Basic Search Results - Formatting & Accuracy

Dave Hooton
Dave Hooton MVP Posts: 35,775
edited November 21 in English Forum

Prime objective was to compare Basic Search results with v3 in a collection of Bible Dictionaries. It soon became obvious that getting results from L4 would be difficult:-

  • the results for one book extend over several pages. The header on each page contains what appears to be a cumulative count. The final count for the "God" search is different to v3.
  • No way to get to all the results pages as the page list does not scroll.

current page is 26

Intermediate result - the count on final header was 5995

 

  • I attempted to search individual books but there were no suggestions for several books. Results for those books come from their final header in the collection search e.g. &5995.

 

Basic Search for "God" and "Golden", no stemming (no matching all forms), case insensitive.

where possible results are shown as #hits / #articles


Resource

L3

God

L4

L3

Golden

L4
Bible Readers Companion** 8395 &5995 22 &22
Easton's Bible Dictionary** 1181 &1056 64 &64
Harper's Bible Dictionary 3875/953 2959/873 54/42 53/41
MacArthur Study Bible
12233/1876 8790/1741 75/58 75/58
NAS Exhaustive Concordance ** 18115 cannot get to result
234 &234
Vines's Dict OT & NT Words 3616/968 3033/915 19/12 19/12

 

 

 

 

 

 

 

 

 

 

 

 

** no suggestions for book, cannot search individually.

 

SUGGESTIONS:

  1. Show results for each book as per v3 ie. collapsed with header showing #hits / #articles. This will allow better navigation between books
  2. Improve suggestions for finding certain individual books on Search in
  3. Improve search accuracy

Dave
===

Windows 11 & Android 13

Tagged:

Comments

  • Robert Pavich
    Robert Pavich Member Posts: 5,685 ✭✭✭

    Dave,

    I'm not quite sure what you are getting at but I can comment on one thing: I like the "results" summary in V3 better....in one glance, you immediately know what book you want to look more in depth in.

     

    As you can see; on the right the search results span many OPEN pages...and how do i know how many pages to go to get to the next resource; or even WHAT that resource is?

    image

    Robert Pavich

    For help go to the Wiki: http://wiki.logos.com/Table_of_Contents__

  • Bob Pritchett
    Bob Pritchett Member, Logos Employee Posts: 2,280

    The search engines are different, and may return different counts for various reasons. Bradley may be able to answer as the specifics on some of these cases, just as examples.

    Do you want the 3.x results organization because it's more useful for your normal searching, or simply for doing this type of comparison? (If the latter, this is what we call a 'tester bug" -- it's only a bug to someone doing testing, not in a real user scenario.)

    We thought the 3.0 organization made it hard to find a result that was useful, particularly across a large collection. We do let you see results by book -- though not collapsed -- but figured that would be more useful with small collections of books, not the entire library.

    Is it useful for something other than lining up 3..x and 4.0? 

  • Robert Pavich
    Robert Pavich Member Posts: 5,685 ✭✭✭

    Bob,

    the only thing that I have a hard time with is the "results spread across multiple pages" issue.

    I don't know how far into the list to go to get to the next book OR what the next book result is.

    I would think that one of two ways would be a good way to go about it:

    1.) a kind of summary thing so I know what type of book returned the results

    OR

    2.) Don't "unroll" the results right off of the bat.

    OR

    3.) On the page results; somehow show when the next book begins; different color; bolded number or whatever.

    My point would be that some resources seem to have "better" answers than others..."God" in the book "who's who in christian history" would  probably be a waste of time to look through.

     

    Is there something that I'm missing about this?

    Robert Pavich

    For help go to the Wiki: http://wiki.logos.com/Table_of_Contents__

  • Jim Fulford
    Jim Fulford Member Posts: 38 ✭✭


    the only thing that I have a hard time with is the "results spread across multiple pages" issue.

    I don't know how far into the list to go to get to the next book OR what the next book result is.

    I Agree. I have found that, like google, I stop looking after the first number of pages - which seems counter productive to extracting the most from my library.

    1.) a kind of summary thing so I know what type of book returned the results

    This would be really cool, but I'm not sure how you would accomplish this.


    2.) Don't "unroll" the results right off of the bat.

    I think this is the answer. If you could, similar to v3, expand the results for the book you want to look in ... or if possible, have an "Expand All/Collapse All" feature to get to what v4 is now.


    3.) On the page results; somehow show when the next book begins; different color; bolded number or whatever.

    This would certainly ease the eye strain. [:)]
    Thanks.

    Laptop:Vista Business SP2 32-bit OS / Intel Duo Core P8400@2.26Ghz / 3-Gigs RAM

  • spitzerpl
    spitzerpl Member Posts: 4,998

    Is it useful for something other than lining up 3..x and 4.0? 

