BUG? - Odd Search Results
Can someone help me figure out what I'm doing wrong here. I am doing searches in 1 Peter and my results count is off at times. One example is searching lemma.g:ἀπειλέω - the results say 3 results in 2 verse. But the word only appears twice (Acts 4:17; 1 Peter 2:23). It is counting the occurrence in 2:23 as twice even though the word only appears once.
The same problem arises is searching root.g:ποιεω in 1 Peter. It counts the use of the word in 3:6; 3:17 and 4:19 as 2 occurrences each. Even though the word only appears once in each of those verses.
Ideas of why?
2015 13" MacBook Pro - 2 Ghz Intel i7 - 16 GB RAM - 500GB SSD - 2018 iMac Pro - 3.2GHz 8-core Xeon - both systems running OS 10.14.3 (Mojave)
Comments
-
A screen shot would help alot. You haven't said which Bible you are searching or what options you have set. I get correct results on the NRSV on both cases.
Orthodox Bishop Alfeyev: "To be a theologian means to have experience of a personal encounter with God through prayer and worship."; Orthodox proverb: "We know where the Church is, we do not know where it is not."
0 -
2015 13" MacBook Pro - 2 Ghz Intel i7 - 16 GB RAM - 500GB SSD - 2018 iMac Pro - 3.2GHz 8-core Xeon - both systems running OS 10.14.3 (Mojave)
0 -
2015 13" MacBook Pro - 2 Ghz Intel i7 - 16 GB RAM - 500GB SSD - 2018 iMac Pro - 3.2GHz 8-core Xeon - both systems running OS 10.14.3 (Mojave)
0 -
The "no" splits "uttered threats" into 2 parts - note the left pointing arrow under threats leading to 2 hits for the search engine. I suspect the others are similar cases.
Orthodox Bishop Alfeyev: "To be a theologian means to have experience of a personal encounter with God through prayer and worship."; Orthodox proverb: "We know where the Church is, we do not know where it is not."
0 -
Yes - it is these "splits" that are being counted twice - but previous versions of Logos never did it this way. If you do a cConcordance report and look at the word the counts are different as well. If we are searching for the number of occurrences of a particular Greek word, the fact the the English text "splits" the translation words does not qualify as two occurrences of the Greek wot=rd being searched for?
Any idea when this change was made? I have done these searches for years and only recently have found these results. With these results one would need to verify every occurrence found in the results to get an accurate count.2015 13" MacBook Pro - 2 Ghz Intel i7 - 16 GB RAM - 500GB SSD - 2018 iMac Pro - 3.2GHz 8-core Xeon - both systems running OS 10.14.3 (Mojave)
0 -
Yes - it is these "splits" that are being counted twice - but previous versions of Logos never did it this way.
I thought I recognized the problem because it had always been done this way. We'll have to wait for Dave, Mark, or Bradley to chime in with their memory.
Orthodox Bishop Alfeyev: "To be a theologian means to have experience of a personal encounter with God through prayer and worship."; Orthodox proverb: "We know where the Church is, we do not know where it is not."
0 -
Thanks MJ - I always appreciate your willingness to help! You are a rockstar!
2015 13" MacBook Pro - 2 Ghz Intel i7 - 16 GB RAM - 500GB SSD - 2018 iMac Pro - 3.2GHz 8-core Xeon - both systems running OS 10.14.3 (Mojave)
0 -
As far as I can remember this has always worked this way - one of the reasons for recommendations, at various points, to use original language texts for this type of searching. And using Multiple Books display (or the equivalent in earlier versions) to see matching results in English translations
0 -
I'll have to politely disagree. I have used and taught this process in my classes for a very long time now. I have literally hundreds of searches like this saved in Favorites that list the word and the number of times it appears in a particular book. Those same searches are now giving me different results when these "split" translations occur. For years, when you compare the number of occurrences between the Concordance tool and these Morph searches - the numbers have matched. That is no longer the case and this seems to be a relatively recent change.
2015 13" MacBook Pro - 2 Ghz Intel i7 - 16 GB RAM - 500GB SSD - 2018 iMac Pro - 3.2GHz 8-core Xeon - both systems running OS 10.14.3 (Mojave)
0 -
We made some changes back in 10.2 so that when searching a translated resource like this, disjoint translation text like this will now get counted and displayed as two separate results. Prior to this point, all but the first portion would not even have been found as a result. This fix is also important because it allows the INTERSECTS operator to work properly on the entire hit.
As Graham indicated, if you care about counts for original language data (or care about original language ordering) you should always be searching an original language resource.
