Don't understand Morph Search results--Misunderstanding or Error / Bug?

Page 1 of 1 (6 items)
This post has 5 Replies | 1 Follower

Posts 73
Terry M. Moore | Forum Activity | Posted: Sat, Nov 10 2012 11:20 PM

I am using L4 Platinum with a number of additional resources running in 64-bit Windows 7. I am trying to do a Morph Search to find all verbs whose lexical form/lemma ends with μαι in the Lexham Greek-English Interlinear New Testament (LGNTI) and don't understand the results I'm getting.

My initial Morph search setup was Search All Morph Text in All Passages in LGNTI with Logos Greek Morphology with the search parameter lemma:*mai. This search returned 4,119 results in 3,114 verses. I then clicked on Analysis and 104 seconds later it displayed. As I looked through the results, I saw that the search also returned unwanted infinitives and participles.

I reran my search this time with the search parameter lemma:*mai@V??[^NP]. This search returned 5,347 results in 2,189 verses. The Analysis view showed that the unwanted infinitives and participles had been excluded from the search as I wanted. However, what I don't understand is how could excluding infinitives and participles cause the results to increase from 4,119 hits to 5,347?

Can anyone explain what I may be missing in interpreting the results? Or, can they confirm that there is an error/bug in this somewhere? Sad

Thanks,
--Terry

 

Posts 15805
Forum MVP
Keep Smiling 4 Jesus :) | Forum Activity | Replied: Sat, Nov 10 2012 11:47 PM

Terry M Moore:
However, what I don't understand is how could excluding infinitives and participles cause the results to increase from 4,119 hits to 5,347?

The results are combining the totals from lemma search and morph search.

Using NA28 in Logos 5, a Morph search for lemma:*μαι finds 4,119 results in 3,114 verses.  Adding Morph criteria to search lemma:*μαι@V??[^NP] found 5,218 results in 2,189 verses.  Looking at search results using Verses showed two lines for each verb; hence combined result is effectively 2,609 results in 2,189 verses.

Keep Smiling Smile

Posts 73
Terry M. Moore | Forum Activity | Replied: Sun, Nov 11 2012 1:37 AM

Thanks, Keep Smiling, but I don't get it. I don't have NA28 or L5 so I can't duplicate your searches, but I did look at mine some more.

The Verse view is as shown below. You can see it says 4,119 results in 3,114 verses. I don't understand what you said about Verses showing two lines for each verb. The results are the same if I change "All Morph Text" to either "Surface Text" or "Manuscript Form."

 

Then if I look at Analysis view I get the following:

Here you can see in the top line it has the same 4,119 results in 3,114 verses, but if you look at the detailed results there are 4,266 verb results (more than the top results number). If you add up the results for each mood they total 4,266. So how does one explain 4,119 vs. 4,226?

There is a similar discrepancy if @V??[^NP] is added to the search parameter. In the below Analysis view it shows 5,347 at the top and 2,746 total results in the detailed results (less than the top results number this time)--

What does make sense here is that the detailed results drop from 4,226 to 2,746 when infinitives and participles are omitted.,

What doesn't make sense is:

  1. That the number of results at the top in Analysis view increases when infinitives and participles are excluded.
  2. That the top and detailed results don't agree in Analysis view.
  3. That in order to get what appears to be the correct results, I have to go to Analysis view and look at the detailed results.
  4. That the results in Verse view don't appear to be connected to anything identifiable.

As an engineer by education and a programmer by vocation (though long retired), I still think there's a bug here.

If I'm still missing something, I appreciate your patience and would like to hear what you or others may have to contribute.Tongue Tied

Blessings,
--Terry

Posts 13417
Mark Barnes | Forum Activity | Replied: Sun, Nov 11 2012 2:17 AM

Terry,

If you look closely you'll see you're using transliterated text for the final search, and Greek text for the first two searches. You should always use Greek text for consistency, though Logos does guess what you mean correctly. 