    I would love to have an Itunes type view, where the book covers for those found in the search appear without much explanation unless you hover over it.  Clicking on a book would jump to the location of that book in the current "list" format.

    The other problem I have with the current list is when many search results are returned from the same book. Once you are halfway through the results for that book its easy to loose sight of which book you are in. I would prefer to see the same type of thing that is done with the Bible search. Their the current Translation title stays at the top. I would love for the current book to "stick" to the first line. I'm sure that makes no sense...but thats ok :-)

  • Dave Hooton
    Dave Hooton MVP Posts: 35,775

    The search engines are different, and may return different counts for various reasons. Bradley may be able to answer as the specifics on some of these cases, just as examples.

    The differences for the word "God" are significant, and need explanation or fixing.

    Do you want the 3.x results organization because it's more useful for your normal searching, or simply for doing this type of comparison? (If the latter, this is what we call a 'tester bug" -- it's only a bug to someone doing testing, not in a real user scenario.)

    We thought the 3.0 organization made it hard to find a result that was useful, particularly across a large collection. We do let you see results by book -- though not collapsed -- but figured that would be more useful with small collections of books, not the entire library.

    Is it useful for something other than lining up 3..x and 4.0?

    1. BUG: I can't get to all the results for "God"

    2. What does the number of results in the header of each book represent when it changes with each page?

    3. Is the final number the total number of results for the book?

    4. The results from "golden" were spread over 8 pages but i have to scroll to find the results in a book of interest. The v3 overview simplified this task and gave me a single summary of the results for that book.

    I have absolutely no idea what you mean by "tester bug", but the current presentation of results gives me no idea about their meaning, let alone their accuracy.

    Later: as I reflect on what you said I'm still puzzled by the term "tester bug", just because I express a preference for another format. If there was a concise presentation of results as per v3 I would not have discovered that the Search in box does not reveal certain individual resources!

     

    Dave
    ===

    Windows 11 & Android 13

  • Robert Pavich
    Robert Pavich Member Posts: 5,685 ✭✭✭

    The search engines are different, and may return different counts for various reasons. Bradley may be able to answer as the specifics on some of these cases, just as examples.

    Do you want the 3.x results organization because it's more useful for your normal searching, or simply for doing this type of comparison? (If the latter, this is what we call a 'tester bug" -- it's only a bug to someone doing testing, not in a real user scenario.)

    We thought the 3.0 organization made it hard to find a result that was useful, particularly across a large collection. We do let you see results by book -- though not collapsed -- but figured that would be more useful with small collections of books, not the entire library.

    Is it useful for something other than lining up 3..x and 4.0? 

     

    Ok...I'll do you a favor; you fly me to Washington, put me up in a moderately swanky hotel and a mid sized rental car and I'll hang at the Logos offices for a week and we can talk and sort this all out...what do ya say?

     

    Uh...I probably need to be paid also...I don't have any vacation left [;)]

     

    Robert Pavich

    For help go to the Wiki: http://wiki.logos.com/Table_of_Contents__

  • Bradley Grainger (Logos)
    Bradley Grainger (Logos) Administrator, Logos Employee Posts: 11,969

    I'm going to break my replies up into individual findings as I investigate the multitude of issues reported in this post. [:)]

    the results for one book extend over several pages. The header on each page contains what appears to be a cumulative count

    This is a bug; thanks for reporting it.


    Dave Hooton said:

    No way to get to all the results pages as the page list does not scroll.

    This is a known issue ("Searches with many results may cause the result page selector to be clipped."); I think it will be fixed in Beta 4.

  • Robert Pavich
    Robert Pavich Member Posts: 5,685 ✭✭✭

    Bradley,

    I notice you convienently side-stepped the issue of flying me to the Logos headquarters to help you sort this out.. [:O]

    Robert Pavich

    For help go to the Wiki: http://wiki.logos.com/Table_of_Contents__

  • Bradley Grainger (Logos)
    Bradley Grainger (Logos) Administrator, Logos Employee Posts: 11,969


    Basic Search for "God" and "Golden", no stemming (no matching all forms), case insensitive.


    Apparently the stemming library we're using changed between LDLS3 and Logos 4, so that some of the words are stemmed differently; also, the definition of "nostem" is slightly different to "match all word forms":


    Stemming / Match all

    LDLS3

    Logos 4

    nostem / unchecked

    God God's

    God

    Stem / checked

    God God's gods

    God God's gods godly

    In LDLS3, "God’s" was returned even for "nostem", but in Logos 4, it's considered a different "word form" and not returned. Also, Logos 4 is returning "godly" (but not "godliness"...)  in a "match all word forms" search for "God".

