SUMO community discussions

Has the Search function been changed, poor results

  1. Update We finally got a string / phrase search and boolean operators in Sumo advanced search. But not until after the Google search was made useless by removing yet more indexing.

    • See Change in support forum threads indexed by Google or other search engines: Only solved threads will be shown /forums/contributors/710806


    Search results are often very disappointing. This one is for a single word ancestry I had hoped it would cope with that without any problems.

    • Using the search bar on the forum I get a single hit why ?
    • Same using Firefox's Searchbar with Sumo's "Search using Firefox help"
    1. https://support.mozilla.org/en-US/search?esab=a&q=ancestry.+
      screenshot attached showing it is supposedly an all products & Show everything search
    2. Try the search engine from the site in Firefox's search bar again the same single hit (/questions/981210)
      https://support.mozilla.org/en-US/search?q=ancestry+&w=3&qs=plugin
      Again similar screen display
    3. Now try from advanced search
      https://support.mozilla.org/en-US/search?q=ancestry&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2
      That's more like it 95 hits.
    4. Google exceeds 1300 certainly the first few pages seem good hits

    THE single hit sumo-search-why-single hit

    Update We finally got a string / phrase search and boolean operators in Sumo advanced search. But not until after the Google search was made useless by removing yet more indexing. *See ''Change in support forum threads indexed by Google or other search engines: Only solved threads will be shown'' </ br> [/forums/contributors/710806] Search results are often very disappointing. This one is for a single word ''ancestry'' I had hoped it would cope with that without any problems. *Using the search bar on the forum I get a single hit why ? *Same using Firefox's Searchbar with Sumo's ''"Search using Firefox help"'' #https://support.mozilla.org/en-US/search?esab=a&q=ancestry.+ <br />screenshot attached showing it is supposedly an ''all products'' & ''Show everything'' search # Try the search engine from the site in Firefox's search bar again the same single hit ([/questions/981210])<br /> https://support.mozilla.org/en-US/search?q=ancestry+&w=3&qs=plugin <br /> Again similar screen display #Now try from advanced search <br />https://support.mozilla.org/en-US/search?q=ancestry&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 <br />That's more like it 95 hits. #Google exceeds 1300 certainly the first few pages seem good hits THE single hit [[Image:sumo-search-why-single hit]]

    Modified by John99 on

  2. If I do this google search,

    I get lots better results so there's definitely a problem with the search here.

    If I do this google search, *https://www.google.com/search?&q=ancestry+site:support.mozilla.org I get lots better results so there's definitely a problem with the search here.
  3. An Advanced Search on the Support Questions tab with no boxes checked returns 95 results:

    https://support.mozilla.org/en-US/search?
    q=ancestry.&
    num_voted=0&
    num_votes=&
    asked_by=&
    answered_by=&
    q_tags=&
    created=0&
    created_date=&
    updated=0&
    updated_date=&
    sortby=0&
    a=1&
    w=2#support

    But I don't know whether this is a new problem.

    An Advanced Search on the Support Questions tab with no boxes checked returns 95 results: [https://support.mozilla.org/en-US/search?q=ancestry.&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2#support https://support.mozilla.org/en-US/search?<br>q=ancestry.&<br>num_voted=0&<br>num_votes=&<br>asked_by=&<br>answered_by=&<br>q_tags=&<br>created=0&<br>created_date=&<br>updated=0&<br>updated_date=&<br>sortby=0&<br>a=1&<br>w=2#support] But I don't know whether this is a new problem.

    Modified by jscher2000 - Support Volunteer on

  4. If you do an Advanced search and screen out Archived threads, then you only get two results:

    https://support.mozilla.org/en-US/search?q=ancestry&is_archived=-1&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2

    I'm guessing that a normal search screens out archived threads by default ... but that doesn't explain why a "normal" search only comes up with one of the two questions:

    https://support.mozilla.org/en-US/search?esab=a&q=ancestry

    If you do an Advanced search and screen out Archived threads, then you only get two results: https://support.mozilla.org/en-US/search?q=ancestry&is_archived=-1&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 *[/questions/975948] *[/questions/981210] I'm guessing that a normal search screens out archived threads by default ... but that doesn't explain why a "normal" search only comes up with one of the two questions: https://support.mozilla.org/en-US/search?esab=a&q=ancestry *[/questions/981210]
  5. I did a bugzilla search for SUMO search bugs in 2013 and found:

    It was resolved "INVALID" by Kadir Topal so maybe a "normal" SUMO search is filtering out threads with zero helpful votes as well as threads that are archived.

