mlphon 3.1.3

Malayalam phonetic analyser

This is python interface for the Malayalam phonetic analyser - mlphon.

Installation

Requires Python 3.9–3.12. Install using pip or uv:

$ pip install mlphon

or with uv:

$ uv pip install mlphon

Syllablize a Malayalam Word

The following python snippet will split a word in Malayalam script into syllables.

from mlphon import PhoneticAnalyser
mlphon = PhoneticAnalyser()
mlphon.split_to_syllables('കേരളം')

It will give the result

[‘കേ’, ‘ര’, ‘ളം’]

Phonetically analyse a Malayalam Word

from mlphon import PhoneticAnalyser
mlphon = PhoneticAnalyser()
mlphon.analyse('കേരളം')

It gives the result as a sequence of ipa and associated phonetic tags.

[{‘phonemes’: [{‘ipa’: ‘k’, ‘tags’: [‘plosive’, ‘voiceless’, ‘unaspirated’, ‘velar’]}, {‘ipa’: ‘eː’, ‘tags’: [‘v_sign’]}]}, {‘phonemes’: [{‘ipa’: ‘ɾ’, ‘tags’: [‘flapped’, ‘alveolar’]}, {‘ipa’: ‘a’, ‘tags’: [‘inherentvowel’]}]}, {‘phonemes’: [{‘ipa’: ‘ɭ’, ‘tags’: [‘lateral’, ‘retroflex’]}, {‘ipa’: ‘a’, ‘tags’: [‘inherentvowel’]}, {‘ipa’: ‘m’, ‘tags’: [‘anuswara’]}]}]

Malayalam g2p : Grapheme to Phoneme conversion

from mlphon import PhoneticAnalyser
mlphon = PhoneticAnalyser()
mlphon.grapheme_to_phoneme('കാറ്റ്')

It gives the ipa sequence as output.

[‘kaːṯṯə’]

Malayalam p2g : Phoneme to Grapheme conversion

from mlphon import PhoneticAnalyser
mlphon = PhoneticAnalyser()
mlphon.phoneme_to_grapheme('paːlə')

It gives the corresponding grapheme sequences as output.

[പാല്’]

Command Line Interface for the above operations: mlphon

usage:

mlphon [-h] [-s] [-a] [-p] [-pe string] [-se string] [-g] [-i INFILE]
        [-o OUTFILE] [-v]

optional arguments:
-h, --help            show this help message and exit
-s, --syllablize      Syllablize the input Malayalam string
-a, --analyse         Phonetically analyse the input Malayalam string
-p, --tophoneme       Transcribe the input Malayalam grapheme to phoneme
                        sequence
-pe string, --phoneme_end string
                        String to be inserted at end of phoneme
-se string, --syllable_end string
                        String to be inserted at end of syllable
-g, --tographeme      Transcribe the input phoneme sequence to Malayalam
                        grapheme
-i INFILE, --input INFILE
                        source of analysis data
-o OUTFILE, --output OUTFILE
                        target of generated strings
-v, --verbose         print verbosely while processing

For example to perform g2p operation on a set of words stored in input.txt with one Malayalam word per line,

mlphon -p -pe " " -se "." -i path/to/inputfile.txt -o path/to/outputfile.txt

Inputfile contents:

cat path/to/inputfile.txt
അകത്തുള്ളത്
അകപ്പെട്ടത്
അകലെ

Outputfile contents:

അകത്തുള്ളത് a .k a .t̪ t̪ u .ɭ ɭ a .t̪ ə .
അകപ്പെട്ടത്        a .k a .p p e .ʈ ʈ a .t̪ ə .
അകലെ    a .k a .l e .

Application: Using mlphon to create a phonetic lexicon

A typical use case of phonetic analysis is to create a phonetic lexicon to be used in Automatic Speech Recognition or Text to Speech Synthesis. The phonetic representation with each phoneme separated by a space can be obtained as below:

from mlphon import PhoneticAnalyser, split_as_phonemes
mlphon = PhoneticAnalyser()
analysis = mlphon.analyse('എന്നാൽ')
for result in analysis:
  split_as_phonemes(result)

It results in the output, two different valid phoneme sequences:

‘e n̪ n̪ aː l’

‘e n n aː l’

The phonetic representation with each syllable separated by a space can be obtained as below:

from mlphon import PhoneticAnalyser, split_as_syllables
mlphon = PhoneticAnalyser()
analysis = mlphon.analyse('ഇന്ത്യയുടെ')
for result in analysis:
  split_as_syllables(result)

It results in the output:

‘i n̪t̪ja ju ʈe’

To get phonemes and syllables with user defined end-marker strings as below:

from mlphon import PhoneticAnalyser, phonemize
mlphon = PhoneticAnalyser()
analysis = mlphon.analyse('ഇന്ത്യയുടെ')
for result in analysis:
  phonemize(result, " ", ".")

It results in the output with a ‘space’ after every phoneme and a ‘period’ after every syllable

‘i .n̪ t̪ j a .j u .ʈ e .’