    So, which search is "correct"?

  • Bradley Grainger (Logos)
    Bradley Grainger (Logos) Administrator, Logos Employee Posts: 11,969


    I notice you convienently side-stepped the issue of flying me to the Logos headquarters to help you sort this out.. Surprise


    I'm still on the third line of Dave's post; I haven't made it to yours yet!

  • Todd Phillips
    Todd Phillips Member Posts: 6,735 ✭✭✭

    Apparently the stemming library we're using changed between
    LDLS3 and Logos 4, so that some of the words are stemmed differently;
    also, the definition of "nostem" is slightly different to "match all
    word forms":

    Stemming / Match all

    LDLS3

    Logos 4

    nostem / unchecked

    God God's

    God

    Stem / checked

    God God's gods

    God God's gods godly

    In
    LDLS3, "God’s" was returned even for "nostem", but in Logos 4, it's
    considered a different "word form" and not returned. Also, Logos 4 is
    returning "godly" (but not "godliness"...)  in a "match all word forms"
    search for "God".

    So, which search is "correct"?

     

    I'd have to go with the LDLS3 method:

    Nostem: God matching God's is ok because it might be a contraction of "God is", but I'd be satisfied with Logos4 "nostem" if it was more limited on the stemming side.

    Stem: Godly is getting too far away from the original. A noun should match noun forms.  Mutating it to other types of words is trying to over-guess the user, IMO.  What's to stop the algorithm from including godliness or ungodly or godlike, etc?

    (Actually this makes me want both v3 and v4 stemming algorithms as options.)

    MacBook Pro (2019), ThinkPad E540

  • Bradley Grainger (Logos)
    Bradley Grainger (Logos) Administrator, Logos Employee Posts: 11,969


    Stem: Godly is getting too far away from the original.


    Yes, "godly" -> "god" seems like overstemming. We'll correct this for Beta 5.

  • Bradley Grainger (Logos)
    Bradley Grainger (Logos) Administrator, Logos Employee Posts: 11,969

    • I attempted to search individual books but there were no suggestions for several books.



    I assume this means:

    1. You clicked the resource picker in the Basic Search panel (the blue underlined text that lists "Entire Library", "All Open Resources", etc.)
    2. You started typing the name of a resource from your library
    3. No matches were shown.

    If so, this should be fixed in Beta 5.

  • Bradley Grainger (Logos)
    Bradley Grainger (Logos) Administrator, Logos Employee Posts: 11,969

    Basic Search for "God" and "Golden", no stemming (no matching all forms), case insensitive.

    I have ignored the rest of the "God" searches, because they're probably caused by the same underlying difference in stemming. As for the "golden" search, the extra hit in LDLS is from "Late Monarchy Jerusalem [Page 474]" (libronixdls:jump|pos=LLS-AOL%3A2041%3Cr1561%3E.25601.0|res=LLS%3A14.0.3). I haven't investigated in detail yet, but it seems as if the resource must have metadata attached to the image that includes the text "Golden Gate" (present in the upper-right corner of the image). This is apparently not getting indexed by Logos 4. We will investigate this further.

  • Dave Hooton
    Dave Hooton MVP Posts: 35,775

    it seems as if the resource must have metadata attached to the image that includes the text "Golden Gate" (present in the upper-right corner of the image). This is apparently not getting indexed by Logos 4.

    I think this will affect a number of map resources with place names e.g. Logos Deluxe Mapset has no Indexes, whilst v3 has 5.

    Dave
    ===

    Windows 11 & Android 13

  • Dave Hooton
    Dave Hooton MVP Posts: 35,775

    I assume this means:

    1. You clicked the resource picker in the Basic Search panel (the blue underlined text that lists "Entire Library", "All Open Resources", etc.)
    2. You started typing the name of a resource from your library
    3. No matches were shown.

    If so, this should be fixed in Beta 5.

    Thanks.

    Dave
    ===

    Windows 11 & Android 13

  • Robert Pavich
    Robert Pavich Member Posts: 5,685 ✭✭✭

    If so, this should be fixed in Beta 5.

     

    So I guess I should abandon my dream of visiting the Logos headquarters...

    Robert Pavich

    For help go to the Wiki: http://wiki.logos.com/Table_of_Contents__

  • spitzerpl
    spitzerpl Member Posts: 4,998

    If so, this should be fixed in Beta 5.