Andrew Batishko | Logos software developer
0 -
So in essence that defeats the benefits of the Reverse Interlinear concept. If it counts ONE occurrence of an original language word now as TWO occurrences of that word (which is what is being asked in the query) the search is now incorrect and inaccurate. One of the benefits of your product over other products has been the ability to accurate original language work from the English translation. This change now tells people a word appears a certain number of times when it actually does not.
I agree that it has been frustrating at times that these searches would only show the first portion of the result in the English BUT theses searches resulted in the correct number of times the word was being used in the text. That accuracy could be verified by other tools like the Concordance tool. Now these two tools are showing different results. The loss of that accuracy is disappointing and from my perspective could be argued these searches now produce incorrect result.2015 13" MacBook Pro - 2 Ghz Intel i7 - 16 GB RAM - 500GB SSD - 2018 iMac Pro - 3.2GHz 8-core Xeon - both systems running OS 10.14.3 (Mojave)
0 -
So in essence that defeats the benefits of the Reverse Interlinear concept.
I've always looked at counts from the Reverse Interlinears as meaningless as the RI is off standard critical texts not the text actually used by the translators. Then again, I rarely care about count but rather about density where this change may have a significant impact. However, it is easier for me to identify split occurrences affecting the count than missed INTERSECT relationships.
This is not to say I don't understand your point. I don't know if the internal coding supports both scenarios. But it does illustrate the diversity of users Logos has to try to satisfy.
from my perspective could be argued these searches now produce incorrect result.
whereas from my perspective I could argue that you misunderstand what they are counting.
Orthodox Bishop Alfeyev: "To be a theologian means to have experience of a personal encounter with God through prayer and worship."; Orthodox proverb: "We know where the Church is, we do not know where it is not."
0 -
Hi Michael,
Thanks for raising the issue. It's a tough one. There's no perfect way to treat an interlinear because one language can never exactly "fit" another language.
I quite understand why you'd assume that searching for ἀπειλέω in a reverse interlinear is going to return a count of how many times that word occurs. But that's not what happens. Instead, you get a count for how many blocks of English text that word is aligned to - because in a reverse-interlinear the English text is primary, so the counts occur there.
For that reason, we've always advised users that if they want accurate counts, they should use an original language Bible. There's no way to guarantee correct counts for every type of search when using original languages in a reverse interlinear. For some searches, there's not even a way to define what that correct count should be! You were fortunate in the past that your searches in RIs generated the counts that you were expecting, but other users' searches didn't.
When we made this change, we took the view that we should optimize for the most relevant use case - combining English/Greek in the same query. That's what reverse interlinears are designed to enable.
As Graham has said above, if the counts are important, I would suggest using a Greek Bible with the English text in parallel. It gives the best of both worlds. I'm sorry that would mean you'd have to update all your favorites.
0 -
Mark,
First let me say that I am not trying to be difficult and I can't imagine that issues that arise from trying to make the RI work as well as it does.
Can you help me understand the difference in what the Concordance Tool does versus what a morph search does? When I sort a Concordance Tool report from the NASB by ROOT in 1 Peter it shows poieo appears 15 times. When I do the morph search it shows 20. Why the different results?2015 13" MacBook Pro - 2 Ghz Intel i7 - 16 GB RAM - 500GB SSD - 2018 iMac Pro - 3.2GHz 8-core Xeon - both systems running OS 10.14.3 (Mojave)
0 -
Mark,
First let me say that I am not trying to be difficult and I can't imagine that issues that arise from trying to make the RI work as well as it does.
Can you help me understand the difference in what the Concordance Tool does versus what a morph search does? When I sort a Concordance Tool report from the NASB by ROOT in 1 Peter it shows poieo appears 15 times. When I do the morph search it shows 20. Why the different results?2015 13" MacBook Pro - 2 Ghz Intel i7 - 16 GB RAM - 500GB SSD - 2018 iMac Pro - 3.2GHz 8-core Xeon - both systems running OS 10.14.3 (Mojave)
0 -
When we made this change, we took the view that we should optimize for the most relevant use case - combining English/Greek in the same query. That's what reverse interlinears are designed to enable.
lemma.g:ἀπειλέω produces two hits in 1 Pe 2:23 as discussed above
lemma.g:λόγος produces 2 hits in 1 Pe 3:1, but why is "to"-->13 not included when it points to the first "word"?
Dave
===Windows 11 & Android 13
0 -
lemma.g:λόγος produces 2 hits in 1 Pe 3:1, but why is "to"-->13 not included when it points to the first "word"?
That's because we only index the reverse interlinear data on the primary alignment terms, not the secondary alignment terms (note the lighter blue background on "to" in the reverse interlinear pane).
Andrew Batishko | Logos software developer
0 -
That's because we only index the reverse interlinear data on the primary alignment terms,
Thank you for the clarification.
Dave
===Windows 11 & Android 13
0
