Issues with Fuzzy Bible Search

Sean Boisen
Sean Boisen Member, Logos Employee Posts: 1,452
edited November 2024 in English Forum

(Forking a new thread from https://community.logos.com/forums/p/187018/1095333.aspx#1095333 and subsequent comments to focus the conversation on Fuzzy Bible Search and possible improvements)

«1

Comments

  • Sean Boisen
    Sean Boisen Member, Logos Employee Posts: 1,452

    Lee Gordon reports here that Google search for "verse demon man leaps" finds the desired verse (Acts 19:16) but that he can't get the same result from Logos. 

    I can confirm that Acts 19:16 doesn't occur in the first 2 dozen results for "demon man leaps" ("verse" would just be noise for Logos): i'll add this to our test sets, and see if we can figure out why. 

  • DMB
    DMB Member Posts: 13,715 ✭✭✭

    Fuzzy search heuristics normally focus on least-likely and similar, then constraining to most likely. I suspect the Logos approach is opposite the google, by intent, and a different overall goal (looks like similar wording approach).

    "If myth is ideology in narrative form, then scholarship is myth with footnotes." B. Lincolm 1999.

  • Sean Boisen
    Sean Boisen Member, Logos Employee Posts: 1,452

    There's more information about the workings of Fuzzy Bible Search here: https://community.logos.com/forums/t/144666.aspx

    Fuzzy Search pools the content from a large number of Bibles. But for this case, none of those Bibles use "demon" in its translation of Acts 19:16, and we don't have the knowledge that demon and evil spirit are equivalent here. (for the query "evil spirit man leaps", Acts 19:16 is the first result)

  • James Hudson
    James Hudson Member Posts: 337 ✭✭✭

    we don't have the knowledge that demon and evil spirit are equivalent

    Could you not leverage the LCV for cases such as these?

  • Sean Boisen
    Sean Boisen Member, Logos Employee Posts: 1,452

    we don't have the knowledge that demon and evil spirit are equivalent

    Could you not leverage the LCV for cases such as these?

    We have resources like LCV and the Bible Sense Lexicon that would help us address this issue. But it still takes development work to analyze the texts, determine what counts as close in meaning, process queries, etc. The thing we don't have (though Google does) is billions of searches and click-through data: our usage is nowhere near that scale, and that scale (along with an army of PhDs and servers) is a lot of what enables their good results. 

  • James Hudson
    James Hudson Member Posts: 337 ✭✭✭

    True about Google usage data. But with the Bible at least the result set is finite - it is a much smaller area of focus- there are at least only about 31000 total possible outcomes to either accept or reject. Anyway, sure you are continually tweaking your algorithms.

    Thanks for all you do! Hope you are hard at work with Logos 9!!!!

  • David Ames
    David Ames Member Posts: 2,977 ✭✭✭

    Fuzzy Search pools the content from a large number of Bibles. But for this case, none of those Bibles use "demon" in its translation of Acts 19:16, and we don't have the knowledge that demon and evil spirit are equivalent here. (for the query "evil spirit man leaps", Acts 19:16 is the first result)

    demon NEAR man NEAR leap [Everything]

    Acts 19:16
    16 καὶ ἐφαλόμενος ὁ ἄνθρωπος ἐπʼ αὐτοὺς ἐν ᾧ ἦν τὸ πνεῦμα τὸ πονηρόν, κατακυριεύσας ἀμφοτέρων ἴσχυσεν κατʼ αὐτῶν ὥστε γυμνοὺς καὶ τετραυματισμένους ἐκφυγεῖν ἐκ τοῦ οἴκου ἐκείνου 1
    Nestle, E., Nestle, E., Aland, B., Aland, K., Karavidopoulos, J., Martini, C. M., & Metzger, B. M. (1993). The Greek New Testament (27th ed., Ac 19:16). Stuttgart: Deutsche Bibelgesellschaft.

    (16) And the man in whom was the evil demon leaped upon them, and overpowered them, and threw them down: and they fled out of the house denuded and bruised. 1
    Murdock, J. (Trans.). (2001). The Syriac New Testament: Translated into English from the Syriac Peshitto Version (Ac 19:16). Piscataway, NJ: Gorgias Press.