    Here's an Advanced search on "ancestry" filtering for threads that are NOT archived, that have helpful votes:

    https://support.mozilla.org/en-US/search?q=ancestry&is_archived=-1&has_helpful=1&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2

    The singe result you get is the one thread (same as a "Normal" search):

    I did a [https://bugzilla.mozilla.org/buglist.cgi?list_id=9032010&resolution=---&resolution=FIXED&resolution=INVALID&resolution=WONTFIX&resolution=DUPLICATE&resolution=WORKSFORME&chfieldto=Now&query_format=advanced&chfieldfrom=2013-01-01&component=Search&product=support.mozilla.org bugzilla search for SUMO search bugs in 2013] and found: *[https://bugzilla.mozilla.org/show_bug.cgi?id=670428 Bug 670428 - Include threads with zero helpful votes in elastic search results] It was resolved "INVALID" by Kadir Topal so maybe a "normal" SUMO search is filtering out threads with zero helpful votes as well as threads that are archived. Here's an Advanced search on "ancestry" filtering for threads that are NOT archived, that have helpful votes: https://support.mozilla.org/en-US/search?q=ancestry&is_archived=-1&has_helpful=1&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 The singe result you get is the one thread (same as a "Normal" search): *[/questions/981210]
  6. P.S. The google search https://www.google.com/search?&q=ancestry+site:support.mozilla.org finds five unarchived threads, going by the dates shown on the first results page:

    (Google's search gives you the date the thread was created, which is left out of SUMO searches.)

    I couldn't figure out why SUMO's Advanced search for unarchived threads containing the term "ancestry" only comes up with two results, not five.

    P.S. The google search https://www.google.com/search?&q=ancestry+site:support.mozilla.org finds five unarchived threads, going by the dates shown on the first results page: *support.mozilla.org/questions/981210 ''6 days ago'' *support.mozilla.org/questions/975948 ''Oct 31, 2013'' *support.mozilla.org/questions/967603 ''Aug 10, 2013'' *support.mozilla.org/questions/966995 ''Aug 5, 2013'' *support.mozilla.org/questions/977807 ''Nov 17, 2013'' (Google's search gives you the date the thread was created, which is left out of SUMO searches.) I couldn't figure out why SUMO's Advanced search for unarchived threads containing the term "ancestry" only comes up with two results, not five.

    Modified by AliceWyman on

  7. Posted by cor-el today, in the Forum Guidelines: "Moderator actions" discussion: /forums/contributors/709033?&page=6#post-56321

    cor-el said

    It would be fine if the search results pages would show whether a thread is locked or archived.
    Moderators would benefit from such status info.
    Currently it only shows solved and no reply status.
    This also helps if you search for posts by some one who has posted duplicate or spam threads.
    Now you have to open each thread to see if they are already locked in case locked threads are included in the search.
    Posted by cor-el today, in the Forum Guidelines: "Moderator actions" discussion: [/forums/contributors/709033?&page=6#post-56321] ''cor-el said'' <blockquote> It would be fine if the search results pages would show whether a thread is locked or archived.<br /> Moderators would benefit from such status info.<br /> Currently it only shows solved and no reply status.<br /> This also helps if you search for posts by some one who has posted duplicate or spam threads.<br /> Now you have to open each thread to see if they are already locked in case locked threads are included in the search. </blockquote>
  8. Yeah, lot's of filtering is going on in different parts of the site. It's all documented in the filters section of this document: http://kitsune.readthedocs.org/en/latest/searchchapter.html#searching-on-the-site

    Do you see an issue with the filtering? Most threads are filtered out because they are archived. They are archived because they are old, and we considered old threads to do more harm than good. After all, they will often apply to Firefox versions that are more than 3 versions outdated.

    Some threads might not be displayed, because they are not marked as solved and don't have a "helpful" vote. We could talk about how useful it is to filter those out indeed.

    Yeah, lot's of filtering is going on in different parts of the site. It's all documented in the filters section of this document: http://kitsune.readthedocs.org/en/latest/searchchapter.html#searching-on-the-site Do you see an issue with the filtering? Most threads are filtered out because they are archived. They are archived because they are old, and we considered old threads to do more harm than good. After all, they will often apply to Firefox versions that are more than 3 versions outdated. Some threads might not be displayed, because they are not marked as solved and don't have a "helpful" vote. We could talk about how useful it is to filter those out indeed.
  9. One more thing to add, we haven't changed the search feature in quite some time as far as I can remember.

