Copied to clipboard

Flag this post as spam?

This post will be reported to the moderators as potential spam to be looked at


  • Kev 16 posts 67 karma points
    Sep 17, 2014 @ 17:39
    Kev
    0

    Searching by name not including some nodes in results

    Hi,

    I'm having issues with search results, it picks up most things but the big problem is searching for events. Some terms display the event & others don't. See below for tests & a link.

    http://www.fernehamhall.co.uk/search

    Search term tests:

    • The Sound of Music (no results)
    • sound (The Sound of Music is shown)
    • music (The Sound of Music is shown)
    • sound of (no results)
    • Foster and allen (no results)
    • foster (event shown).

    Other terms work absolutely fine though e.g.

    • worst serial killers
    • havant chamber

    Does anyone have any suggestions on what could be the issue?

    p.s. I've already rebuild the index and no luck.

    Thanks in advance

  • Matt Brailsford 4125 posts 22222 karma points MVP 9x c-trib
    Sep 17, 2014 @ 18:31
    Matt Brailsford
    0

    Hi Kev,

    It's hard to say without looking into the index to see what is going on. The best I can suggest is to install Luke (https://code.google.com/p/luke/) and then debug into ezSearch.cshtml to just before the query runs and copy the generated query into Luke and see what part is erroring.

    If you can do that, maybe copy the lucene query here and I'll see if I can spot anything obvious.

    Matt

  • keilo 568 posts 1023 karma points
    Sep 17, 2014 @ 18:32
    keilo
    0

    Not sure if you have seen this patch to better support phrases in ezsearch (added by user DougMac)

     

    https://github.com/DougMac/ezSearch/commit/a74644d5dbebba1772f3eebe8a4d1a669f8a32f9

  • Matt Brailsford 4125 posts 22222 karma points MVP 9x c-trib
    Sep 17, 2014 @ 18:45
    Matt Brailsford
    0

    Hi Keilo, I don't think Dougs fix would make much of a difference here. As Kev's searches are not phrase searches, ie he isn't wrapping the phrases in quotes, the fix by doug wouldn't apply.

    Ultimately, ezSearch splits multi term searches not wrapped in quotes into OR tokens, so what Kev is searching for should yield results.

    Matt

  • keilo 568 posts 1023 karma points
    Sep 17, 2014 @ 19:05
    keilo
    0

    Hi Matt

    You have a point. As he stated others with multi terms do return results whereas certain instances it doesnt - all without quotes.

    I was playing around with the ezsearch as I saw the post and recently Doug's updates. Offtopic but I was trying to figure out if I can put a simple dropdown where user can indicate search by date (most recent) or most relevant. Not sure if you have implemented something similar with it?

    cheers

  • Doug Mackay 56 posts 155 karma points
    Sep 18, 2014 @ 12:23
    Doug Mackay
    1

    I think the issue here is that the Umbraco ExternalSearcher based on the Lucene standard analyser has a default set of stop words 

    "a", "an", "and", "are", "as", "at", "be", "but", "by", "for", "if", "in", "into", "is", "it", "no", "not", "of", "on", "or", "such", "that", "the", "their", "then", "there", "these", "they", "this", "to", "was", "will", "with"

    As ezSearch is doing a grouped or on the keywords if a keyword is also a default stop word then no results will be returned.

    Currently testing this theory. I think the easiest thing to do is for ezSearch to strip these words from a keyword search (they're fine in phrase search). It is possible to specify the stop words in the Lucene analyser consturctor but this would require a change in the Umbraco core.

     

  • Matt Brailsford 4125 posts 22222 karma points MVP 9x c-trib
    Sep 18, 2014 @ 12:49
    Matt Brailsford
    101

    Hmm, I think you might be right Doug. Setup a node named "The Sound of Music" on my test install, and searching for "Sound of Music" does fail, but if I change line 63 of ezSearch.cshtml to the following:

    model.SearchTerms = Tokenize(model.SearchTerm)
        .Where(x => !Lucene.Net.Analysis.StopAnalyzer.ENGLISH_STOP_WORDS_SET.Contains(x));
    

    it does indeed return results (after stripping out any stop words). I guess I need to work out if stripping stop words is a good idea or not. I think the only extra thing to do would be to check after tokenization to make sure we still have a search term to search for (ie, make sure all tokens weren't stop words) and if it's now emtpy, don't show any results (basically one more big if statement round everything).

    Anyway, I think the code above should give Kev the results he needs for now though.

    Matt

  • Jamie Hall 7 posts 77 karma points
    Mar 30, 2017 @ 12:37
    Jamie Hall
    0

    Worked perfectly for me!

  • Doug Mackay 56 posts 155 karma points
    Sep 18, 2014 @ 14:22
    Doug Mackay
    1

    I think that will break phrase searches that contain stop words.

    I've put my changes here https://github.com/DougMac/ezSearch/compare/patch-3?expand=1. Needs a bit more testing before a pull request.

  • Matt Brailsford 4125 posts 22222 karma points MVP 9x c-trib
    Sep 18, 2014 @ 14:31
    Matt Brailsford
    0

    Doug, would it? The stop words are only stop words if they are on their own. Passing a phrase into the where statement would not match an entry in the set so wouldn't get removed and so should still go through.

    I'll have to take a look at your patch a bit later as it's showing quite a lot of changes and I haven't got much time to look through it at the moment.

    Matt

  • Doug Mackay 56 posts 155 karma points
    Sep 18, 2014 @ 14:56
    Doug Mackay
    0

    Yes you're right for some reason I was thinking the contains in the where statement was checking the opossite way.

    Ignore my patch (there was a lot of changes reported due to the additional if statement but the main change was in the Tokenize method)

  • Matt Brailsford 4125 posts 22222 karma points MVP 9x c-trib
    Sep 18, 2014 @ 15:15
    Matt Brailsford
    0

    Hey Doug,

    Cool. Yea, I've moved the where statement into the Tokenize method now so it sounds like we are pretty much inline with one and other now so all good.

    If I get some time tonight, I'll pull down your pull requests and push these all out as an update.

    Many thanks

    Matt

  • Kev 16 posts 67 karma points
    Sep 19, 2014 @ 16:58
    Kev
    0

    Thank you so much for this fix guys!

    I really appreciate the time & effort.

    Kev

Please Sign in or register to post replies

Write your reply to:

Draft