  • Donnie Hale
    Donnie Hale Member Posts: 2,036

    Fuzzy search is getting better. But even last week it didn't come up with a verse and I went to DuckDuckGo (I very rarely search w/ google any longer). I think the biggest challenge is that web search engines get immediate feedback on the accuracy of their results based on the links the user clicks on. So it's not just a full-text search (w/ adjustments for h1/h2/h3, etc.). I don't know how FL would be able to as easily and as usefully incorporate that into its ranking of search results.

    So the direct answer to your question is: online search gives better results when I'm having a hard time remembering the exact word/words or nearness of words.

    -Donnie

  • Al Het
    Al Het Member Posts: 206

    Sean,

    I've been recommending on these forums that people forget about Logos and use Google to search for Bible verses for years.  The problem is, in Logos, you have to know what type of search you want to do (like a "fuzzy" search) and different ways to insert the search perimeters, etc. If I'm using my Logos library to search a variety of different things, in various different ways every day, I might be able to make Logos give me what I want pretty quickly.  Because I don't use search often enough, I don't even remember how to do a fuzzy search from one use to the next.

    However, when I can't remember a specific verse or reference for a verse, and I know I don't have the wording correct, I can nearly always find that verse in seconds with Google.  I open my browser, and usually only type a couple of words before listings come up with the reference I want in it. 

    So, if I can't remember exactly, but I want to find the verse that says something about being new in Christ, I can do a fuzzy search in Logos.  If I haven't done one in a while, I might have to google how to do that.  Then, I type in "new in Christ."  Actually, I was just surprised when I did this in Logos (after googling how to do a fuzzy search) that the verse I was after came up 11 verses down in the search.  That's not bad at all, given how poorly I was remembering 2 Cor. 5:17. 

    However, when I opened my browser (which opens MUCH faster than Logos) and typed the words "new in Christ," the passage I want is the first one that comes up.  I can't remember a time that I was looking for a verse that I only remembered loosely that the result didn't come up within the first several results.  And with the results, I pretty much never need to even click on the link - the preview gives the info I need.

    I don't know if Logos will ever be able to match this.  However, I do know that even back in the late 90's, when Google wasn't good at Bible searches, I could nearly always find a verse far more easily with a very basic Quickverse program than I could with Logos.  It had a simple search interface that nearly always found the right verse.  Not sure why.  I've always assumed it was that Logos was too detailed and "advanced" to do such simple searches easily. 

    Over the years, so many people have asked about this in these forums, and I know that the "fuzzy search" was your attempt to address this.  People who ask on these forums are told how to appropriately search for the passage, perhaps using certain characters or formatting, and other really great advice on how to use Logos for this. It seems most Logos userse would much rather use their Logos to search for Bible verses than Google.  Having just used Logos 8 for the first time for this before writing this post, it seems to be getting a lot better.  Still, I think ease of use and effectiveness need to increase to rival using Google for this particular task.

    Hope this input is helpful, not just more noise to wade through for you.

    Thanks for your interest in making Logos better.

  • Sean Boisen
    Sean Boisen Member, Logos Employee Posts: 1,452

    Thanks Al, that's a very helpful response. You're right that we can't necessarily expect to match Google results: their resources are many many times larger than ours, including the number of users and searches they perform every day. But we want to do the best we can and continue improving (including making fuzzy search easier to perform so you don't have to remember how to access it or technical search details). 

    If you have specific cases that don't work in Logos or work poorly, feel free to post them to this thread so we can triage them. 

  • Graham Criddle
    Graham Criddle MVP Posts: 32,681

    Al Het said:

    So, if I can't remember exactly, but I want to find the verse that says something about being new in Christ, I can do a fuzzy search in Logos.  If I haven't done one in a while, I might have to google how to do that.  Then, I type in "new in Christ."  Actually, I was just surprised when I did this in Logos (after googling how to do a fuzzy search) that the verse I was after came up 11 verses down in the search.  That's not bad at all, given how poorly I was remembering 2 Cor. 5:17

    Interestingly, the example you quote performs rather well in Logos - in the actual search as opposed to the fuzzy one

    This particular search syntax is finding all verses with all three terms present.

