Search results sorting in Oxygen

Note that content weight boost does not apply to Oxygen content, because for Relevanssi, it’s not post content, it’s custom field content (because it comes from the ct_builder_shortcodes custom field). That might explain the low positions.

Can you please show me screenshots of the debug screen with the different custom field indexing settings?

Hi, Mikko. Good to know about the weight boost. Makes sense. What’s strange is that the position isn’t low if I index VISIBLE fields.

This is what I get if I set indexing to SOME fields and specify ct_builder_shortcodes: https://postimg.cc/sG7cfXqZ

But the page does not show at the top spot.

If I change the indexing to VISIBLE fields, I get this in debug: https://postimg.cc/Q9Rmr543

And the page does show at the top spot.

Thanks for any insights.

That’s interesting; that should not make any difference. Is the post relevancy score the same in both cases? Is there any difference in that? You can see that easily in the Relevanssi admin search (Dashboard > Admin search).

So if I set indexing custom fields to VISIBLE, the post relevancy is 7406.79

If I set it to SOME and specify ct_builder_shortcodes, the relevancy score 352.57

What the keyword is you’re searching for?

“Next-Level Technology for Clinicians in Every Department” – it’s a headline on a clinical solutions page.

It almost looks like it’s a weighting issue. When I look at the results, everything above my clinical solutions page is a post of some kind…podcast, blog post, etc that is using wordpress content editor. the Oxygen builder content that is in the custom field would appear to be weighted less.

But what I can’t understand is why that would change when i include VISIBLE custom fields in the index, and my clinical solution shoots to the top of the results.

I’ve also tried it with Boost Exact Matches on and off, with no change.

Yeah, I don’t understand this either. Let’s study further. What does this print out?

add_filter( 'relevanssi_match', 'rlv_test_match' );
function rlv_test_match( $match ) {
    if ( xxx == $match->doc ) {
        var_dump( $match );
    }
    return $match;
}

Replace xxx with the post ID of this post and then run a search with this both ways. Is there a difference?

indexing SOME custom fields (ct_builder_shortcodes)

