Jürg Schneider

The first Grammar of the Malay language, 1736

When Europeans – seafarers, traders and colonists – arrived in Southeast Asia and the Indonesian archipelago in the 16th century, they needed Malay primarily in two areas – to interact with the courts in Sumatra and the Malay peninsula, and in in trade and communication across the Archipelago. But there was no formal written grammar.
From the title page of Werndly's Malay Grammar

“Grammar” is the syste­ma­tic frame­work of rules that governs the struc­tu­re, com­po­si­ti­on, and use of a natu­ral lan­guage. For humans stu­dy­ing lan­guages, a text­book grammar has often been a key resour­ce to advan­ce in the lear­ning pro­cess. Not so for AI: it is gram­ma­ti­cal­ly com­pe­tent wit­hout using a for­mal grammar becau­se its gram­ma­ti­cal „beha­vi­or“ emer­ges from trai­ning on vast amounts of texts. Modern “Lar­ge Lan­guage Models” (LLMs) thus do not usual­ly con­tain a trans­pa­rent, human-rea­da­ble grammar modu­le – they are able to impli­ci­t­ly app­ly gram­ma­ti­cal rules wit­hout sta­ting them.

When Euro­peans – sea­fa­rers, trad­ers and colo­nists – arri­ved in Sou­theast Asia and the Indo­ne­si­an archi­pe­la­go in the 16th cen­tu­ry, they nee­ded Malay pri­ma­ri­ly in two are­as – to inter­act with the courts in Suma­tra and the Malay pen­in­su­la, and in in trade and com­mu­ni­ca­ti­on across the Archi­pe­la­go whe­re it had been used as a lan­guage of com­mu­ni­ca­ti­ons for cen­tu­ries. To learn Malay, an oral approach ser­ved them per­fect­ly well for a long time – picking up this lan­guage in con­ver­sa­ti­on, using word­lists and so on – even­tual­ly spea­king in a cor­rect or under­stan­da­ble way.

The need for for­mal lan­guage ins­truc­tion – and for a for­mal grammar – emer­ged with colo­ni­al expan­si­on and mis­sio­na­ry acti­vi­ties. At around 1680, mem­bers of the Dutch cler­gy embark­ed on a trans­la­ti­on of the bible into Malay. Malay had to be taught to offi­ci­als. Yet it took ano­ther 50 years until the first grammar of Malay was prin­ted in Euro­pe with funds of the Dutch East Indies Com­pa­ny – in 1736 in Amster­dam.

This grammar had been pro­du­ced by a Swiss-Dutch pastor and Ori­en­ta­list, Georg Hein­rich Wernd­ly (1693−1744). As a pastor, he had lived bet­ween 1719–1726 in Padang, Makas­sar and Bata­via (today’s Jakar­ta) and sub­se­quent­ly given the task to com­ple­te the Bible trans­la­ti­on into Malay. His fami­ly roots are in Zurich from whe­re his father had moved to Lin­gen, Ger­ma­ny, working as a pastor.

For his grammar, Wernd­ly made sure that Malay text samples are pre­sen­ted both in Roman and Ara­bic script. At the time, Malay was writ­ten pri­ma­ri­ly in Ara­bic script, and know­ledge of that script was essen­ti­al. For that pur­po­se, the prin­ter used type­sets that were adapt­ed or new­ly crea­ted for this grammar. Werndly’s tex­tu­al sources – the exam­p­le sen­ten­ces – are most­ly taken from Malay lite­ra­tu­re and reli­gious books. He expo­ses us to the Malay of the Sul­tana­tes and courts much more than to that of the mar­ket.

I am inte­re­sted in the histo­ry of the Malay lan­guage becau­se of my work on Suma­tran „sin­gers of tales“. This week, I con­sul­ted a copy of Wernd­ly­’s grammar in the rea­ding room of the Ber­ne Uni­ver­si­ty Library…It was an impec­ca­ble copy, after 300 years of sto­rage (one out of two that I could loca­te in Switz­er­land). When I men­tio­ned this to the clerk on checking out, his remark was “Not sure if our digi­ta­li­zed books will last that long”.

Befo­re finis­hing this post, I put a que­sti­on to AI – „which grammars were used to train LLMs on Malay?“ The ans­wer was: „The­re is no publicly docu­men­ted list of tra­di­tio­nal Malay or Indo­ne­si­an grammars that were direct­ly inser­ted into, or syste­ma­ti­cal­ly used to train, the major gene­ral-pur­po­se AI models“. Howe­ver, the ans­wer to the prompt lists the lan­guage cor­pus data­ba­ses – one of them with 4 bil­li­on words – that were used.