A few more tagging mistakes/ inconsistencies in Andersen Forbes

Since the developers are said to watch the forum I post a few more errors for their consideration:

Jer 33:5 את פגרי האדם should be tagged as "second obj Gram" cf. Gen 26:15, 21:19, 42:25. At the moment it is tagged as dir obj gram.

Josh 11:20 את ישראל should be tagged as את 2 and not as אות1 cf. Gen 14:2 if one is to make sense of the verse. If it is tagged as את 2 it should no longer be tagged as dir obj. (and indeed it is not a direct object!)

Ezek 20:23 בגוים and Ps 106:27 בגוים tagged as dir obj. should be "Loc Gram" cf. Lev 26:33

I think that the statistical rate of inconsistencies must be much higher than 0.1%. In this search alone I got 4 wrong results out of 62 hits - apx. 6 percent which is statistically significant.

Find more posts tagged with

Comments

Dean Forbes

Hello, David--

Many thanks for sending along the crop of errors that you detected. They have been corrected in the master files. They were, for the most part, places where the computer parser blindly produced multiple direct objects in embedded obliqueness structures. The parser was allowed to do this to capture suspended and resumed arguments.The downside of so doing was too many false positives like the ones that you detected.

If you have other errors in need of fixing, we'd be grateful to receive them.

============================================================

You asserted: "I think that the statistical rate of inconsistencies must be much
higher than 0.1%. In this search alone I got 4 wrong results out of 62
hits - apx. 6 percent which is statistically significant."

Your assertion is based on 62 hits (in a highly biased and truly minuscule sample). My assertion is based on 40-hours of careful checking of "text surrogates" involving 27,014 opportunities for error. (See the paper that I cited earlier in the forum.) I must confess that I grant greater confidence to my assertion than to yours in this matter...

But, many thanks for your inputs!

Dean

David Knoll

https://community.logos.com/discussion/comment/247714#Comment_247714

See the paper that I cited earlier in the forum

In my defence, this book is not available in any academic library in Israel (which is quite rare). I asked for it to be ordered to Tel Aviv University and as soon as the Independence day holiday is over I'll do the same thing for the Hebrew University but it takes about 6 months for a book to arrive.

Your assertion is based on 62 hits (in a highly biased and truly minuscule sample). My assertion is based on 40-hours of careful checking of "text surrogates" involving 27,014 opportunities for error. (See the paper that I cited earlier in the forum.) I must confess that I grant greater confidence to my assertion than to yours in this matter...

I think that there are two different things to measure. One thing is how consistent the database is in order to get the general picture for papers like your Hebrew spelling book (A great book BTW). But the other thing to measure is how accurate specific searches are. Most people who use bible software packages are interested in specific questions and not the general trend in the Hebrew Bible. I speak from my own experience here when I say that you need to work on that.

If I look for the question how many times the word X is the direct object of the verb y and I get a 5% error rate then the overall 0.1% is of no consequence to me. But perhaps your database was not intended for such searches and the problem lies in my expectations.