Malayalam
Malayalam before and after the script split it in two
A 1971 reform simplified Malayalam print by breaking up hundreds of the old script’s distinct conjunct ligatures into their separate letters, so a book set before the reform and a book set after it can look like two different scripts for the same language. Evidano reads both, along with the incised palm-leaf manuscripts the printed forms grew out of, setting the transcription beside the image so every old-style ligature and every rounded modern letter can be checked.
Sample pages
Real pages from public collections, shown beside their transcriptions. Pages whose transcription is still being checked are marked.
Source · CC BY-SA 4.0 · Rajeshodayanchal
Transcription being verified
An undated, unidentified personal or family manuscript; no published edition or catalogue entry has been located to check against.
What the model is told to watch for: Produce a diplomatic transcription in Malayalam.
Transcription being verified
The Internet Archive’s OCR of this page is unreliable for the Malayalam entries; no independently proofread transcription of this dictionary has been located.
What the model is told to watch for: Produce a diplomatic transcription in Malayalam.
Source · CC BY-SA 4.0 · Sathish Kalathil
Transcription being verified
Prathibhavam is a defunct, self-published monthly with no digital archive located; no separate transcription exists for this issue.
What the model is told to watch for: Produce a diplomatic transcription in Malayalam.
Source · CC BY-SA 3.0 · Arunkumar P. R.
Transcription being verified
A single municipal notice board; no separate published transcription exists for it.
What the model is told to watch for: Produce a diplomatic transcription in Malayalam.
Source · Public domain · Scanned from the Malayalam Wikisource copy of the 1936 edition
Transcription
The whole of printed page 45, as proofread and then validated on Malayalam Wikisource (quality level 4 of 4) against this same scan. The text keeps the book’s old orthography and its Malayalam numerals; wiki markup is removed and paragraph breaks, not the printed line breaks, are what the Wikisource text records.
നാവായിക്കുളം, ചാത്തന്നൂർ ഇവയെക്കടന്നു കൊല്ലത്തു എത്തുന്നു. ഇതു് അവിടെനിന്നും വടക്കോട്ടുപോയി നീണ്ടകരപ്പാലം കടന്നു കൃഷ്ണപുരം, കായംകുളം, ഹരിപ്പാടു്, പുറക്കാടു, അമ്പലപ്പുഴ, ഇവയിൽ കൂടി ആലപ്പുഴ ചെല്ലുന്നു. ഈ പാതയെ ചേർത്തലവഴി വടക്കോട്ടു അതിർത്തിവരെ നീട്ടിയിട്ടുമുണ്ടു്. തിരുവനന്തപുരത്തുനിന്നു കൊല്ലത്തേയ്ക്കു ദൂരം ൪൫-മൈലും അവിടുന്നു് ആലപ്പുഴയ്ക്കു ൫൩-മൈലും, അവിടുന്നു അതിർത്തിയ്ക്കു ൨൮ മൈലുമാണു്. മറ്റു ചെറുതരം റോഡുകളിൽ പ്രാധാന്യം കൂടിയവയെ താഴെ വിവരിക്കുന്നു. (എ) നാഗർകോവിൽനിന്നും ഇരണിയൽവഴി ഒരു പിരിവു കുളച്ചലിലേയ്ക്കു്. നീളം ൧൪-മൈൽ. (ബി) തൊടുവെട്ടിയിൽ (കുഴിത്തുറയ്ക്കു് ഒരു മൈൽ കിഴക്കു) നിന്നു തിരുവട്ടാർ കുലശേഖരം ഇവയെ കടന്നു പേച്ചിപ്പാറയ്ക്കു്. ദൂരം ൧൪-മൈൽ. (സി) കൊല്ലത്തുനിന്നും കിഴക്കോട്ടു് കിളികൊല്ലൂർ, കുണ്ടറ, കൊട്ടാരക്കര, പുനലൂർ ഇവയിൽകൂടി തെന്മലയ്ക്കു സമീപത്തുവച്ചു തിരുവനന്തപുരം ചെങ്കോട്ട റോഡിൽ ചേരുന്നു. ദൂരം ൪൧-മൈൽ. ചെങ്കോട്ടവരെ ൫൯-മൈൽ. (ഡി) കായംകുളത്തുനിന്നും തെക്കുകിഴക്കായി പള്ളിക്കൽ, അടൂർ, പത്തനാപുരം ഇവയെക്കടന്നു പുനലൂരേയ്ക്കു് ദൂരം ൩൫-മൈൽ. (ഇ) കോട്ടയത്തുനിന്നും കിഴക്കോട്ടുള്ള റോഡ്:-വാഴൂർ, കാഞ്ഞിരപ്പള്ളി, മുണ്ടക്കയം ഇവിടങ്ങളിൽകൂടി പീരുമേട്ടിൽ എത്തുന്നു. അവിടെനിന്നും കിഴക്കോട്ടുപോയി പെരിയാർ കടന്നു ഗൂഡലൂർതാവളം വഴി കുമിളിയിൽകൂടി അതിർത്തി കടന്നു് അമിയനായ്ക്കനൂർ റെയിൽവേ സ്റ്റേഷനിലേയ്ക്കു പോകുന്നു. അമിയനായ്ക്കനൂർ വരെ ആകെ ദൂരം ൧൪൧ മൈൽ. അതിർത്തിവരെ ൯൦ മൈൽ. (എഫ്) എറ്റുമാനൂർനിന്നു കിഴക്കോട്ടുപോയി പാലായിൽ എത്തി അവിടന്നു വടക്കോട്ടുചെന്നു തൊടുപുഴവഴി മൂവ്വാറ്റുപുഴ വച്ചു് മെയിൻറോഡിൽ ചേരുന്നു. ദൂരം പാലായ്ക്കു ൧൦ മൈലും തൊടുപുഴയ്ക്കു ൨൮ മൈലും, മുവ്വാറ്റുപുഴയ്ക്കു ൪൧ മൈലുമാണു്. (ജി) എം.സി. റോഡിൽ ഏറ്റുമാനൂർനിന്നു വടക്കുപടിഞ്ഞാറായി ഒരു റോഡു കടുത്തുരുത്തി, വടയാറു്, വൈയ്ക്കം ഇവ കടന്നു കൊച്ചിക്കു പോകുന്നു. വൈയ്ക്കത്തിനു വടക്കു ഇത്തിപ്പുഴപ്പാലവും അതുകഴിഞ്ഞു മുറിഞ്ഞപുഴ പൂത്തോട്ടാൽ എന്നു രണ്ടു കടത്തുകളും ഉണ്ടു്. ദൂരം വൈയ്ക്കത്തേയ്ക്കു ൧൮ മൈലും അതിർത്തിക്കു ൩൦ മൈലുമാകുന്നു.
Transcription from താൾ:Geography textbook 4th std tranvancore 1936.djvu/52, Malayalam Wikisource (CC BY-SA 4.0).
What Malayalam text looks like on the page
Malayalam’s letters are markedly round even by the standards of neighbouring South Indian scripts, a shape usually traced to centuries of writing on palm leaf: a stylus called a narayam incises the leaf rather than inking it, and a straight or angular stroke would tear along the grain, so the script settled into loops and curves that a metal point can trace safely. The finished leaf, or ola, had the incised grooves rubbed with soot or a similar pigment to make the scratched letters legible, and leaves were bound through one or two holes into a taliyola book. This tradition, and the palm-leaf format itself, carried literary, medical, astrological and land-record texts in Kerala for centuries, well into the age of print.
Print inherited a script that had grown, letter by letter, a separate compound glyph for a very large number of consonant clusters, so that a complete traditional Malayalam typeface needed several hundred distinct pieces of type rather than a small alphabet combined systematically. Vowel signs, too, could attach to a consonant in shapes specific to that letter rather than a single consistent form. In 1971 the Government of Kerala adopted a simplified script that broke most of these ligatures apart into their component consonants and regularised the vowel signs, cutting the type inventory dramatically and making typewriters and, later, computer fonts practical. A book, deed or inscription set before 1971 — and some material set afterwards by publishers who kept the older convention on principle — uses letterforms a font built only for the reformed script does not contain at all.
Modern printed Malayalam, in newspapers, signs, notices and forms across Kerala, uses the reformed, disconnected letterforms almost universally today, still built from the same round bowls and loops but composed far more predictably from a small set of consonants, vowel signs and a plain virama.