Your results look completely wrong, but the problem is that Logos indicates two matches when you search for a lemma with a morphology (the lemma is one hit, and the morphology another). That's counter-intuitive, but I'm afraid that's the way it is.

Mark

 

Posts 73
Terry M. Moore | Forum Activity | Replied: Sun, Nov 11 2012 1:33 PM

Hi Mark, thanks for weighing in. You actually confirmed there is a bug here. Let me begin by saying that I like Logos and find it to be a  valuable tool. But with any project this big there are bound to be things that will come up and have to be fixed or clarified. So I am not being critical. With that in mind, as a former programmer, if my boss questioned results of code I'd written and I told him "I'm afraid that's the way it is" I would probably have been out of a job. It should be clear that anyone doing a search like I am is not wanting to double count hits (one in lemma and one in morphology). If it is really necessary for the programmer to produce a double count for some internal reason, that count just needs to be divided by 2 before being displayed to the user--a pretty simple solution. Take a look at the below example. I've made my search results smaller this time by restricting the search to just the imperative mood to keep the numbers manageable. I've also changed the search text to the SBLGNT to make things as simple as possible:

The top Analysis results is 502, which is 2x the results I actually wanted as shown by 251 results for imperative verbs (double count as you pointed out). So the correct 251 number is available. However, if one only looked at Verses they would only see 501 hits. If a user needed a count, how would they know they have to divide by 2 or go to Analysis? That is not user friendly, helpful, or intuitive.

When I looked at the above Analysis I was initially thrown off by the unwanted indicative and subjunctive moods appearing. But a little investigation showed that these were the cases where a verb could also be parsed in a way in addition to what I searched for. All the indicative and subjunctive results also appear under imperative. So this is helpful and appropriate in Analysis view. The counts for indicative and subjunctive are included in the verb total, but apparently not in the 2x Analysis results.

Mark, I know you're connected at Logos, so if you can pass this along to the development folks I would appreciate it, and I expect they and other users would too.

Thanks and God bless,
--Terry

P.S. I've been working with the NT resources based on the SBLGNT like the LEB. Since it is based on the same Greek text I would expect identical results from the same search in each of them, but they each produce different results. I'm going to take a closer look at the LEB results and if there appears to be a bug I will make a new posting.

Posts 73
Terry M. Moore | Forum Activity | Replied: Sun, Nov 11 2012 2:53 PM

I have done the followup I indicated I would be doing in my previous posting.

A Morph Search done in the SBLGNT produces the following results:

The discrepancy in the 502 results in Analysis results vs. the 251 for imperative verbs is discussed in my previous post and will hopefully be passed on to the folks in development as a bug.

When the same search is done in the Lexham English Bible (LEB) the following results are returned:

I find it interesting that the Analysis results are 251 as they should be per the previous search's results under imperative verbs. Evidently they were not double counted at the Analysis results level as discussed in previous posts. There are, however, a couple of possible problems here. Because SBLGNT and LEB utilize the same Greek text, one would expect both to return the same search results, but that is not true:

  1. The imperative total has changed from 251 to 253 with the LEB, despite the Analysis level total being 251. This change is due to present, either middle or passive, imperative increasing from 144 results to 146. 
  2. Adjective results are displayed. Though the results count does not appear to be added into any other count.

In both of the issues identified above the problem arises from the idiomatic expression in 1 Co 10:32, ἀπρόσκομι γίβεσθε, translated "Give no offense." In the adjective section of the results all three words in the translation appear:

Likewise, under verb, imperative, present, either middle or passive 1 Co 10 appears three times, once for each translated word:

This needs to be cleaned up somehow. There is only one verb in the idiom, so there should only be one occurrence for it in the verb list. This would correct the verb results count. Also the adjective section could just be suppressed. 

Comments are welcome. If the right person sees this posting and agrees something needs to be fixed, hopefully they can forward it on to development.

Thanks,
--Terry

Page 1 of 1 (6 items) | RSS