All learning paths Learn See Try Make Settle the transliteration rules What this lesson is about Decide once how a name crosses between scripts, write it down, and hold to it across the whole batch.
A batch of Tamil shop names must also appear in Latin letters. Four people will type them. What do you settle first?
A Decide once how names, numbers and borrowed words cross between scripts. B Write the vowel and consonant rules, and settle each name by the typist's ear. C Start typing, and add a rule to the sheet each time two people differ.
Check my recall Why this matters The Aksharantar work covers 21 languages and 12 scripts, and tests foreign words and rare words separately. Those are the words your rule has to name, because that is where two people will differ.
Write a transliteration sheet that holds across a whole batch.
The Aksharantar work covers 21 languages and 12 scripts, and tests foreign words and rare words separately. Those are the words your rule has to name, because that is where two people will differ.
Decide once how names, numbers and borrowed words cross between scripts. The hard cases are personal names, place names and borrowed English words. One rule used everywhere beats the best rule used half the time. Task A batch of Tamil shop names has to be written in Latin letters as well.
Weak approach Each person spells the names the way they think they sound.
Stronger approach One sheet decides it. Long vowels are doubled. A town name is spelt the way its signboard spells it. English words stay English.
Why the stronger approach works The second batch can be searched and joined to other records. The first cannot, because one shop has four spellings.
Try a changed situation You have moved from shop names to spoken phone numbers and dates inside Marathi transcripts. Which part of the sheet has to be written first?
Make something yourself Write a transliteration sheet for one language pair: your rules, then twenty worked examples. At least five of them must be personal or place names.
Lesson 5 of 12 on Bharat-language data craft next Handle mixed script and spelling Part 2 ends in a work sample you can send
Read the primary or official source: Madhani et al., Aksharantar (arXiv)