Comments on Fuzzy Search

1. I suggest that Everything is not the best place for Fuzzy as results from other searches are not relevant and it cannot be customised for Fuzzy without detracting from the usual searches I conduct. So why not have a {Fuzzy} extension, which would confirm its uniqueness without compromise e.g. I'd know that a previous search from history was Fuzzy as opposed to Everything.
2. "the woman at the well"
Congratulations on finding the pericope in ESV, but Jn 3:36 being ranked #1 along with Jn 4:1 are hardly relevant!?
3. "wheat and tares"
Congratulations on finding "tares" in the ESV (which uses "weeds") but why is only one verse listed and why is Job #1?
Dave
===
Windows 11 & Android 13
Comments
-
[Y]
Meanwhile, Jesus kept on growing wiser and more mature, and in favor with God and his fellow man.
International Standard Version. (2011). (Lk 2:52). Yorba Linda, CA: ISV Foundation.
MacBook Pro MacOS Sequoia 15.4 1TB SSD
0 -
The search is placed in Everything because that is the default search. The vast majority of people using the search panel use this search, and the fuzzy search is a mechanism that is designed to be simple and just return results. Using a extension is not appropriate because the results for this search can not be combined with other search syntax. This is much more analogous to Media or books.logos.com which don't support full syntax, but just retrieve results that match the provided query as well as possible.
Most of the ranking complaints are because you are viewing the results in your preferred Bible, while the results rank highly due to some other Bible with different wording. This is actually the point of the search is to be able to find the verse without knowing which translation to look in.
#2 is due to a Bible having the following as the text of John 3:36: "That is why whoever accepts and trusts the Son gets in on everything, life complete and forever! And that is also why the person who avoids and distrusts the Son is in the dark and doesn’t see life. All he experiences of God is darkness, and an angry darkness at that.” The Woman at the Well". This most likely indicates that "The Woman at the Well" is a pericope which is incorrectly marked as part of the Bible text. I'll track that down so we can fix it.
#3 is because the 1890 Darby Bible has the following text for Job 31:40: "Let thistles grow instead of wheat, and tares instead of barley. The words of Job are ended." The Matthew verse actually seems to also be the result of finding a match in a pericope.
Andrew Batishko | Logos software developer
0 -
What would it look like to include in the results some type of notation that indicates which translation matched most closely (if not the default), including a link to that translation or a comparison with the n most closely matching versions?
0 -
The search is placed in Everything because that is the default search. The vast majority of people using the search panel use this search, and the fuzzy search is a mechanism that is designed to be simple and just return results. Using a extension is not appropriate because the results for this search can not be combined with other search syntax. This is much more analogous to Media or books.logos.com which don't support full syntax, but just retrieve results that match the provided query as well as possible.
This comes across as a little defensive to a legitimate criticism. I am surprised that most people use the Everything search. I find it almost useless in its massiveness, and feel like a fuzzy search makes much more sense as part of a Bible search. I get that an extension makes it harder to use, but to have Fuzzy Search included in "All Bibles" or "Top Bibles" or even a new "Unsure" would seem to be much more intuitive than the current placement.
Most of the ranking complaints are because you are viewing the results in your preferred Bible, while the results rank highly due to some other Bible with different wording. This is actually the point of the search is to be able to find the verse without knowing which translation to look in.
I may be wrong, but it seems like it would make more sense to prioritize the more common Bibles. It is unlikely someone is thinking of the Darby Bible in a half remembered verse, so a possible hit there should be weighted substantially lower than a hit in the ESV, NIV or KJV. So Matthew should be listed first in this example, with translation as one more consideration in the confidence assigned.
Using Logos as a pastor, seminary professor, and Tyndale author
0 -
For people with a small library, the Everything search makes sense, However, I think the following would improve the usefulness:
- also include the Fuzzy Bible search in the Bible Search as that is where people would think to look for it if they aren't an Everything user
- have the routine include the user's Bible prioritization in ordering from most to least likely matches
- have the ability to export the results to a passage list as there is no way to see the full list ... or to remember where in the "random" page breaks some particular result occurs.
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 -
Randy W. Sims said:
What would it look like to include in the results some type of notation that indicates which translation matched most closely (if not the default), including a link to that translation or a comparison with the n most closely matching versions?