object(stdClass)#2912 (22) { ["doc"]=> string(3) "398" ["term"]=> string(5) "level" ["term_reverse"]=> string(5) "level" ["content"]=> string(1) "0" ["title"]=> string(1) "0" ["comment"]=> string(1) "0" ["tag"]=> string(1) "0" ["link"]=> string(1) "0" ["author"]=> string(1) "0" ["category"]=> string(1) "0" ["excerpt"]=> string(1) "0" ["taxonomy"]=> string(1) "0" ["customfield"]=> string(1) "1" ["mysqlcolumn"]=> string(1) "0" ["taxonomy_detail"]=> string(0) "" ["customfield_detail"]=> string(0) "" ["mysqlcolumn_detail"]=> string(0) "" ["type"]=> string(4) "post" ["item"]=> string(1) "0" ["tf"]=> float(1) ["taxonomy_score"]=> float(0) ["weight"]=> float(44.0707360822) } object(stdClass)#3391 (22) { ["doc"]=> string(3) "398" ["term"]=> string(10) "clinicians" ["term_reverse"]=> string(10) "snaicinilc" ["content"]=> string(1) "0" ["title"]=> string(1) "0" ["comment"]=> string(1) "0" ["tag"]=> string(1) "0" ["link"]=> string(1) "0" ["author"]=> string(1) "0" ["category"]=> string(1) "0" ["excerpt"]=> string(1) "0" ["taxonomy"]=> string(1) "0" ["customfield"]=> string(1) "3" ["mysqlcolumn"]=> string(1) "0" ["taxonomy_detail"]=> string(0) "" ["customfield_detail"]=> string(0) "" ["mysqlcolumn_detail"]=> string(0) "" ["type"]=> string(4) "post" ["item"]=> string(1) "0" ["tf"]=> float(3) ["taxonomy_score"]=> float(0) ["weight"]=> float(132.21202127) } object(stdClass)#3929 (22) { ["doc"]=> string(3) "398" ["term"]=> string(10) "department" ["term_reverse"]=> string(10) "tnemtraped" ["content"]=> string(1) "0" ["title"]=> string(1) "0" ["comment"]=> string(1) "0" ["tag"]=> string(1) "0" ["link"]=> string(1) "0" ["author"]=> string(1) "0" ["category"]=> string(1) "0" ["excerpt"]=> string(1) "0" ["taxonomy"]=> string(1) "0" ["customfield"]=> string(1) "3" ["mysqlcolumn"]=> string(1) "0" ["taxonomy_detail"]=> string(0) "" ["customfield_detail"]=> string(0) "" ["mysqlcolumn_detail"]=> string(0) "" ["type"]=> string(4) "post" ["item"]=> string(1) "0" ["tf"]=> float(3) ["taxonomy_score"]=> float(0) ["weight"]=> float(132.211988241) } object(stdClass)#4233 (22) { ["doc"]=> string(3) "398" ["term"]=> string(10) "technology" ["term_reverse"]=> string(10) "ygolonhcet" ["content"]=> string(1) "0" ["title"]=> string(1) "0" ["comment"]=> string(1) "0" ["tag"]=> string(1) "0" ["link"]=> string(1) "0" ["author"]=> string(1) "0" ["category"]=> string(1) "0" ["excerpt"]=> string(1) "0" ["taxonomy"]=> string(1) "0" ["customfield"]=> string(1) "1" ["mysqlcolumn"]=> string(1) "0" ["taxonomy_detail"]=> string(0) "" ["customfield_detail"]=> string(0) "" ["mysqlcolumn_detail"]=> string(0) "" ["type"]=> string(4) "post" ["item"]=> string(1) "0" ["tf"]=> float(1) ["taxonomy_score"]=> float(0) ["weight"]=> float(44.0706459486) }

Indexing VISIBLE I get this
object(stdClass)#2900 (22) { ["doc"]=> string(3) "398" ["term"]=> string(5) "level" ["term_reverse"]=> string(5) "level" ["content"]=> string(1) "0" ["title"]=> string(1) "0" ["comment"]=> string(1) "0" ["tag"]=> string(1) "0" ["link"]=> string(1) "0" ["author"]=> string(1) "0" ["category"]=> string(1) "0" ["excerpt"]=> string(1) "0" ["taxonomy"]=> string(1) "0" ["customfield"]=> string(2) "21" ["mysqlcolumn"]=> string(1) "0" ["taxonomy_detail"]=> string(0) "" ["customfield_detail"]=> string(0) "" ["mysqlcolumn_detail"]=> string(0) "" ["type"]=> string(4) "post" ["item"]=> string(1) "0" ["tf"]=> float(21) ["taxonomy_score"]=> float(0) ["weight"]=> float(925.850115451) } object(stdClass)#3405 (22) { ["doc"]=> string(3) "398" ["term"]=> string(10) "clinicians" ["term_reverse"]=> string(10) "snaicinilc" ["content"]=> string(1) "0" ["title"]=> string(1) "0" ["comment"]=> string(1) "0" ["tag"]=> string(1) "0" ["link"]=> string(1) "0" ["author"]=> string(1) "0" ["category"]=> string(1) "0" ["excerpt"]=> string(1) "0" ["taxonomy"]=> string(1) "0" ["customfield"]=> string(2) "63" ["mysqlcolumn"]=> string(1) "0" ["taxonomy_detail"]=> string(0) "" ["customfield_detail"]=> string(0) "" ["mysqlcolumn_detail"]=> string(0) "" ["type"]=> string(4) "post" ["item"]=> string(1) "0" ["tf"]=> float(63) ["taxonomy_score"]=> float(0) ["weight"]=> float(2777.5463867) } object(stdClass)#3976 (22) { ["doc"]=> string(3) "398" ["term"]=> string(10) "department" ["term_reverse"]=> string(10) "tnemtraped" ["content"]=> string(1) "0" ["title"]=> string(1) "0" ["comment"]=> string(1) "0" ["tag"]=> string(1) "0" ["link"]=> string(1) "0" ["author"]=> string(1) "0" ["category"]=> string(1) "0" ["excerpt"]=> string(1) "0" ["taxonomy"]=> string(1) "0" ["customfield"]=> string(2) "63" ["mysqlcolumn"]=> string(1) "0" ["taxonomy_detail"]=> string(0) "" ["customfield_detail"]=> string(0) "" ["mysqlcolumn_detail"]=> string(0) "" ["type"]=> string(4) "post" ["item"]=> string(1) "0" ["tf"]=> float(63) ["taxonomy_score"]=> float(0) ["weight"]=> float(2777.54573631) } object(stdClass)#4290 (22) { ["doc"]=> string(3) "398" ["term"]=> string(10) "technology" ["term_reverse"]=> string(10) "ygolonhcet" ["content"]=> string(1) "0" ["title"]=> string(1) "0" ["comment"]=> string(1) "0" ["tag"]=> string(1) "0" ["link"]=> string(1) "0" ["author"]=> string(1) "0" ["category"]=> string(1) "0" ["excerpt"]=> string(1) "0" ["taxonomy"]=> string(1) "0" ["customfield"]=> string(2) "21" ["mysqlcolumn"]=> string(1) "0" ["taxonomy_detail"]=> string(0) "" ["customfield_detail"]=> string(0) "" ["mysqlcolumn_detail"]=> string(0) "" ["type"]=> string(4) "post" ["item"]=> string(1) "0" ["tf"]=> float(21) ["taxonomy_score"]=> float(0) ["weight"]=> float(925.848247482) }