    One more thing to add, we haven't changed the search feature in quite some time as far as I can remember.
  10. AliceWyman said

    If you do an Advanced search and screen out Archived threads, then you only get two results: https://support.mozilla.org/en-US/search?q=ancestry&is_archived=-1&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2

    AliceWyman said

    P.S. The google search https://www.google.com/search?&q=ancestry+site:support.mozilla.org finds five unarchived threads, going by the dates shown on the first results page: (Google's search gives you the date the thread was created, which is left out of SUMO searches.)

    I couldn't figure out why SUMO's Advanced search for unarchived threads containing the term "ancestry" only comes up with two results, not five.

    Can you figure that one out, Kadir?

    ''AliceWyman [[#post-56183|said]]'' <blockquote> If you do an Advanced search and screen out Archived threads, then you only get two results: https://support.mozilla.org/en-US/search?q=ancestry&is_archived=-1&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 *[/questions/975948] *[/questions/981210] </blockquote> ''AliceWyman [[#post-56186|said]]'' <blockquote> P.S. The google search https://www.google.com/search?&q=ancestry+site:support.mozilla.org finds five unarchived threads, going by the dates shown on the first results page: *support.mozilla.org/questions/981210 ''6 days ago'' *support.mozilla.org/questions/975948 ''Oct 31, 2013'' *support.mozilla.org/questions/967603 ''Aug 10, 2013'' *support.mozilla.org/questions/966995 ''Aug 5, 2013'' *support.mozilla.org/questions/977807 ''Nov 17, 2013'' (Google's search gives you the date the thread was created, which is left out of SUMO searches.) <br><br> I couldn't figure out why SUMO's Advanced search for unarchived threads containing the term "ancestry" only comes up with two results, not five. </blockquote> Can you figure that one out, Kadir?
  11. I was wondering, why I was getting wildly different results, but I think I understand now. It seems like the labels on advanced search are misleading.

    There are 3 settings for filters:

    "yes" means: Only include questions that match this filter. In the cases of "archived": Only show me archived threads.
    "no" means: Only show me questions that do not match this filter (pretty much what you'd expect)
    "don't filter" means: Show me questions that do and do not match this filter, eg. questions that are archived and those that are not.

    By default the search will assume that all filters are set to "don't filter".

    That said, the actual issue here is that "ancestry" and "ancestry.com" are different words for our search engine, while Google considers both cases. You can try it out with a search for ancestry.com

    I filed a bug for that issue: https://bugzilla.mozilla.org/show_bug.cgi?id=957658

    I was wondering, why I was getting wildly different results, but I think I understand now. It seems like the labels on advanced search are misleading. There are 3 settings for filters: '''"yes"''' means: Only include questions that match this filter. In the cases of "archived": Only show me archived threads. <br> '''"no"''' means: Only show me questions that do not match this filter (pretty much what you'd expect)<br> '''"don't filter"''' means: Show me questions that do and do not match this filter, eg. questions that are archived and those that are not. By default the search will assume that all filters are set to "don't filter". That said, the actual issue here is that "ancestry" and "ancestry.com" are different words for our search engine, while Google considers both cases. You can try it out with a search for [https://support.mozilla.org/en-US/search?q=ancestry.com&is_archived=-1&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 ancestry.com] I filed a bug for that issue: https://bugzilla.mozilla.org/show_bug.cgi?id=957658

    Modified by Kadir Topal on

  12. Note that there are cases where people leave out the space after a dot and the forum software linkifies such text as an URL. The word .it is a case where you see this a lot.

    There is also a problem with searching for words with a hyphen in it. I.e. you cant search for a pref like full-screen-api prefs (full-screen-api.enabled) or other (common) words with a hyphen in it. Each single word can be quite common, but the combination makes them a unique word.

    So if I do a search for a word with a hyphen or dot in it then I expect to find results with such a full word combination and not with all the parts that make up this word.

    Note that there are cases where people leave out the space after a dot and the forum software linkifies such text as an URL. The word .it is a case where you see this a lot. There is also a problem with searching for words with a hyphen in it. I.e. you cant search for a pref like full-screen-api prefs (full-screen-api.enabled) or other (common) words with a hyphen in it. Each single word can be quite common, but the combination makes them a unique word. So if I do a search for a word with a hyphen or dot in it then I expect to find results with such a full word combination and not with all the parts that make up this word.