This is a nice idea. I'll pass it along and see if it's feasible.
Andrew Batishko | Logos software developer
0 -
Justin Gatlin said:
The search is placed in Everything because that is the default search. The vast majority of people using the search panel use this search, and the fuzzy search is a mechanism that is designed to be simple and just return results. Using a extension is not appropriate because the results for this search can not be combined with other search syntax. This is much more analogous to Media or books.logos.com which don't support full syntax, but just retrieve results that match the provided query as well as possible.
This comes across as a little defensive to a legitimate criticism. I am surprised that most people use the Everything search. I find it almost useless in its massiveness, and feel like a fuzzy search makes much more sense as part of a Bible search. I get that an extension makes it harder to use, but to have Fuzzy Search included in "All Bibles" or "Top Bibles" or even a new "Unsure" would seem to be much more intuitive than the current placement.
My apologies for being defensive. I wanted to provide some explanation, but was under a time constraint and needed to respond quickly. In retrospect, I probably should have waited to respond.
I do plan on forwarding this thread tomorrow to the people who make design decisions. We do actually value getting immediate feedback on this feature, particularly since it went into the beta cycle a little later than we normally like.
Justin Gatlin said:Most of the ranking complaints are because you are viewing the results in your preferred Bible, while the results rank highly due to some other Bible with different wording. This is actually the point of the search is to be able to find the verse without knowing which translation to look in.
I may be wrong, but it seems like it would make more sense to prioritize the more common Bibles. It is unlikely someone is thinking of the Darby Bible in a half remembered verse, so a possible hit there should be weighted substantially lower than a hit in the ESV, NIV or KJV. So Matthew should be listed first in this example, with translation as one more consideration in the confidence assigned.
I'll pass this suggestion along.
Andrew Batishko | Logos software developer
0 -
Justin Gatlin said:
so a possible hit there should be weighted substantially lower than a hit in the ESV, NIV
Sorry but ESV and NIV are not versions I would be apt to remember ... I use them primarily for testing for the forums - never for study or memorization ... I would be interested in the ranking routine recognizing my personal prioritization but not some generalized "most common" list. I was raised on a KJV version that had updated language, used NEB in college classes, RSV/NRSV/JB/NJB/NAB/NABRE and Community Bible at various times since, and would most often be looking for a half-remembered verse of a small group participant who may have used pretty much any version.
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 -
MJ. Smith said:
For people with a small library, the Everything search makes sense, However, I think the following would improve the usefulness:
- also include the Fuzzy Bible search in the Bible Search as that is where people would think to look for it if they aren't an Everything user
- have the routine include the user's Bible prioritization in ordering from most to least likely matches
- have the ability to export the results to a passage list as there is no way to see the full list ... or to remember where in the "random" page breaks some particular result occurs.
We do actually have a spec that integrates Fuzzy Bible Search into the Bible Search, but we didn't have a chance to get to it this beta cycle.
It was actually my mistake for the results to show up in a paged format. There was supposed to be a "more" link that added additional results. This will be fixed in the next beta. I'm curious if this is fixed if there is still any value in saving to a Passage List. It seems like many of the lower ranked results don't tend to have much value.
Andrew Batishko | Logos software developer
0 -
a more link that added additional results.
This will be sufficient for me.
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 -
This is actually the point of the search is to be able to find the verse without knowing which translation to look in.
There have been some good suggestions to work around this 'point' because it definitely needs attention. Fuzzy, like Morph Query is not self-explanatory, and complaints/suggestions should be followed up appropriately.
#3 is because the 1890 Darby Bible has ...The Matthew verse actually seems to also be the result of finding a match in a pericope.
Does that explain why only the verse is quoted vs. the pericope at #2?
Having Fuzzy in Bible Search is definitely better than Everything, provided that it is accessed similar to Everything and is not part of a normal Bible Search.
Dave
===Windows 11 & Android 13
0 -
To throw in another opinion, I too would support that having Fuzzy Search under Bible search is a better place for it. Without knowing that a thread had come out explaining how to use it, the first place I looked was whether there was a new search type, like Morph, Syntax, etc. When I did not see it there, I looked under Bible. I never thought to look under Everything.
