⵉ ⵜⴻⵜⵜⴰⵕⵓⴹ?
D tabrat i tettaruḍ?
Are you writing a letter?
7028599
ⵉⵜⵜⴱⵉⵏ ⴷ ⴷ ⵉⵍⵎⵥⵉ, ⵓⵔ ⵉⵣⵎⵉⵔ ⴰⵔⴰ ⴰⴷ ⵢⵉⵍⵉ ⴰⴽⵜⵔ ⵉⵡ.
Ittbin-d d ilemẓi, ur izmir ara ad yili akter-iw.
He looks young. He cannot be older than I.
7028617
ⴰⵏⵡⴰ ⴰⵔ ⴰⴷ ⵜⵃⵎⵎⵍⴹ, ⵎⴰ ⵜⵓⵖ ⵜⵎⴰⵍⴹ?
Anwa ar ad tḥemmeleḍ, ma tuɣ tmaleḍ?
We no longer love when we are in love.
7028640
ⴰⵎⴽ ⴰⵔⴰ ⵜⵃⵎⵎⵍⴹ ⵜⴰⵇⵛⵉⵛⵜ ⵉⴷⴳ ⵓⵔ ⴷ ⵉⴱⴰⵏ ⵡⴰⵔⴰ?
Amek ara tḥemmleḍ taqcict ideg ur-d iban wara?
How can one love a girl who remains hidden?
7028649
ⴰⵎⴰⵏ ⵓⵔ ⴷ ⵜⵓⴳⵉⵎ ⵉⵎⵎⴰ, ⵓⵔ ⵜⴽⴽⵙⵏ ⴼⴰⴷ ⴰⵢ ⴰⵜⵎⴰ.
Aman ur-d tugim yemma, ur tekkesen fad ay-atma.
The water that my mother will not give me from her jug will never quench my thirst.
7030045
ⵍⴰⵢⵍⴰ ⵜⵙⵍⴰ ⵉ ⵍⵀⴰⵔⴰⴷⵊ.
Layla tesla i lharaǧ.
Layla heard the noise.
7030050
ⵍⵀⴰⵔⴰⴷⵊ ⴰⴳⵉ ⵉⵙⵄⵓⵥⵓⴳ ⵉⵎⵥⴰⵖ
Lharaǧ-agi yesɛuẓug imeẓaɣ
This noise is deafening.
7030054
ⵍⴰⵢⵍⴰ ⵜⵙⵍⴰ ⵉ ⵍⵀⴰⵔⴰⴷⵊ ⴷⵉ ⴱⵕⵕⴰ.
Layla tesla i lharaǧ di beṛṛa.
Layla heard a noise outside.
7030122
ⵙⵓ ⴰⵎⴰⵏ.
Sew aman.
Drink water.
7030125
ⵉⵎⵎⵓⵜ?
Yemmut?
Dead?
7030126
ⵜⵎⵎⵓⵜ?
Temmut?
Dead?
7030130
ⵜⵎⵎⵓⵜⴹ?
Temmuteḍ?
Are you dead?
End of preview. Expand in Data Studio

Dataset Card for Kabyle Latin-to-Tifinagh Parallel Corpus

This dataset provides a parallel corpus of the Kabyle language ($\text{Taqbaylit}$), pairing native Latin-based orthography with automated, context-aware Neo-Tifinagh transliterations. It is built by processing clean source sentences through an algorithmic engine designed to preserve grammar structures and protect foreign loanwords.

Dataset Details

Dataset Sources

  • Repository: [More Information Needed]
  • Source Project: Tatoeba Translation Corpus

Uses

Direct Use

  • Training and fine-tuning machine translation models (e.g., Latin-to-Tifinagh tokenizers).
  • Building automated transliteration tools and orthographic normalizers for Amazigh.

Out-of-Scope Use

This dataset is optimized specifically for standard latin orthography with slight modifications to get clearn Tifinagh orthography version.

Dataset Structure

The dataset consists of a tab-separated values (.tsv) file containing the following fields:

  • Sentence ID: Unique identifier inherited from the Tatoeba database.
  • Language Code: The ISO 639-3 code for Kabyle (kab).
  • Tifinagh Text: The processed Neo-Tifinagh transliteration column.
  • Latin Text: The original source Kabyle sentence written in the standard Latin script.

Dataset Creation

Curation Rationale

While the Kabyle language is widely written using a modified Latin alphabet, Neo-Tifinagh holds official and cultural prominence in the region of North Africa. This dataset bridges the script divide algorithmically, creating a high-fidelity parallel text repository that filters out foreign names and isolates native phonetic behaviors.

Source Data

Data Collection and Processing

The underlying source sentences are pulled from the Kabyle subset of the Tatoeba project. The text was passed through a specialized Python processing pipeline applying several linguistic constraints:

  • Grammatical Suffix Merging: Structural dashes (-) used for indirect objects, prepositions, or kinship suffixes (e.g., fell-ak, baba-s, tama-m) are cleanly removed and merged into a single word token (yielding ⴱⴰⴱⴰⵙ instead of ⴱⴰⴱⴰ ⵙ). All other structural punctuation dashes are globally normalized to spaces.
  • Contextual Schwa ($\mathbf{e}$ / ⴻ) Optimization: The engine automatically strips the schwa () if it appears at the absolute beginning of a word. Internally, it strips when sandwiched between two different consonants, but strictly preserves it between identical consonants.
  • Foreign Word Isolation: To prevent phonetic corruption of loanwords as much as possible, the system flags and skips transliteration entirely (keeping words in Latin script) if a token meets any of these criteria:
    • Contains French/Foreign accented vowels (é, è, à, etc.).
    • Contains the vowel o or O.
    • Terminates with the letter e or E.
    • Is an all-caps acronym (e.g., MRI, M.R.I.).
    • Contains a standalone uppercase letter inside or at the end of a sentence (e.g., vitamins like C).
  • Targeted Bypasses: Standalone instances of high-frequency words like Ait/ait are explicitly targeted to resolve to ⴰⵢⵜ. Ubiquitous name patterns common in Tatoeba text like Tom, tom, Ṭom, and ṭum bypass the foreign letter filters and map directly to ⵟⵓⵎ.
  • Direct Digraph Conversions: Common digraph inputs map explicitly to unified sounds without triggering foreign blocks (sh/ch $\rightarrow$ ⵙⵀ/ⵛⵀ, and ph $\rightarrow$ ).

Who are the source data producers?

The original source sentences are crowdsourced from independent contributors and native speakers on the open-source Tatoeba platform.

Personal and Sensitive Information

The source sentences consist of general conversational text, educational translations, and daily idioms. There is no intentionally collected private, personal, or sensitive information.

Bias, Risks, and Limitations

Because the script relies on a rule-based algorithmic filter to catch foreign structures, exceptional edge cases might slip through:

  • A native word spelling that happens to match an accidental foreign filter rule could remain in Latin text.
  • Unusual foreign names missing the standard vowel markers (o, trailing e, accents) might be forcefully transliterated into Tifinagh incorrectly.

Citation

BibTeX:

@misc{kabyle_latin_tifinagh_2026,
  title={Kabyle Latin-to-Tifinagh Parallel Corpus},
  author={Abdelhaque Id Ali},
  year={2026},
  publisher={Hugging Face Hub}
}
Downloads last month
41