    Modified by cor-el on

  13. Kadir, Thanks for getting to the bottom of this.

    I chose this example not because I was necessarily looking for a TLD but because the search feature is not too useful although this seemed something it should have been able to handle.

    Hey if we had the ability to search for a quoted string we could solve the TLD issue and try other simple and presumably common searches & have them return meaningful results from the aaq forum.

    Trying force SUMO to find TLDs instead of words sounds difficult, and may it not also make results worse.

    Of course contributors will have long since learnt to use Google instead. It is the forum users & silent visitors that will get caught out & expect something useful. Some comparative examples with Sumo, Sumo Advanced & Google results, with number of hits (1000 is bad it maxed out).

    Lets remember many users will expect the Sumo search to work like Google and will be used to adding more terms or quoting strings to refine results. Our search will to them look broken. Add just a second word and the results are a mess and even longer but with less true hits. Contributors will understand why this happens and just use Google. Users will just give up.

    (Yes I can see the

    • Advantage of filtering the basic search results to show helpful answers.
    • Disadvantage of boolean search; or to a minor extent, string search in some cases )

    I also see Google results are far from perfect, but compared with Sumo results I doubt anyone really tries to use the Sumo search unless there is no alternative.

    I would not say it seems likely to meet the SLA http://kitsune.readthedocs.org/en/latest/sla.html

    Search availability 
    Search should work and return useful results.
    Kadir, Thanks for getting to the bottom of this. I chose this example not because I was necessarily looking for a TLD but because the search feature is not too useful although this seemed something it should have been able to handle. Hey if we had the ability to search for a quoted string we could solve the TLD issue and try other simple and presumably common searches & have them return meaningful results from the aaq forum. Trying force SUMO to find TLDs instead of words sounds difficult, and may it not also make results worse. Of course contributors will have long since learnt to use Google instead. It is the forum users & silent visitors that will get caught out & expect something useful. Some comparative examples with Sumo, Sumo Advanced & Google results, with number of hits (1000 is bad it maxed out). Lets remember many users will expect the Sumo search to work like Google and will be used to adding more terms or quoting strings to refine results. Our search will to them look broken. Add just a second word and the results are a mess and even longer but with less true hits. Contributors will understand why this happens and just use Google. Users will just give up. (Yes I can see the * Advantage of filtering the basic search results to show helpful answers. * Disadvantage of boolean search; or to a minor extent, string search in some cases ) I also see Google results are far from perfect, but compared with Sumo results I doubt anyone really tries to use the Sumo search unless there is no alternative. I would not say it seems likely to meet the SLA http://kitsune.readthedocs.org/en/latest/sla.html ''Search availability <br />Search should work and return useful results.''
  14. Update We finally got a string / phrase search and boolean operators in Sumo advanced search, at the end of 2014. But not until after the Google search was made useless by removing yet more indexing.

    • See Change in support forum threads indexed by Google or other search engines: Only solved threads will be shown /forums/contributors/710806

    The changes will make the figures in the links under totally different


    Should anyone wish to discuss search results, here are some examples.

    Comparative search results. Number of hits.
    Row Search Term S-ord S-ad . . . Go G”Q”
    #<template>abc d
    01 ancestry.com001143 1.6k 1.5k
    02 ancestry001 095 1.6k 1.6k
    03 facebook3571000 173k 173k
    04 facebook.com022 374173k 173k
    05 babylon toolbar1000 1000 2.8k 1.3k
    06 babylon018 928 5.0k 5.0k
    07 toolbar1000 1000 82k 74k
    08 bookmarks10001000101k 120k
    09 bookmarks toolbar1000100016.7k 12.3k
    10 conduit493507k 7k
    11 conduit plugin100010003.6k 721
    12 Flashplayer Crash7421000147k 5.7k
    13 plugin10001000160k 160k
    14 crash642100058k 58k
    15 FlashPlayer RealPlayer21710004.3k Nill
    16 New Tab1000100070k 52k
    17 <topcrash>27610003k 3k
    18 full-screen-api10001000894 14
    # <crash-sig>010601 01

    Google may not be perfect on Sumo searches, but Sumo search sure could benefit from some improvements.

    One of the hard coded 'crash stats searches actually works most just max out with any proper hits buried. Many reasonable searches form the forum also max out. I know we have metrics on actual search terms used. Does anyone ever looked at the searches tried and the check whether the results are useful.

