andreihar / taibun Goto Github PK

Describe the feature
Current conversion to Zhuyin requires a 800 word long csv file with all possible syllables. It is preferred to have a conversion table to be much smaller, enabling to move the table to a dictionary within the code.

Expected behavior
Converter with Zhuyin system produces the same results as currently, but the chart within zhuyin.json file is decreased in size and moved moved to the in-code Python dictionary.

Relevant methods

• __tailo_to_zhuyin

Describe the bug
The converter throws an error when converting words with "m" as the nucleus (e.g. 毋). The transliteration of initial "n-" in incorrect (e.g. 耐 as nâi instead of lnâi). The conversion of final "-t", "-k", "-p" is incorrect (converts to "-d", "-g", "-b").

To Reproduce
Steps to reproduce the behavior:

1. Initialise Converter with 'Pingyim' as the system
2. Prompt '毋'
3. See error
4. Prompt '耐'
5. See incorrect transliteration of initial 'n-'
6. Prompt '法', '色', '給'
7. See incorrect transliteration of final '-t', '-k', '-p'

Expected behavior
Initial "n-" should convert to "ln-", final "-t", "-k", "-p" should convert to same values instead of "-d", "-g", "-b". Conversion of characters with "m" as the nucleus shouldn't throw an error.

Describe the feature
The current dataset doesn't include ~10k words that were not processed from the Taiwanese-Chinese Online Dictionary and doesn't include a big chunk of the iTaigi Chinese-Taiwanese Comparison Dictionary. Additionally, many proper nouns in the dataset can be derived using other entries and can yield unsatisfactory tokenisation results (e.g. 中國文化大學 should be tokenised as 中國 文化 大學, not 中國文化大學).

Expected behavior
More words are added and unnecessary proper nouns whose romanisation can be derived using other entries in the data are removed.

Relevant files
*data/words.json

Describe the feature
Add International Phonetic Alphabet as a transliteration system to the library.

Expected behavior
A function that can transliterate from Tai-lo to IPA is added.

Additional context
International Phonetic Alphabet
Taiwanese Phonology

Describe the feature
Currently, the sandhi flag applies sandhi rules locally within a word, i.e. changes tone within the word of every single syllable except for the first one. This is different from real sandhi rules, where changes are applied to every single syllable of the sentence, not just single words. Sandhi rules of the library should be modified to reflect the actual sandhi rules of the Taiwanese language.

Expected behavior
Properly applies sandhi rules to words depending on their position in the sentence.

Relevant methods

• __tone_sandhi

Additional context
Taiwanese Sandhi rules

Describe the feature
Current test cases are sufficient enough to test for the correctness of Tailo and Zhuyin systems but lack in testing other systems. Similar test cases should be added for other transliteration systems.

Expected behavior
Test cases are added for POJ, TLPA, Pingyim, and Tongiong transliteration systems.

Relevant files

• tests

Describe the feature
Currently, it is possible to convert to different transliteration systems only by converting from Chinese characters. Should add the possbility from converting from Tailo to other transliteration systems, bypassing Chinese characters.

Expected behavior
There exists a method/class that will convert from Tailo to other transliteration systems

Is your feature request related to a problem? Please describe.
The code is cluttered and can be organised better