Errors in transliterations about kawazu HOT 4 OPEN

thanks Cutano. I am running people and company names though it so it is hitting other ones as well that are not fixed but that fix did work for a lot of the names in the batch i was testing on.

below is a few more that a failing. It is not all of them but i am hoping that it is enough to give a good idea of the issue. The first fix seem to reduce the errors by over 50%
日本コンクリート工業（株）
東京コスモス電機（株）
日本アジア投資（株）
日精エー・エス・ビー機械（株）
（株）三ッ星
三ッ矢産業
三ツ川工業所
三ツ川浩一
白神自然学校一ッ森校

from kawazu.

i think i located the issue. hopefully this makes sense since i can't write or read japanese myself. The issue was that the Utilities.GetTextType method returns the wrong response cases where a Kana character is actually concidered a Kanji character like in the name "袖ケ浦港運". In that usecase the "ケ" should have been treated as Kanji and Utilities.GetTextType should have returned PureKanji but instead returns KanjiKanaMixed that then breaks the conversion. If you force it to PureKanji the conversion seems to work.

from kawazu.

i think i located the issue. hopefully this makes sense since i can't write or read japanese myself. The issue was that the Utilities.GetTextType method returns the wrong response cases where a Kana character is actually concidered a Kanji character like in the name "袖ケ浦港運". In that usecase the "ケ" should have been treated as Kanji and Utilities.GetTextType should have returned PureKanji but instead returns KanjiKanaMixed that then breaks the conversion. If you force it to PureKanji the conversion seems to work.

Yes, I checked the call stack and found the problem was caused by the ambiguity of kana "ケ". In common circumstances, it pronounced as "ke", but in the example that you offered, it is "ge", which directly caused the mismatch of the method IndexOf().
I'm working on this problem currently but don't know how to solve this in a decent way yet.

from kawazu.

I updated the nuget package and solved the problem temporarily by filtering the ケ relating results, but there could be other problems. Right now, it is just a temporary solution.

from kawazu.