    Update We finally got a string / phrase search and boolean operators in Sumo advanced search, at the end of 2014. But not until after the Google search was made useless by removing yet more indexing. *See ''Change in support forum threads indexed by Google or other search engines: Only solved threads will be shown'' </ br> [/forums/contributors/710806] The changes will make the figures in the links under totally different --------------- Should anyone wish to discuss search results, here are some examples. {| |+ Comparative search results. Number of hits. !Row !! Search Term!! S-ord !! S-ad . . .!! Go !! G”Q” |- |#||<template>||[/en-US/search?esab&q=XXX a]||[https://support.mozilla.org/en-US/search?q=xxxx&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 b]||[https://www.google.com/search?&q=XXXX+site:support.mozilla.org c] ||[https://www.google.com/search?&q=%22XXXX%22+site:support.mozilla.org d] |- | 01 ||ancestry''.''com||[/en-US/search?esab&q=ancestry 001]||[https://support.mozilla.org/en-US/search?q=ancestry.com&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 143 ]||[https://www.google.com/search?&q=ancestry.com+site:support.mozilla.org 1.6k ] ||[https://www.google.com/search?&q=%22ancestry.com%22+site:support.mozilla.org 1.5k ] |- | 02 ||ancestry||[/en-US/search?esab&q=ancestry 001 ]||[https://support.mozilla.org/en-US/search?q=ancestry&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 095 ]||[https://www.google.com/search?&q=ancestry+site:support.mozilla.org 1.6k ] ||[https://www.google.com/search?&q=%22ancestry%22+site:support.mozilla.org 1.6k ] |- | 03 ||facebook||[/en-US/search?esab&q=facebook 357]||[https://support.mozilla.org/en-US/search?q=facebook&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000 ]||[https://www.google.com/search?&q=facebook+site:support.mozilla.org 173k ] ||[https://www.google.com/search?&q=%22facebook%22+site:support.mozilla.org 173k] |- | 04 ||facebook''.''com||[/en-US/search?esab&q=facebook.com 022 ]||[https://support.mozilla.org/en-US/search?q=facebook.com&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 374]||[https://www.google.com/search?&q=facebook.com+site:support.mozilla.org 173k ] ||[https://www.google.com/search?&q=%22facebook.com%22+site:support.mozilla.org 173k ] |- | 05 ||babylon toolbar||[/en-US/search?esab&q=babylon%20toolbar 1000 ]||[https://support.mozilla.org/en-US/search?q=babylon%20toolbar&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000 ]||[https://www.google.com/search?&q=babylon%20toolbar+site:support.mozilla.org 2.8k ] ||[https://www.google.com/search?&q=%22babylon%20toolbar%22+site:support.mozilla.org 1.3k] |- | 06 ||babylon||[/en-US/search?esab&q=babylon 018 ]||[https://support.mozilla.org/en-US/search?q=babylon&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 928 ]||[https://www.google.com/search?&q=babylon+site:support.mozilla.org 5.0k ] ||[https://www.google.com/search?&q=%22babylon%22+site:support.mozilla.org 5.0k ] |- | 07 ||toolbar||[/en-US/search?esab&q=toolbar 1000 ]||[https://support.mozilla.org/en-US/search?q=toolbar&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000 ]||[https://www.google.com/search?&q=toolbar+site:support.mozilla.org 82k ] ||[https://www.google.com/search?&q=%22toolbar%22+site:support.mozilla.org 74k ] |- | 08 ||bookmarks||[/en-US/search?esab&q=bookmarks 1000]||[https://support.mozilla.org/en-US/search?q=bookmarks&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000]||[https://www.google.com/search?&q=bookmarks+site:support.mozilla.org 101k] ||[https://www.google.com/search?&q=%22bookmarks%22+site:support.mozilla.org 120k] |- | 09 ||bookmarks toolbar||[/en-US/search?esab&q=bookmarks%20toolbar 1000]||[https://support.mozilla.org/en-US/search?q=bookmarks%20toolbar&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000]||[https://www.google.com/search?&q=bookmarks%20toolbar+site:support.mozilla.org 16.7k] ||[https://www.google.com/search?&q=%22bookmarks%20toolbar%22+site:support.mozilla.org 12.3k] |- | 10 ||conduit||[/en-US/search?esab&q=conduit 49]||[https://support.mozilla.org/en-US/search?q=conduit&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 350]||[https://www.google.com/search?