Why ordinary OCR struggles here
Two orthographies, one language
A model trained mainly on modern reformed print has effectively never seen most of the old script’s conjunct ligatures, since the reform removed them from ordinary use rather than merely restyling them. A pre-1971 book, or a document a traditional publisher still sets in the old way, needs the ligatures recognised as themselves, not decomposed into letters that were never printed that way.
Ligatures that do not build from their parts
Where the reformed script constructs a cluster from recognisable consonants, the old script often gives the same cluster its own unrelated compound shape. Recognising one of several hundred such ligatures depends on having seen that specific shape before, not on reading the consonants that happen to make it up.
Vowel signs that vary by the letter they attach to
In pre-reform Malayalam, a vowel sign is not always one consistent mark: its shape can depend on which consonant it is attached to. A model expecting the reformed script’s regular, consonant-independent vowel signs misreads these older, letter-specific forms.
Incised, rubbed and aged palm leaf
A taliyola leaf carries no ink: contrast comes only from pigment rubbed into a scratched groove, which fades, and from a leaf surface that darkens, curls and cracks over time. Round letters already prone to running together in a fast hand become harder still to separate once the leaf itself has aged.
Who works with this material
Palm-leaf and manuscript archives
Kerala’s libraries, temples and private households hold large numbers of taliyola manuscripts covering literature, medicine, astrology and land records. A transcription that a specialist can check against the leaf turns an unread bundle into searchable text.
Legal, land-record and temple-trust archives
Deeds, grants and registers drawn up before 1971 use the old script’s ligatures throughout, and property or heritage disputes often turn on the exact wording of one such document. A transcription that keeps the old orthography intact, rather than silently modernising it, is the version worth checking against the original.
Publishers and type designers working with traditional orthography
Some religious and classical publishers still set new editions in the old script by choice, and a designer restoring or comparing historical typefaces benefits from a transcription that identifies each ligature rather than dissolving it into letters the old font never used.
Newspaper and periodical digitisation
An archive covering the years around 1971 can hold issues in both orthographies within the same run. A transcription pipeline that recognises which script a given issue uses keeps the whole archive searchable without one convention overwriting the other.
Getting the best transcription
Say which orthography the page uses
Tell the model whether the material is pre-1971 old-script Malayalam or the modern reformed script, since the two use different ligature and vowel-sign inventories that should not be forced into each other.
Decide how to render old-script ligatures
Choose whether a ligature that has no equivalent in modern print should be kept as a labelled compound form or expanded into the consonants it represents, and apply the same choice consistently across a document.
Name the material for palm-leaf pages
Mention that the source is incised palm leaf rather than ink on paper, so faint or rubbed-out letters are flagged as uncertain instead of being filled in from a guess about what a printed page would say.
Start with the most heavily ligatured page
In a batch that spans the 1971 divide, test the oldest or most conservatively typeset page first. If its ligatures are recognised correctly there, the more modern pages in the same batch need only a light check.
Frequently asked questions
- Can it read pre-1971 Malayalam, not just the modern reformed script?
- Yes. Tell it the page uses the old orthography and it reads the traditional conjunct ligatures and letter-specific vowel signs directly rather than expecting the simplified forms the 1971 reform introduced.
- What happens with an old-script ligature that no modern font contains?
- It is recognised as the compound letter it is rather than forced into a modern equivalent. You can ask for it kept as a labelled ligature or expanded into its component consonants, whichever suits your project.
- Can it transcribe an incised palm-leaf manuscript?
- Yes, from a clear photograph of the leaf. Mentioning that the source is palm leaf rather than paper helps it treat faint, rubbed or cracked passages as uncertain rather than guessing at them from context.
- Will a document that mixes old and new orthography be read consistently?
- Yes, provided you flag where the convention changes, since a reprint or an archive spanning 1971 can genuinely switch script partway through and each part should be read in its own terms.
- Does it need to know the material — palm leaf, paper or a printed sign — beforehand?
- It helps. Palm leaf, aged paper and a modern printed sign fail in different ways — faded incision, foxing, glare — and naming the material lets the model apply the right expectations for what damage looks like.
Related scripts and pages
Transcribe your Malayalam material
Upload a palm-leaf photograph, an old-script book, or a modern sign and check the Malayalam transcription against the image before you export it.