    But I recognise that having to remember how to do Logos searches doesn't always make it easy.

    Al Het said:

    It seems most Logos userse would much rather use their Logos to search for Bible verses than Google

    I'm intrigued by this as I recognise myself in the description.

    I guess it's a combination of nearly always having Logos open when I am doing any type of study and typically wanting to do something with search results within Logos once I have found them. 

  • Randy
    Randy Member Posts: 112

    Lee Gordon reports here that Google search for "verse demon man leaps" finds the desired verse (Acts 19:16) but that he can't get the same result from Logos. 

    I can confirm that Acts 19:16 doesn't occur in the first 2 dozen results for "demon man leaps" ("verse" would just be noise for Logos): i'll add this to our test sets, and see if we can figure out why. 

    One thing that seemed evident from the search results, is that it appeared to mostly ignore the word "leaps". Every result that I was able to view, except the one from Job, only highlighted the words "demon" and "man":

    Does Logos only search for the first two words? Does it weigh verbs the same as other words? Should it give higher priority to results that match all the search input (e.g. verses that have all three words, instead of just two)?

    Another issue, is that the list contains more results than I'm able to view. When doing the search above, I can only hit the "more>>" option twice, and then the list goes clear to the bottom of the screen, and you can no longer scroll down any further or see the rest of the results. 

  • Randy
    Randy Member Posts: 112

    Lee Gordon reports here that Google search for "verse demon man leaps" finds the desired verse (Acts 19:16) but that he can't get the same result from Logos. 

    I can confirm that Acts 19:16 doesn't occur in the first 2 dozen results for "demon man leaps" ("verse" would just be noise for Logos): i'll add this to our test sets, and see if we can figure out why. 

    Here's another search issue that just came to my attention:

    I restricted the results to the "New Testament", then it starts listing verses from Genesis. If you want to restrict it to the New Testament, how do you do this? Am I doing it wrong, or is it showing me results it shouldn't be showing me?

  • Graham Criddle
    Graham Criddle MVP Posts: 32,681

    It was reported three years ago - https://community.logos.com/forums/p/148370/922009.aspx#922009 - that Fuzzy search ignores range parameters. I don’t believe this has changed

    But a regular, non-Fuzzy, search does and would be better for this particular example.

  • Randy
    Randy Member Posts: 112

    It was reported three years ago - https://community.logos.com/forums/p/148370/922009.aspx#922009 - that Fuzzy search ignores range parameters. I don’t believe this has changed

    But a regular, non-Fuzzy, search does and would be better for this particular example.

    Okay thanks.

  • Dave Hooton
    Dave Hooton MVP Posts: 35,800

    Sean

    It would be less confusing if Fuzzy was a Search on its own i.e. alongside Bible, Basic, Media, etc. Because it searches for a match in OT/NT/Apocrypha and ignores 'normal' Search criteria. Being part of a 'normal' Search, I have to remember to switch it off when I don't want results.

    Dave
    ===

    Windows 11 & Android 13

  • Robert Kelbe
    Robert Kelbe Member Posts: 614 ✭✭✭

    I will preface this by saying that the Fuzzy Bible Search often works really well, and I am grateful for it!

    Here is an example where I couldn't find it with the Fuzzy Bible Search, but when I searched in Google, it was the first result. I could have played around with the search parameters, but it was easier to just try Google.

    Additionally, others have mentioned that it froze as they were trying to see all the results in Fuzzy Bible Search. I can confirm this... It seems like the scroll bar is missing, so once you click "More" so that the results go off the page you can no longer scroll to see the results. 

  • DMB
    DMB Member Posts: 13,715 ✭✭✭

    I couldn't find it with the Fuzzy Bible Search

    Not disagreeing with the need for a more robust FBS. And not crashing would a good thing.

    But technically, FSB returned the correct answer from 1Esdras. And scholars note the match as well, to Matt. So, quite impressive.