&q=conduit+site:support.mozilla.org 7k] ||[https://www.google.com/search?&q=%22conduit%22+site:support.mozilla.org 7k] |- | 11 ||conduit plugin||[/en-US/search?esab&q=conduit%20plugin 1000]||[https://support.mozilla.org/en-US/search?q=conduit%20plugin&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000]||[https://www.google.com/search?&q=conduit%20plugin+site:support.mozilla.org 3.6k] ||[https://www.google.com/search?&q=%22conduit%20plugin%22+site:support.mozilla.org 721] |- | 12 ||Flashplayer Crash||[/en-US/search?esab&q=Flashplayer%20Crash 742]||[https://support.mozilla.org/en-US/search?q=Flashplayer%20Crash&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000]||[https://www.google.com/search?&q=Flashplayer%20Crash+site:support.mozilla.org 147k] ||[https://www.google.com/search?&q=%22Flashplayer%20Crash%22+site:support.mozilla.org 5.7k] |- | 13 ||plugin||[/en-US/search?esab&q=plugin 1000]||[https://support.mozilla.org/en-US/search?q=plugin&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000]||[https://www.google.com/search?&q=plugin+site:support.mozilla.org 160k] ||[https://www.google.com/search?&q=%22plugin%22+site:support.mozilla.org 160k] |- | 14 ||crash||[/en-US/search?esab&q=crash 642]||[https://support.mozilla.org/en-US/search?q=crash&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000]||[https://www.google.com/search?&q=crash+site:support.mozilla.org 58k] ||[https://www.google.com/search?&q=%22crash%22+site:support.mozilla.org 58k] |- | 15 ||FlashPlayer RealPlayer||[/en-US/search?esab&q=FlashPlayer%20RealPlayer 217]||[https://support.mozilla.org/en-US/search?q=FlashPlayer%20RealPlayer&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000]||[https://www.google.com/search?&q=FlashPlayer%20RealPlayer+site:support.mozilla.org 4.3k] ||[https://www.google.com/search?&q=%22FlashPlayer%20RealPlayer%22+site:support.mozilla.org Nill] |- | 16 ||New Tab||[/en-US/search?esab&q=New%20Tab 1000]||[https://support.mozilla.org/en-US/search?q=New%20Tab&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000]||[https://www.google.com/search?&q=New%20Tab+site:support.mozilla.org 70k] ||[https://www.google.com/search?&q=%22New%20Tab%22+site:support.mozilla.org 52k] |- | 17 ||<topcrash><!--js::GCMarker::processMarkStackTop(js::SliceBudget&)-->||[/en-US/search?esab&q=js::GCMarker::processMarkStackTop(js::SliceBudget&) 276]||[https://support.mozilla.org/en-US/search?q=js::GCMarker::processMarkStackTop(js::SliceBudget&)&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000]||[https://www.google.com/search?&q=js::GCMarker::processMarkStackTop(js::SliceBudget&)+site:support.mozilla.org 3k] ||[https://www.google.com/search?&q=%22js::GCMarker::processMarkStackTop(js::SliceBudget&)%22+site:support.mozilla.org 3k] |- | 18 ||full-screen-api||[/en-US/search?esab&q=full-screen-api 1000]||[https://support.mozilla.org/en-US/search?q=full-screen-api&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 1000]||[https://www.google.com/search?&q=full-screen-api+site:support.mozilla.org 894] ||[https://www.google.com/search?&q=%22full-screen-api%22+site:support.mozilla.org 14] |- | # ||<crash-sig>||[/en-US/search?esab&q=CoreFoundation@0x11e8b 01]||[https://support.mozilla.org/en-US/search?q=CoreFoundation@0x11e8b&num_voted=0&num_votes=&asked_by=&answered_by=&q_tags=&created=0&created_date=&updated=0&updated_date=&sortby=0&a=1&w=2 06]||[https://www.google.com/search?&q=CoreFoundation@0x11e8b+site:support.mozilla.org 01] ||[https://www.google.com/search?&q=%22CoreFoundation@0x11e8b%22+site:support.mozilla.org 01] |- |} Google may not be perfect on Sumo searches, but Sumo search sure could benefit from some improvements. One of the hard coded '''crash stats'' searches actually works most just max out with any proper hits buried. Many reasonable searches form the forum also max out. I know we have metrics on actual search terms used. Does anyone ever looked at the searches tried and the check whether the results are useful.

