Hello, all! Please excuse the late response. In our latest exchange, on 12 Sept. 2020 at 9:01 AM, Nicolas Legrand < nicolas.legrand@bulac.fr> wrote: A good day, אהלן, こんにちは,
Ho, mine also :).
8-) Been there.
[skip, skip, skip]
Tuning Zebra to search CKJ or languages written with arabic script was a real pain with Zebra, it is a no brainer with Elastic Search. With the ICU module enabled, it works very well for CKJ and handles glyphe similarities. For example, the Library of Congress catalogues Farsi with alef maksura U+0649 instead of yeh U+06CC. We imported the farsi records from the Library of Congress and we were unable to find the documents searching with a farsi keyboard yielding the letter yeh. You can parameter Zebra to handle this and say U+0649 = U+06CC. With Elastic Search and ICU, you don't have to, it just works. We lost some day the possibilities to search CKJ with Zebra and didn't understand how to get it back. Zebra is certainly a very good search engine. But it's weird and hard to tune. We don't even have to ask ourselves how to tweak Elastic Search to do it. It just works.
Note that for Chinese, enabling QueryAutoTruncate with Elastic Search may lead to weird results when you type a full chinese name or title. As of 18.05 this is the case, I didn't check yet if this improved since then. We enabled it only when “*” is added at the end of a word
Ugh! Thank you for your account. It will help us as we further implement Kona on our site. We'll be following one of the first rules of software updating: Update or upgrade on a copy.
Best regards, יאללה ביי, それでは、また,
You, too. -- 気を付けて。 /ki wo tukete/ = Take care. -- Charles. Charles Kelley, MLS PSC 704 Box 1029 APO AP 96338 Charles Kelley Tsukimino 1-Chome 5-2 Tsukimino Gaadenia #210 Yamato-shi, Kanagawa-ken 〒242-0002 JAPAN +1-301-741-7122 [US cell] +81-80-4356-2178 [JPN cell] mnogojazyk@aol.com [h] cmkelleymls@gmail.com [p] linkedin.com/in/cmkelleymls <http://www.linkedin.com/in/cmkelleymls> Meeting Your Information Needs. Virtually.