While the idea of putting it where a newbie will see it is good, to me the more compelling argument is "Say you want to find a Bible verse you can't remember exactly. Go to Basic, and search everything..." What? It makes more sense to say "Say you want to find a bible verse you can't remember. Do a Bible search, type in what you remember, and there it is in the first section of results!"
I also like the idea of saying what Bible the result came from in the results. I can always get back to my Bible and it could very well be I remember a phrase I could not find because it read it in a Bible translation I don't use very often. Nice suggestion!
As for the results - I have had great luck in using it so far, it's really good. Now maybe this is because I go back to Libronix (or at least my crusty memory of it by now lol), and I am used to using Google - but my scenario for Fuzzy search, and what I think most people desire it for, is "What was that Bible verse exactly, it is something like <.....>". So if I can type in my phrase and see it in the first page of results, which I can scan quickly, that is a great success! I know what I am looking for - it does not have to be first. It is FUZZY.
My fear is by trying to make things too exact in the results returned, too prioritized, making the right answer always first, it will no longer work as well. If you see it quickly it doesn't matter if it is first or third. Embrace the fuzziness!
0 -
1. Thank you very much for bringing back Fuzzy Search. The long wait (from Libronix 3 to Logos 7) is over.
2. Does this Fuzzy Search have a similar mechanism as the old Libronix one? That is, was the old one also based on searching for matching words in multiple English versions? (I always suspected there might have been some referencing to a thesaurus or dictionary of synonyms of sorts under the hood in the old Fuzzy Search. Was there / is there?)
3. As such, was this only designed to work with English Bibles? Would it work with Spanish Bibles, or Chinese Bibles (when there are multiple translations available in Logos), when my memory is sloppy with the exact detailed biblical wording?
4. I agree it makes more sense for this to be under Bible Search than Basic Search.
5. While I'm at it, would you consider changing "Basic Search" to "General Search" or "Search All"? To me, Basic Search has always felt like a misnomer because (a) the algorithm underneath it is anything but basic; it is just as sophisticated; (b) even the list of potential search options available is the full complement of options and features, so it is by no means Basic.
Thanks,
Peter
0 -
I believe, or at least I hope, that the fuzz won't be limited to bibles. Hoping fuzzy bible is just the initial roll-out. So, while it does belong in Bible search, it MAY also belongs under general search.
I am a bit concerned about the interface and feedback to let users know when fuzzy search is working and when it is not (for example when complex search terms are included as I think was mentioned as planned behavior in another thread), and why. Simply returning no results is ambiguous: is it bad syntax, no results, or too complex?
0 -
Any response from FL about these questions/comments?
PL said:2. Does this Fuzzy Search have a similar mechanism as the old Libronix one? That is, was the old one also based on searching for matching words in multiple English versions? (I always suspected there might have been some referencing to a thesaurus or dictionary of synonyms of sorts under the hood in the old Fuzzy Search. Was there / is there?)
3. As such, was this only designed to work with English Bibles? Would it work with Spanish Bibles, or Chinese Bibles (when there are multiple translations available in Logos), when my memory is sloppy with the exact detailed biblical wording?
4. I agree it makes more sense for this to be under Bible Search than Basic Search.
5. While I'm at it, would you consider changing "Basic Search" to "General Search" or "Search All"? To me, Basic Search has always felt like a misnomer because (a) the algorithm underneath it is anything but basic; it is just as sophisticated; (b) even the list of potential search options available is the full complement of options and features, so it is by no means Basic.
0 -
Sorry to be late to the party, folks. I was out sick with a cold week before last and then I was busy catching up .First of all, we didn't do a great job getting out ahead of this feature to explain it. Please bear with us. I'll try to summarize the questions that are raised in this thread as best I can.
Shouldn't this go in Bible Search instead?
We think it should go in both Basic and Bible Search. We put it in Everything Search first because that was easiest, in large part because Bible Search results views are much older code that is highly optimized to do exactly what it does right now, and also, it's much higher risk to mess around with. In any case, were we to put it into Bible Search today, it would display the same results as it would in Everything Search.
Will I be able to fuzzy search a specific Bible version?
Let me wait to answer that until further down.
How can I coax it produce better results?