    Modified by John99 on

  15. yeah, the search capabilities are rather disappointing still. just today i was failing in finding a old answer of mine which would have had a useful screenshot attached (it was apparently already archived, so no help via google). because it's not possible to use search operators like AND & NOT, with the prospect of sifting through hundreds of posts i quickly gave up - for a matter which should just take a few seconds otherwise...

    yeah, the search capabilities are rather disappointing still. just today i was failing in finding a old answer of mine which would have had a useful screenshot attached (it was apparently already archived, so no help via google). because it's not possible to use search operators like AND & NOT, with the prospect of sifting through hundreds of posts i quickly gave up - for a matter which should just take a few seconds otherwise...
  16. Thanks for feedback on search. It's an area that is really important, but also really hard to get right. Just a comment: The number of hits are actually not that important, most people won't read past the first few entries.

    I understand that we need better capabilities for advanced search though. Please add use cases to this Etherpad: https://etherpad.mozilla.org/sumo-advanced-search-use-cases We'll talk about the roadmap again in a few weeks and can take that Etherpad as the base for a discussion of whether we should put search improvements on the roadmap.

    Thanks for feedback on search. It's an area that is really important, but also really hard to get right. Just a comment: The number of hits are actually not that important, most people won't read past the first few entries. I understand that we need better capabilities for advanced search though. Please add use cases to this Etherpad: https://etherpad.mozilla.org/sumo-advanced-search-use-cases We'll talk about the roadmap again in a few weeks and can take that Etherpad as the base for a discussion of whether we should put search improvements on the roadmap.
  17. hello kadir, thanks for your answer - my particular wishes are already covered in the etherpad :-)

    maybe another search related issue - with the recent java plugin update there will be an error message for many use-cases that have worked before (this might also have driven up visitor numbers during the past weeks). many people will not be able to determine that this error is not coming from firefox, but from the plugin itself.

    alice has thankfully edited our KB article accordingly, but when you search for the wording of the error message, the relevant article doesn't show up on the first page: https://support.mozilla.org/en-US/forums/contributors/709912?last=56765#post-56762

    hello kadir, thanks for your answer - my particular wishes are already covered in the etherpad :-) maybe another search related issue - with the recent java plugin update there will be an error message for many use-cases that have worked before (this might also have driven up visitor numbers during the past weeks). many people will not be able to determine that this error is not coming from firefox, but from the plugin itself. alice has thankfully edited our KB article accordingly, but when you search for the wording of the error message, the relevant article doesn't show up on the first page: https://support.mozilla.org/en-US/forums/contributors/709912?last=56765#post-56762
  18. Kadir Topal said

    The number of hits are actually not that important, most people won't read past the first few entries.

    You could try looking at the first page of each of those results. By adding links we know if anyone should do comparisons what they are looking at.

    Nearly always Google is far better.

    Additionally I believe you have data on the real searches conducted.

    • Did you actually look at the search terms used in practice, and whether for say the top 20 searches the results are of any use ?

    Of course that may be influenced by knowledgeable contributors deliberately restricting searches to something that may work, rather than what they would like to try to find.

    I added the numbers for a few reasons

    • It clearly demonstrates that many searches maxing out at 1000.
      That is a big problem.
      • Maxed out searches certainly may not be combined with any other search term or the results will get even worse.
      • As you say who looks at more than the first page or at most the first three pages. Especially if even those results are not the expected proper hits.
    • In many cases users may try to narrow down a search by adding more terms. Clearly we know the opposite will and does happen.
    • Google does have disadvantages
      • does not look at sub domains
      • unable to restrict to specific AAQ forum, or KB articles
      • picks up details from the added System Info
      • unable to index some areas

    Just a thought could we leverage the Google searches by putting an identifier string in AAQ, KB and Contribs as a label within the indexed post or article. Anything likely to be unique.

    Say SUMOAAQ2014, SUMOKB2014, SUMOCONTRIB2014 Those wishing to do an advanced search could then easily add that as a search term. It may be a rather naive idea, but it is simple to implement and I can not think of any disadvantage or reason why it should not work.

    The system information being indexed is probably not a problem, I am guessing it only really shows up lower down in any pages of hits.