Ok, so the difference is that with visible fields indexed, the words “level” and “technology” appear 21 times and the words “clinicians” and “department” appear 63 times, and with only the one field indexed, they appear once and three times respectively. That’s why the weights are 21 times bigger.

With only one field indexed, Relevanssi just sees the content you build from the ACF fields in the ct_builder_shortcodes, but if you have all visible fields indexed, Relevanssi reads in both the source ACF fields and the complete content, leading to seeing the same content multiple times.

Part of the problem here is that big part of the weights does not work as expected in Oxygen, because Oxygen does not have post content as such, everything is just custom fields. All that custom field content has by default fairly light weight.

In general it would be best to only index the ct_builder_shortcodes field if all field content is reflected there. The main question here is now what is boosting the other content before the better matches. What is giving weight for the posts higher in the rankings?

The posts are pulling the content from WordPress content editor, not Oxygen custom fields. The page is pulling all its content from Oxygen custom field. So if custom fields are “fairly light weight” by default, that would explain it right?

Yes, that’s all – I thought all content in Oxygen came from the custom fields, but if you have something that’s in post content and got plenty of content weight, that’s all this needs. Drop the content weight to 1, and it’ll be on the same level as the custom fields.

If necessary, here are tools for increasing custom field weight.

Okay, so I definitely will adjust the weights. Thank you for helping me figure that out.

What’s confusing me is why when set to “VISIBLE”, the result weights are 21 times higher for the page. There are no additional fields that would contain those phrases. the only ACF fields I have in place are for whether to show the author’s name. And the wordpress content area is empty on that page.

There are 9 page revisions. The first being empty. Would it be reading page revisions when “VISIBLE” is set?

• This reply was modified 4 years ago by rvturnage.

No, it’s not reading revisions (those are different posts with different post IDs, and separate entities as far as Relevanssi is considered).

It’s hard to say why without knowing what happens in the indexing. This indexing debugging trick might shed some light on it. Maybe it’s some ACF meta fields? I don’t know, and the default debugger doesn’t help here either.