More words is usually better. Or different words. But beyond that, there's not much in terms of settings that you can tweak to manipulate the results. This is by design, and is in part a consequence of the way most fuzzy text retrieval algorithms work. As Don Awalt correctly notes, the point of fuzzy searching is that it's imprecise. It is, to some extent, a guessing game on both ends: You're making your best guess at what will find what you wanted to find, and the results are the server's best guess at what you meant.
How does it work? / Is it like the Libronix Fuzzy Search?
Libronix used a certain single algorithmic technique to rank candidate verses against a query. Without getting too deep into the detail, this iteration of "Fuzzy Search" is a service that takes a set of words and sends them to a server that uses multiple different text retrieval and ranking techniques then aggregates those results to arrive at one set of ranked results. So the underlying technology doesn't use an algorithm so much as a cluster of algorithms.
This gives us a lot of freedom to fine tune the results on the server side and to try out the latest developments in the field at will. If one technique is contributing quirky results, we can drop it from the roster or lower its weight in the collation algorithm so that it only counts a little bit.
This verse isn't related, why did it come up?
Given the above, any specific explanation we may give for how a given verse arrived in a given set of results for a given query is likely to be overgeneralized enough that it's a little bit wrong. [:)] Also, the exact matching algorithm is a moving target, by design. Also, we've discovered a few bugs since the beta release.
The most helpful feedback at this point is "I searched for X, but Y showed up instead of Z." That will set off an investigation on our end that will probably result in some fine tuning. This may make the "X input returns Z output" situation better, but may make other things worse. We'll keep at it until we get a good balance. While each of the retrieval and ranking techniques is a science, balancing them to get good results in various situations is an art. The more examples we have, the better we'll get.
Will it work with non-English Bibles?
There's nothing stopping it. I don't know if the necessary server work has been done to make it happen.
Will we be able to Fuzzy Search non-Bibles?
Not yet, but it's possible given the technology approach outlined above. The reason we wouldn't just automatically make it work on every resource is that techniques for retrieving Bible verses (very small, highly marked candidate texts that are strongly correlated across many versions) are not necessarily going to be great for finding, say, sentences in any book.
So, will we be able to fuzzy search a specific Bible version?
That's an internal discussion that is ongoing. It's possible with the current technology as implemented to limit the retrieval engine to just work with data culled from one resource. But our tests show that doing so severely limits the effectiveness of the results. Too severely, as it stands. The choice we face isn't between results that match a given version poorly and results that would match it better; rather, it's between some false positives gathered from other translations and no results at all for many searches.
So for now, the set of results you get is a version-agnostic set of verses, and your choice of Bible doesn't influence the retrieval engine at all, but is just the choice of which Bible to render the returned verses. (The documentation really ought to point this out.)
But we totally understand that context matters, that is, that you're going to want searching NASB to look a little different than searching the Message and both of those to look a little different than searching All Bibles. If we were to do anything, it would be to nudge up results that are a closer match for the Bible you're searching. (And defining what "closer match" means in practice would bethe key issue to solve.) The good news is that since this is server technology, we can tweak it without having to update the client software.
I probably missed something.
0 -
Eli Evans (Faithlife) said:
I probably missed something.
One query from my original post
Why is the pericope Jn 4:7-26 listed for "woman at the well" vs. the first verse Mt 13:24 only for "wheat and tares". Highlighting the hits (Beta 4) doesn't improve the presentation as Mt 13:24 has no highlights. And Jn 3:36, 4:1 only highlight "the".
In Job 31:40 shouldn't "weeds" also be highlighted?
ESV is my bible.
Dave
===Windows 11 & Android 13
0 -
Thanks Eli for the reply. I am still having really good luck using it.
IMHO, many questions can be answered by thinking about, "What verse(s) was that I am trying to remember...." - the impreciseness. IMHO if that is really embraced, then:
(1) It is not helpful to search by Bible. Once I have the "Aha" moment, I know it's Matthew 6:32-35 I was searching for - then there are other tools that can help me compare Bible versions, etc.
(2) It doesn't matter what the other results are - they could be the craziest results in the world, but if on the first page, or top 10, I SPOT THE ANSWER I am looking for, who cares?. There it is! Why would I care about any other results?
(3) If it takes advantage of some of the tagging like "person", then if I am remembering "Jesus said....<...>", and the answer I search is "He replied .....<...>" that would be really nice.