    ''Kadir Topal [[#post-56782|said]]'' <blockquote> The number of hits are actually not that important, most people won't read past the first few entries. </blockquote> You could try looking at the first page of each of those results. By adding links we know if anyone should do comparisons what they are looking at. Nearly always Google is far better. <br /> Additionally I believe you have data on the real searches conducted. *Did you actually look at the search terms used in practice, and whether for say the top 20 searches the results are of any use ? Of course that may be influenced by knowledgeable contributors deliberately restricting searches to something that may work, rather than what they would like to try to find. <u>I added the numbers for a few reasons</u> * It clearly demonstrates that many searches maxing out at 1000. <br />That is a big problem. **Maxed out searches certainly may not be combined with any other search term or the results will get even worse. **As you say who looks at more than the first page or at most the first three pages. Especially if even those results are not the expected proper hits. * In many cases users may try to narrow down a search by adding more terms. Clearly we know the opposite will and does happen. * Google does have disadvantages ** does not look at sub domains ** unable to restrict to specific AAQ forum, or KB articles ** picks up details from the added System Info ** unable to index some areas Just a thought could we leverage the Google searches by putting an identifier string in AAQ, KB and Contribs as a label within the indexed post or article. Anything likely to be unique. Say SUMOAAQ2014, SUMOKB2014, SUMOCONTRIB2014 Those wishing to do an advanced search could then easily add that as a search term. It may be a rather naive idea, but it is simple to implement and I can not think of any disadvantage or reason why it should not work. The system information being indexed is probably not a problem, I am guessing it only really shows up lower down in any pages of hits.
  19. Actually, we looked at our search terms last week. Joni even combined them to find out what the most often looked for topics are: https://docs.google.com/a/mozilla.com/spreadsheet/ccc?key=0AtmG206rGF1odFR1RHlzOEl3MlQ5aWZCUnp1WDNQX2c#gid=0 (It's currently protected, but I'll ask her to open it up).

    Improving search is not as straight forward as it seems. While "update" might be the most often used search term, it still accounts for only ~2% of searches. And even then, it's used in many different variations, eg. "updated browser", "update Firefox", "updating to latest version". So, you might be tempted to improve search for "update", but what you actually want to do is improve search for "[some text] update (stemmed, to include updating, updates, updated, etc) [some text]" And that is not straight forward at all.

    That said, it's an important thing to get right and one of the areas that Joni will be looking at. Expect great things from her :)

    Actually, we looked at our search terms last week. Joni even combined them to find out what the most often looked for topics are: https://docs.google.com/a/mozilla.com/spreadsheet/ccc?key=0AtmG206rGF1odFR1RHlzOEl3MlQ5aWZCUnp1WDNQX2c#gid=0 (It's currently protected, but I'll ask her to open it up). Improving search is not as straight forward as it seems. While "update" might be the most often used search term, it still accounts for only ~2% of searches. And even then, it's used in many different variations, eg. "updated browser", "update Firefox", "updating to latest version". So, you might be tempted to improve search for "update", but what you actually want to do is improve search for "[some text] update (stemmed, to include updating, updates, updated, etc) [some text]" And that is not straight forward at all. That said, it's an important thing to get right and one of the areas that Joni will be looking at. Expect great things from her :)
  20. Questions about the Data

    1. quotation marks Does the data show users attempting to use quotes.
      Or is that not available because for instance they are stripped out before you collect the results.
    2. tasks In the data are the tasks a guess at what may have been looked for or are they the frequently entered strings eg
    update Firefox	17177	1.78%	
    check for updates, manual update, automatic update, disabling updates		
    

    From your comments I imagine many searches contain multiple terms; not single words. That is where we start very quickly to produce bad hits because Sumo ORs the terms and buries the required hits whilst maxing out.

    Thanks for the link to the data.

    Back to the fact that being able to search

    • quoted strings
    • boolean [term 1] AND [term 2]

    May possibly help in getting good results. As I said before I think many users will be used to using Google and Google by default ANDs search terms. I imagine many will also be aware of Google's ability to use a quoted string. Right now anyone using the term update in the search terms makes the results even worse with every additional word added. (Other than possibly for specially curated search terms if KB articles are prioritised)

    Questions about the Data # '''quotation marks''' Does the data show users attempting to use quotes. <br />Or is that not available because for instance they are stripped out before you collect the results. # '''tasks''' In the data are the ''tasks'' a guess at what may have been looked for or are they the frequently entered strings eg update Firefox 17177 1.78% check for updates, manual update, automatic update, disabling updates From your comments I imagine many searches contain multiple terms; not single words. That is where we start very quickly to produce bad hits because Sumo ORs the terms and buries the required hits whilst maxing out. Thanks for the link to the data. Back to the fact that being able to search *quoted strings *boolean [term 1] AND [term 2] May possibly help in getting good results. As I said before I think many users will be used to using Google and Google by default ANDs search terms. I imagine many will also be aware of Google's ability to use a quoted string. Right now anyone using the term update in the search terms makes the results even worse with every additional word added. (Other than possibly for specially curated search terms if KB articles are prioritised)
  1. 1
  2. 2