Trouble searching ... zu... in romaji. Broken rōmaji input in search box

[bengaskin]

To find 続く (つづく) I have to type in rōmaji as: tsudzuku or tsuduku or tuduku etc.
Trying "tsuzuku" doesn't work, I get the explanation "no results".

It's a bit counter-intuitive. I would expect tsuzuku to work, especially as this is how aedict itself displays this word in rōmaji(!).

Can you please check rōmaji input is working as expected?

[mvysny]

Unfortunately, both づ and ず share the same sound in Japanese : zu. Since the default romanization system in aedict (Hepburn, it's most widely used) is based on pronunciation rather than uniqueness, it captures both characters as zu. You can google for wiki hiragana table, you'll see that. The same goes for ji. You can also check the kana table built-in in aedict. You can see that to write づ in Hepburn, you need to use xzu (or du as you discovered - this comes from Nihon Shiki). You can also check the search query displayed - it will show the hiragana constructed from romaji and you can check easily whether the reconstruction was perfect or not. To remedy this, you have several options : 1. Switch to a different romanization system. Aedict support Nihon Shiki and kunrei Shiki. You can then consult the aedict kana table on how to write characters in romaji. For example づ is written as du. 2. Use Hepburn and tend to write xzu or du. The x letter is useful as it may be also used in other cases, for example xna will force んあ instead of な, similar goes to nya vs xnya. If you clear the recently viewed entries, aedict will display more hints regarding this. Best regards, Martin

…

On Dec 2, 2016 5:48 AM, "bengaskin" ***@***.***> wrote: To find 続く (つづく) I have to type in rōmaji as: tsudzuku or tsuduku or tuduku etc. Trying "tsuzuku" doesn't work, I get the explanation "no results". It's a bit counter-intuitive. I would expect tsuzuku to work, especially as this is how aedict itself displays this word in rōmaji(!). Can you please check rōmaji input is working as expected? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#714>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAPgiyeKwzg_A7qE-kLcvf5vHSjJda-Yks5rD5ULgaJpZM4LCK-V> .

[bengaskin]

Thanks for the explanation. Long rant below, TL;DR: Sincere thanks for the workaround, but the bug still stands. Sorry to belabor the point, but it still seems like it's technically possible to address. Is it not feasible to do a parallel back end search for all rōmaji instances of zu or ji to check for both ず and づ and for both じ and ぢ? If you wanted to change the menu options to allow for instance sloppy Hepburn input (maybe call this non-strict or "non-strict ji/zu") and strict Hepburn input (maybe call this "strict ji=じ,di=ぢ,zu=ず,du=づ”). I'm sure if you think about it there's many different ways you could name these switches so that they are reasonably self-documenting. Another option would be strict one-one input or allow ambiguous ji/zu input. I would argue that non-strict should be the default (for rōmaji input into the search box). I didn't know what nihon shiki or kun shiki were exactly, I had to look them up. At any rate they don't seem to do what I want, they don't allow tsuzuku from what I can tell. I guess what I'm looking for is a permissive version of Hepburn for _input_, I'm not as worried about output. It seems the way of the world these days that even Google will offer "did you mean..."-style advice for any minor typo. I note that I can already find 料理 in the app by searching for both ryori (this is common not-great roumaji) or ryouri (which I would consider better), so it seems you are already implementing some version of many to one searching. But despite all of the above, I'd also argue that if the app displays the roumaji of 続く as tsuzuku in Hepburn mode, to not accept tsuzuku as input in the search box is undeniably inconsistent and translates as broken for the user experience. Please can you at least for now take this on board as a legitimate bug. ...that if you accept o/ou: ryouri and ryori for りょうり, then you should also accept zu/du for づ. I'm not so worried about accepting shimbun for しんぶん. I guess the app doesn't do it, and could, but it's fine as is. I guess if the app displayed tsuduku or tsudzuku I would feel like it was a little un-user friendly and strict but at least I would understand that it doesn't take tsuzuku as input. But if it displays tsuzuku as output, it _needs_ to accept tsuzuku as input. Many Thanks Ben On Dec 2, 2016 2:25 PM, "Martin Vysny" <notifications@github.com> wrote: Unfortunately, both づ and ず share the same sound in Japanese : zu. Since the default romanization system in aedict (Hepburn, it's most widely used) is based on pronunciation rather than uniqueness, it captures both characters as zu. You can google for wiki hiragana table, you'll see that. The same goes for ji. You can also check the kana table built-in in aedict. You can see that to write づ in Hepburn, you need to use xzu (or du as you discovered - this comes from Nihon Shiki). You can also check the search query displayed - it will show the hiragana constructed from romaji and you can check easily whether the reconstruction was perfect or not. To remedy this, you have several options : 1. Switch to a different romanization system. Aedict support Nihon Shiki and kunrei Shiki. You can then consult the aedict kana table on how to write characters in romaji. For example づ is written as du. 2. Use Hepburn and tend to write xzu or du. The x letter is useful as it may be also used in other cases, for example xna will force んあ instead of な, similar goes to nya vs xnya. If you clear the recently viewed entries, aedict will display more hints regarding this. Best regards, Martin

On Dec 2, 2016 5:48 AM, "bengaskin" ***@***.***> wrote: To find 続く (つづく) I have to type in rōmaji as: tsudzuku or tsuduku or tuduku etc. Trying "tsuzuku" doesn't work, I get the explanation "no results". It's a bit counter-intuitive. I would expect tsuzuku to work, especially as this is how aedict itself displays this word in rōmaji(!). Can you please check rōmaji input is working as expected? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#714>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAPgiyeKwzg_A7qE-

kLcvf5vHSjJda-Yks5rD5ULgaJpZM4LCK-V>

.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#714 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACEj40CGz-4hIN26O7CHfYqBH21SQglnks5rD6vkgaJpZM4LCK-V> .

[mvysny]

Actually, it is technically possible, that is true. And you are right, Aedict actually auto-prolongs syllables in some cases: #308
I'll try to alter the search index so that the tsuzuku entry will be findable both using つづく and つずく (and thus tsuzuku).

[bengaskin]

Thanks! I was also planning on getting back to you about information relating to Heisig English names. An old logged issue. There's an authoritative electronic source. I'll reply back to the original thread so you've got the information but of course up to you whether it's worth looking into and if so please take your time. Personally I've been busy with work but also married very early this year and baby late this year 3 weeks ago. Even still, I might be able to help. Ben

…

On Dec 2, 2016 8:51 PM, "Martin Vysny" ***@***.***> wrote: Actually, it is technically possible, that is true. And you are right, Aedict actually auto-prolongs syllables in some cases: #308 <#308> I'll try to alter the search index so that the tsuzuku entry will be findable both using つづく and つずく (and thus tsuzuku). — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#714 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACEj4-HPs2WPHZNqlMuRcve3rzkhnKxIks5rEAY9gaJpZM4LCK-V> .

[mvysny]

I have one advice for you.
RUN.
Raising a baby is hell. Your life is going to be changed completely and irreversibly. But let's look on the bright side: maybe you're a baby person ;)

[mvysny]

Fixed in Aedict 3.39.28