    "Technically", meaning, 'tracking a fuzzy semantic match'. The google return was a doctrinal match (also correct).  In the earlier discussion, FSB is limited to the text (semantic matching).

    "If myth is ideology in narrative form, then scholarship is myth with footnotes." B. Lincolm 1999.

  • Sean Boisen
    Sean Boisen Member, Logos Employee Posts: 1,452

    Thanks for the detailed report Robert: these specifics really help. 

  • MWW
    MWW Member Posts: 427 ✭✭

    Here is another example of Fuzzy Search not finding a result in what should be a fairly strait forward association. The search = "no man knows the day". 

  • Randy
    Randy Member Posts: 112

    In this example, your search is limited to the KJV. What happens if you do the same search with more versions?

  • Graham Criddle
    Graham Criddle MVP Posts: 32,681

    MWW said:

    Here is another example of Fuzzy Search not finding a result in what should be a fairly strait forward association. The search = "no man knows the day". 

    Strange - on my system, there are results and the specific verse you are searching for, I think, is the fifth one down.

    Were you online when running the search?

  • MWW
    MWW Member Posts: 427 ✭✭

    In this example, your search is limited to the KJV. What happens if you do the same search with more versions?

    See below... If I scroll down near the bottom of the search I find what I am looking for... Fuzzy Search didn't work at all?

  • Randy
    Randy Member Posts: 112

    MWW said:

    In this example, your search is limited to the KJV. What happens if you do the same search with more versions?

    See below... If I scroll down near the bottom of the search I find what I am looking for... Fuzzy Search didn't work at all?

    My guess is the Logos Search feature operates in a more purely analytical way, while Google search results probably incorporate vast amounts of past user experience. That is, they know that's what you're looking for, because thousands of people have already searched for something similar, and ultimately clicked on this link.

  • Graham Criddle
    Graham Criddle MVP Posts: 32,681

    MWW said:

    Fuzzy Search didn't work at all?

    Your screenshot shows it is unavailable:

    1. Were you online when running the search?
    2. What base package(s) do you have?
  • MWW
    MWW Member Posts: 427 ✭✭

    Your screenshot shows it is unavailable:

    1. Were you online when running the search?
    2. What base package(s) do you have?

    1. Yes I was online at the time. 

    2. Base Packages = Standard Diamond, Baptist Diamond, P & C Platinum, AnglicanSilver, Lutheran Bronze, Messianic Silver, Methodist Silver, Reformed Bronze.

  • MWW
    MWW Member Posts: 427 ✭✭

    I just ran the search again and my results were the same as Graham's. Perhaps a hiccup in the cyber universe?

  • Sean Boisen
    Sean Boisen Member, Logos Employee Posts: 1,452

    <snip />

    My guess is the Logos Search feature operates in a more purely analytical way, while Google search results probably incorporate vast amounts of past user experience. That is, they know that's what you're looking for, because thousands of people have already searched for something similar, and ultimately clicked on this link.

    That's largely correct: much of the ranking is based on matching terms in the query against a large collection of Bibles. We hope to incorporate a metric for overall verse popularity, which would probably boost Matt 24:36 higher in the rank. Unfortunately we don't have the mountain of click-through data for Bible queries that Google has. 

  • Sean Boisen
    Sean Boisen Member, Logos Employee Posts: 1,452

    In this example, your search is limited to the KJV. What happens if you do the same search with more versions?

    Currently Fuzzy Bible Search results aren't limited to the range or Bibles specified: results are shown in your preferred version, and only the results in Verses are filtered by range. We may add filtering by range in the future, but the fuzzy search approach assumes you're looking at a lot of Bibles, so that's not likely to change. 

  • Randy
    Randy Member Posts: 112

    (Forking a new thread from https://community.logos.com/forums/p/187018/1095333.aspx#1095333 and subsequent comments to focus the conversation on Fuzzy Bible Search and possible improvements)

    Here's another one. I tried a search for "Eli's sons", to find the account where it says they were rebuked for misbehaving.

    Google gave the exact reference to what I was looking for as the first result.

    Here's Logos: