/Anki-Romaji-Remover

Automatically turn Romaji into Hiragana in an Anki deck

Primary LanguageHTMLMIT LicenseMIT

Anki Romaji Remover

Automatically turn Romaji into Hiragana in an Anki deck.

This is for if you get bothered by Romaji’s inaccuracy, or in my opinion, its somewhat misleading format.

This tool takes the name of a deck and the name of the field with Romaji, and converts it into Hiragana.

Some notes:

  • It preserves most anything which isn’t romaji, so if you have e.g. “setsumei(suru)” it will convert it to “せつめい(する)”
  • ”-” will be removed, however. This is because the romkan converter can get confused by these
  • It doesn’t hurt to run the script again on a deck which has already been wholly or partially converted

Setup

Use

Previewing changes

It is recommended to first run a “soft edit” of the deck. This will output what the script will change each Note to, but will not go through with the edit. For my use case, the command looked like this:

python3 AnkiRomajiRemover.py "A Frequency of Japanese Words" "Romanization" --written-field-name "Lemma" --soft-edit

Where Lemma was the normal written form and Romanization was the field I wanted to replace romaji with kana.

This will output a list of all conversions, as well as any warnings or errors encountered. See “Error Handling” for more details on errors. Here is an example of a successful conversion:

yokan(suru)     -> よかん(する)         (hint '予感（する）')

This says that it will change the “Romanization” field of that note from “yokan(suru)” to “よかん(する)”. The hint provided by the written field name “Lemma” is provided for additional verification.

If there were warnings or errors, I can output only those, for ease of reading (i.e don’t output straightforward/successful conversions):

python3 AnkiRomajiRemover.py "A Frequency of Japanese Words" "Romanization" --written-field-name "Lemma" --soft-edit --only-warnings

Ideally, you should resolve all warnings and errors before running the script without --soft-edit.

Committing the changes

Make sure that your decks are Synced and backed up.

Once you have looked over the changes and decided that you want to make the conversion, run the script without --soft-edit. My use case looked like this:

python3 AnkiRomajiRemover.py "A Frequency of Japanese Words" "Romanization" --written-field-name "Lemma"

Create a backup of your decks before running this!

I am not responsible for damage to your decks. Use this script at your own risk.

Error Handling

Errors sometimes occur in the romaji input, and in conversion. For example:

  • Romaji may not be convertable correctly by romkan: ‘matchi(suru)’ outputs ‘まtち(する)’ (the actual field should be ‘マッチ（する）’, but the ‘t’ is especially problematic in the all-hiragana version)
  • From input “booringu” the converter outputs ボオリング instead of ボーリング

Resolving error with --written-field-name hint

Without a --written-field-name (aka “hint field”), there isn’t much I can do to fix errors. This is because the script doesn’t know what the actual word you are trying to convert the reading for is.

If there is a written field name, I use a few techniques to resolve these errors:

  • If the hint field does not have any kanji, just use the hint field. It’s already readable in its hiragana, katakana, or mixed hiragana-katakana (e.g. verb-ified loan word) representation. Both of the previously mentioned errors would be resolved this way
  • If the hint field contains kanji, use EDict to look up the hint’s reading
  • Finally, if there was anything strange about the error resolution, include the romaji in the field so that the user may decide how to resolve the error

Malformed notes

For my dataset, the script found notes with missing fields. It will error like so:

Error: Empty 'Romanization' found in the following note, which may be malformed:
{'noteId': 1534968932931, 'tags': [], 
'fields': {
'Rank': {'value': '3541', 'order': 0}, 
'Lemma': {'value': '親友shin’yuu', 'order': 1}, 
'Mnemonic Lemma/Kanji': {'value': '', 'order': 2}, 
'Romanization': {'value': '', 'order': 3}, 
'Mnemonic Pronounciation': {'value': '', 'order': 4}, 
'Part of Speech': {'value': 'n.', 'order': 5}, 
'English Gloss': {'value': 'best friend, close friend', 'order': 6}, 
'Illustrative Example': {'value': '二十年来の親友の結婚式に出席した。', 'order': 7}, 
'Illustrative Example Translation': {'value': 'I attended the wedding of my best friend of twenty years.', 'order': 8}, 
'Illustrative Example Pronounciation': {'value': '', 'order': 9}, 
'Illustrative Example 2': {'value': '', 'order': 10}, 
'Illustrative Example 2 Translation': {'value': '', 'order': 11}, 
'Illustrative Example 2 Pronounciation': {'value': '', 'order': 12}}, 
'modelName': 'A Frequency Dictionary of Japanese Words', 'cards': [1534968945014, 1534968945015]}

As you can see, it is a valid error: the Romanization field appears to have been merged with the Lemma field. I will need to fix that note by hand before conversion will work on it.

Fixing notes by hand

As an example, here is how I would fix the Empty 'Romanization' found error in the previous section:

  • Open Anki
  • Click Browse
  • Click the deck in the list on the left with the erroneous card
  • Search some text in the card to find it. In this case “best friend” will get me to the card
  • Look over the fields and change them to correct the error. In this case, I will cut “shin’yuu” from the Lemma field and paste it into the Romanization field