     

    So I guess I should abandon my dream of visiting the Logos headquarters...

    don't worry. Jesus is the Logos. All who are His will be in the Logos headquarters one day :-)

  • Dave Hooton
    Dave Hooton MVP Posts: 35,775

    Apparently the stemming library we're using changed between LDLS3 and Logos 4, so that some of the words are stemmed differently; also, the definition of "nostem" is slightly different to "match all word forms":

     

    Stemming / Match all

    LDLS3

    Logos 4

    nostem / unchecked

    God God's

    God

    Stem / checked

    God God's gods

    God God's gods godly

    In LDLS3, "God’s" was returned even for "nostem", but in Logos 4, it's considered a different "word form" and not returned. Also, Logos 4 is returning "godly" (but not "godliness"...)  in a "match all word forms" search for "God".

    So, which search is "correct"?

    A search for god OR God’s restores confidence as the results from L4 now match L3.

    If consistent throughout, L4 is correct for a nostem search. You get what you ask for, no second guessing the user's intentions. The word picker lists God's in case I want that form.

    As for stemming searches, matching (nearly) all word forms L4 is overstemming with godly. If this correction is made will the stemming algorithm for L4 be the same as L3? What are the possible disadvantages of returning to the L3 algorithm?

     

    Dave
    ===

    Windows 11 & Android 13

  • Dave Hooton
    Dave Hooton MVP Posts: 35,775

    So I guess I should abandon my dream of visiting the Logos headquarters...

    Yup. You get to spend more time with all the new Indexes you requested i.e. longer Indexing sessions.[8-|]

    Dave
    ===

    Windows 11 & Android 13

  • Dave Hooton
    Dave Hooton MVP Posts: 35,775

    I'd be satisfied with Logos4 "nostem" if it was more limited on the stemming side.

    Stem: Godly is getting too far away from the original. A noun should match noun forms.  Mutating it to other types of words is trying to over-guess the user, IMO.  What's to stop the algorithm from including godliness or ungodly or godlike, etc?

    Appreciate your insights, Todd

    Dave
    ===

    Windows 11 & Android 13

  • Robert Pavich
    Robert Pavich Member Posts: 5,685 ✭✭✭

    Robert Pavich

    For help go to the Wiki: http://wiki.logos.com/Table_of_Contents__

  • Bradley Grainger (Logos)
    Bradley Grainger (Logos) Administrator, Logos Employee Posts: 11,969


    As for stemming searches, matching (nearly) all word forms L4 is overstemming with godly. If this correction is made will the stemming algorithm for L4 be the same as L3? What are the possible disadvantages of returning to the L3 algorithm?

    It will be the same for this word.

    We didn't write the stemming algorithm; we use the ones defined by the Snowball project. The author of those tweaks them over time as he finds improvements, so LDLS3 was using whatever was current at the time it was written. We don't have an exhaustive list of how the updates to the algorithm will change stemming. I do know of some improvements: "on" and "one" used to be stemmed the same in LDLS3, and that has now been fixed.

  • Dave Hooton
    Dave Hooton MVP Posts: 35,775


    As for stemming searches, matching (nearly) all word forms L4 is overstemming with godly. If this correction is made will the stemming algorithm for L4 be the same as L3? What are the possible disadvantages of returning to the L3 algorithm?

    It will be the same for this word.

    I don't understand your response. What changes are being proposed for Beta 5?

    Dave
    ===

    Windows 11 & Android 13

  • Bradley Grainger (Logos)
    Bradley Grainger (Logos) Administrator, Logos Employee Posts: 11,969


    As for stemming searches, matching (nearly) all word forms L4 is overstemming with godly. If this correction is made will the stemming algorithm for L4 be the same as L3? What are the possible disadvantages of returning to the L3 algorithm?

    It will be the same for this word.

    I don't understand your response. What changes are being proposed for Beta 5?


    The change in Beta 5 is that "godly" and "godliness" will have the same stem, and it will be different to the stem for "God", "God's", "gods". So, the word "godly" will be stemmed the same as in LDLS3.

  • Dave Hooton
    Dave Hooton MVP Posts: 35,775

    • I attempted to search individual books but there were no suggestions for several books.

    I assume this means:

    1. You clicked the resource picker in the Basic Search panel (the blue underlined text that lists "Entire Library", "All Open Resources", etc.)
    2. You started typing the name of a resource from your library
    3. No matches were shown.

    If so, this should be fixed in Beta 5.

    Still not fixed in Beta 5!

    Dave
    ===

    Windows 11 & Android 13

  • Dave Hooton
    Dave Hooton MVP Posts: 35,775

    the results for one book extend over several pages. The header on each page contains what appears to be a cumulative count

    This is a bug; thanks for reporting it.


    Dave Hooton said:

    No way to get to all the results pages as the page list does not scroll.

    Thanks, the above have been fixed in Beta 5

    Dave
    ===

    Windows 11 & Android 13