how to search hebrew accent on Logos

Vincent Setterholm said:

Accent sensitive searching is slated for version 5.2. I've gotten to play around with a working demo of it. There is one way to do it now: our regular expression (a standard pattern matching syntax) searches are sensitive to all the marks. E.g.: /.*\u05A6.*/

Where the // tell the search engine you want to use regular expression pattern matching, .* means 0 or more characters coming before or after the accent and \u05A6 is the Unicode quad for a particular accent (http://unicode.org/charts/PDF/U0590.pdf). You can just type a Hebrew accent in that location as well, but as you get into writing regular expressions with parens and brackets, things can display a little funny with the collision of right to left and left to right characters, so I often just write in the code point quads.

Regular Expression searches don't get normalized, so if you're doing something more complicated than just searching for one character, the order you put the marks in matters. Right now, our normalization for Hebrew puts the marks in Unicode order, so accents come before vowels which come before sin/shin dots and dageshes. This is backwards from both the intuitive order and the order we actually encode the Hebrew in (for good font display, etc.), so hopefully we will re-implement our normalization routine so that this is more intuitive. (We had our own normalization routine for Hebrew in previous versions, but there was a regression here.)

So regular expressions can find a lot of things if one is willing to dig into how to use them, but there is easier accent-sensitive searching coming.

??  How long have regular expression searches been in Logos 5?  I was not aware that they were there.  (It doesn't seem to be a full implementation.  Some patterns don't seem to work right)

Vincent Setterholm said:

I'm told they were in Logos 4... They're an 'unsupported feature', so there's not much in the way of documentation

I think they were already usable in L3.

Todd Phillips said:

Vincent Setterholm said:

I'm told they were in Logos 4... They're an 'unsupported feature', so there's not much in the way of documentation. I haven't given it much of a stress test (I only recently learned of them myself - I know I tried them before, but I probably assumed they just weren't in there because the normalization changed so my old searches just seemed to not work), but I'm guessing that the same limitations from Libronix DLS 3 are in play, namely that the searches are bound to a single word (as if you had \b on both ends of your query) and I'd be surprised if they could be used to find punctuation characters. But beyond that, I'd be curious to know what patterns you are looking for that aren't working (and whether or not they're really normalization issues, with regards to Hebrew).

I searched for both /s*word/ and /s{0,1}word/ in the ESV (either of which should find all verses with "sword" and/or "word"), but both patterns only found the "sword" verses.

On the other hand, /(sw|w)ord/ worked and found both "sword" and word".

Try "*word" (without quotes).  I'll warn you:  Limit your search to one bible.  I made the mistake of searching all passages in all bibles.

Vincent Setterholm said:

Interesting. Using character class brackets also works: /*word/, but this is a bug. Now you know why it's unsupported!

EDIT: Not sure how to stop the emoticon, but that was supposed to be [ s ] (no spaces).

The problem with that (if you wish to receive hits on either "word" or "sword" is that, if you specify an "s" at the beginning, it will not find the word without the "s."  The same happens if you use "?word" since it REQUIRES a letter for the "?" whereas it may be absent with the use of the asterisk.

I don't know. Bradley pointed out that regular expressions were undocumented, untested, unsupported, etc. etc., so it doesn't shock me that some change made in development could have altered this behavior. Why would they test their new code against a 'feature' that is undocumented, untested, unsupported, etc.?

Obviously, I think Regular Expressions are wicked cool, so I hope we do a real job of supporting them someday - it's probably the biggest thing I miss from Libronix 3. There's a Uservoice suggestion you can cast votes on to indicate your interest in this feature: http://logos.uservoice.com/forums/42823-logos-bible-software-5/suggestions/508925-full-regular-expression-searches.

5.2 did introduce some handy new search syntax for matching vowels and accents, which you can read about here: http://wiki.logos.com/Search_Matching_Commands

However, as far as I know, these match commands don't yet work with wild-card searches (nor the unsupported regular expressions). So while this will let you find a particular word with a particular accent, I don't think it'll let you just find the accent in isolation.

Actually, I'm saying that if you want to do regular expressions on the Hebrew text, you should keep a copy of Libronix 3 installed on your machine in addition to Logos 5. Libronix 3 had much better Regular Expression support than Logos 4/5 ever did. And tell all your friends about that Uservoice link.