Syllabification of Devanagari

Question

I am trying to syllabify devanagari words

धर्मक्षेत्रे -> धर् मक् षेत् रे dharmakeshetre -> dhar mak shet re

wd.split('्')

I get the result as :

['धर', 'मक', 'षेत', 'रे']

Which is partially correct

I try another word कुरुक्षेत्र -> कु रुक् षेत् रे kurukshetre -> ku ruk she tre

['कुरुक', 'षेत', 'रे']

The result is obviously wrong.

How do I extract the syllables effectively?

Answer 1

If you look at your strings character by character

>>> data = "कुरुक्षेत्र"
>>> re.findall(".", data)
['क', 'ु', 'र', 'ु', 'क', '्', 'ष', 'े', 'त', '्', 'र']

And your other string

>>> data = "धर्मक्षेत्रे"
>>> re.findall(".", data)
['ध', 'र', '्', 'म', 'क', '्', 'ष', 'े', 'त', '्', 'र', 'े']

So what you want is probably split these using '् '्. Let's call them notation characters for now. If you print the ord(data[2])for the first notation character, it is 2381. Now if you probe around this value

>>> for i in range(2350, 2400):
...     print(i, chr(i))
...
2350 म
2351 य
2352 र
2353 ऱ
2354 ल
2355 ळ
2356 ऴ
2357 व
2358 श
2359 ष
2360 स
2361 ह
2362 ऺ
2363 ऻ
2364 ़
2365 ऽ
2366 ा
2367 ि
2368 ी
2369 ु
2370 ू
2371 ृ
2372 ॄ
2373 ॅ
2374 ॆ
2375 े
2376 ै
2377 ॉ
2378 ॊ
2379 ो
2380 ौ
2381 ्
2382 ॎ
2383 ॏ
2384 ॐ
2385 ॑
2386 ॒
2387 ॓
2388 ॔
2389 ॕ
2390 ॖ
2391 ॗ
2392 क़
2393 ख़
2394 ग़
2395 ज़
2396 ड़
2397 ढ़
2398 फ़
2399 य़

We are mostly interested in in values between 2362 and 2391. So we create a array of such values

>>> split = ""
>>> for i in range(2362, 2392):
...     split += chr(i)

Next we want to find all pattern with or without a corresponding notation symbol.

>>> re.findall(".[" + split + "]?", "धर्मक्षेत्रे")
['ध', 'र्', 'म', 'क्', 'षे', 'त्', 'रे']
>>> re.findall(".[" + split + "]?", "कुरुक्षेत्र")
['कु', 'रु', 'क्', 'षे', 'त्', 'र']

This should get you close to what you are probably looking for. If you need more complex handling then you will have to go with the link @OphirYoktan posted

Answer 2

Check out unicodedata module.

>>> import unicodedata
>>> word = 'कुरुक्षेत्र'

Names assigned to each character:

>>> for ch in word:
        print(unicodedata.name(ch))

    
DEVANAGARI LETTER KA
DEVANAGARI VOWEL SIGN U
DEVANAGARI LETTER RA
DEVANAGARI VOWEL SIGN U
DEVANAGARI LETTER KA
DEVANAGARI SIGN VIRAMA
DEVANAGARI LETTER SSA
DEVANAGARI VOWEL SIGN E
DEVANAGARI LETTER TA
DEVANAGARI SIGN VIRAMA
DEVANAGARI LETTER RA

General category assigned to each character:

>>> for ch in word:
        print(unicodedata.category(ch))

    
Lo
Mn
Lo
Mn
Lo
Mn
Lo
Mn
Lo
Mn
Lo

FileFormat.info has a list of Unicode character categories.

See if this is what you want to achieve:

import unicodedata

def split_clusters(txt):
    """ Generate grapheme clusters for the Devanagari text."""

    stop = '्'
    cluster = u''
    end = None

    for char in txt:
        category = unicodedata.category(char)
        if (category == 'Lo' and end == stop) or category[0] == 'M':
            cluster = cluster + char        
        else:
            if cluster:
                yield cluster
            cluster = char
        end = char

    if cluster:
        yield cluster

Testing the function:

>>> list(split_clusters('धर्मक्षेत्रे'))
['ध', 'र्म', 'क्षे', 'त्रे']
>>> list(split_clusters('कुरुक्षेत्र'))
['कु', 'रु', 'क्षे', 'त्र']