(4) This is really different than searching for everything "fuzzy" related to a word or phrase. Again, there are much better linguistic-based tools for that in Logos. Unless I am missing something, this has one really important task - "What the heck was that verse I am trying to remember?"
Addendum:
I tried searching for
He said you shall not put the Lord to the test
and at least in the first 20 I did not see Matt 4:7. I changed 'He' to 'Jesus' and it was ranked 15th. Not too bad but could be better.
I also tried searching for
Adam said I was naked and I hid
And it found Gen 3:10 ranked #14 (this is an example of searching for a proper name but the Bible text is 'he'). Not bad, in both cases I would have found what I was looking for but the choices above them look like they could be optimized somewhat.
0 -
Dave Hooton said:Eli Evans (Faithlife) said:
I probably missed something.
One query from my original post
Why is the pericope Jn 4:7-26 listed for "woman at the well" vs. the first verse Mt 13:24 only for "wheat and tares". Highlighting the hits (Beta 4) doesn't improve the presentation as Mt 13:24 has no highlights. And Jn 3:36, 4:1 only highlight "the".
In Job 31:40 shouldn't "weeds" also be highlighted?
ESV is my bible.
Only words in your preferred Bible that match any of the terms in your query will be highlighted. If the verse was chosen due to strong hits in a version different than your preferred Bible, then you may see little or no highlighting.
If multiple verses have very high confidence, then they can be merged together into a single result.
Andrew Batishko | Logos software developer
0 -
Dave Hooton said:Eli Evans (Faithlife) said:
I probably missed something.
One query from my original post
Why is the pericope Jn 4:7-26 listed for "woman at the well" vs. the first verse Mt 13:24 only for "wheat and tares". Highlighting the hits (Beta 4) doesn't improve the presentation as Mt 13:24 has no highlights. And Jn 3:36, 4:1 only highlight "the".
In Job 31:40 shouldn't "weeds" also be highlighted?
ESV is my bible.
The best answer I have right now is: I don't know. I'll see if someone can find out, though.
0 -
PL, I think I answered your questions 2-4, but skipped #5: Would we consider changing the name of Basic Search, because it's anything but.
We agree that "Basic Search" has been a misnomer for a long time now. We've considered changing it several times, but we always argue ourselves back around to wanting to deliver more than just renaming it.
0 -
Don Awalt said:
Thanks Eli for the reply. I am still having really good luck using it.
Good, good. I think you're using it as intended, so that helps. [:)]
There's plenty of room for improvement. We'll look into those examples, thanks!
0 -
Hi Eli,
Thanks for your insights and answers.
I tried typing in some Chinese words to try out Fuzzy Search for the Chinese Bible.
The Fuzzy Search section does not appear when Chinese characters are entered as the search string in Basic / Everything Search.
Would you consider this a Bug?
Thanks,
Peter
0 -
PL said:
I tried typing in some Chinese words to try out Fuzzy Search for the Chinese Bible.
The Fuzzy Search section does not appear when Chinese characters are entered as the search string in Basic / Everything Search.
Would you consider this a Bug?
Fuzzy Search is currently only available for English. We will announce (via the release notes) when more languages are available.
0 -
Only words in your preferred Bible that match any of the terms in your query will be highlighted. If the verse was chosen due to strong hits in a version different than your preferred Bible, then you may see little or no highlighting.
I expected that! Which is why I think highlighting should not be performed as it gives rise to false expectations. I prefer an indication of where the results come from
Dave
===Windows 11 & Android 13
0 -
Don Awalt said:
(3) If it takes advantage of some of the tagging like "person", then if I am remembering "Jesus said....<...>", and the answer I search is "He replied .....<...>" that would be really nice.
This is already happening.
Andrew Batishko | Logos software developer
0 -
Eli Evans (Faithlife) said:
Will we be able to Fuzzy Search non-Bibles?
Not yet, but it's possible given the technology approach outlined above. The reason we wouldn't just automatically make it work on every resource is that techniques for retrieving Bible verses (very small, highly marked candidate texts that are strongly correlated across many versions) are not necessarily going to be great for finding, say, sentences in any book.
It would be very nice to have it expanded to Patristics where we also frequently have multiple translations and imprecise quotations.
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