Friday, July 18, 2008

Quick comment on accuracy

Harry Neumann from Deep Bible Study has another website my bible version with a neat selection tool.  He has some accuracy numbers I don't fully agree with, but are reasonable and independent of my opinions.  By adding his "form" and "function" numbers you get something close to what I've been calling "accuracy".  So here are the stats from his site.

15: NCV, TEV, NEB
14: CEV, NET, GW, NJPS, CJB
13: NAB, NIV, REB, NLT, NIrV
12: NRSV, NKJV, NJB,  
11: KJV, HCSB, TNIV, MSG, RSV, YLT,  TS
10: NASB, ESV, NLV
8: Phillips, TLB

I'm going to rare them myself just so you see that the accuracy issues I brought up in Is the ESV "essentially literal"? hold here as well.   

Great
Bolded in ESV article
Very good 
OK
Poor
Unrated

You'll notice how closely we correspond which proves my point about literalness killing accuracy.  He doesn't have a distinction between great and good.  Also we don't seem to agree at all on paraphrases, but the article didn't really address what quality in a paraphrase meant.  So now if I exclude the paraphrases and the unrated, and remove the good vs. great the chart looks like:

15: NCV, TEV, NEB
14: CEV, NETNJPS,
13: NAB, NIV, REB, NLT
12: NRSV, NKJV, NJB,
11: KJV, HCSB, TNIVRSV, YLT
10: NASB, ESV

Which is really close.  So close in fact that when I compute a statistical r it it .79.  Which means for moderate translations (classes 3-7 from the ESV article) choosing a balanced philosophy explains 79% of the accuracy using Mr. Neumann's statistics.  I couldn't ask for better evidence that attempting to be literal ends up damaging accuracy, just as I had asserted.  

3 comments:

Henry Neufeld said...

It's NJPS. I have an error in the way I entered that in the database which I will get around to fixing soon along with a few others. I have used, but have never reviewed or scored the JPS.

Henry Neufeld said...

Oh, and thanks for the link!

CD-Host said...

OK I'll fix the original.

______

To everyone reading Henry's first comment is referring to a line is answering a question that is no longer in the article.