Free text filter should be made less prominent or removed in annotation search

[cmungall]

E.g., at least make it only one row. Perhaps make it something you explicitly need to open

There are arguments for removing altogether. Some arguments for keeping it:

1 Users who are not steeped in the data models of the GOC may not know exactly how to find what they are looking for. They may just have a string that matches terms, families, genes. They may even be hazy about the distinctions here. But this may be better served in what amigo calls 'medial search'. Note that medial search for http://amigo.geneontology.org/amigo/medial_search?q=foo now provides links for users to pages like http://amigo.geneontology.org/amigo/search/annotation?q=foo which use the free text filter. I think this needs supported, but this can be done without the box.

2 We used to be able to type solr queries in this box. I loved this, and maybe there's a handful of others who would love this. But better served by an optionally opened 'advanced' search

3 Sometimes it's just faster to get at what you want the unstructured route. Facet drilling can be tedious.

Note that this all applies to annotation search. It's obviously more useful when you are searching through a list of things (terms, genes) than associations

[kltm]

1: This is the use case I worry most about--it was a very common one for A1.x that we were trying to meet.

2: I think something along the lines of Gannet would be more appropriate for that.

For 3, while the free-text can be made less prominent, it can't be removed completely until we switch to Solr 4.x and get searches on facets in--there is currently no other way to remove the facet filter cruft and eliminate enough options to get at some things.

I think that berkeleybop/bbop-js#16, #131, and #132 are the main issues here, with this largely being a symptom (beyond prominence in the display).

[dosumis]

Users have no problem navigating the search on QuickGO. This works as I think AmiGO2 should, by auto-suggest allowing users to find 'objects' - ontology terms or gene products. The results displayed are for things indexed to the object found. In contrast, text refinement allows users to find content based on substrings - IMO a huge step back. As mentioned, this type of search is already possible in AmiGO2, it's just non-obvious. I would make this the only option for primary search - displaying the text refinement box only after a first initial auto-suggest search - and thinking carefully about its placement and labelling.

The auto-suggest could be extended in AmiGO2 so that it includes classes mentioned in annotation extension (although we'd need to think carefully about how these results were displayed - perhaps a separate TAB of hits?).

For 3, while the free-text can be made less prominent, it can't be removed completely until we switch to Solr 4.x and get searches on facets in--there is currently no other way to remove the facet filter cruft and eliminate enough options to get at some things.

But limiting functionality may do less harm than confusing users. (BTW - looking forward to seeing searches on facets. What's the timescale for moving to Solr 4 ?)

[kltm]

@dosumis I largely agree with most of your points (again the other tickets). For the last point though, there's an example issue floating around where you could really get things chugging slow by trying to work the facets without getting access to the filter. Given everything, fixes and refinements should just go along with 4.x.

Another aspect is that there are two user groups, one that knows what they're looking for and one that doesn't. It may be better to split in interfaces a bit more for those--with some locking on more to known items. Right now, things are designed more for the latter case.

[ValWood]

I never did comment on this.
I don't see the positives for free text searching. I'd need to see an example where it was useful to change my mind.

So far, all of the text searches where I have (inadvertently) done a free text search have only retrieved many results which were only indirectly related to my query term. I have documented a couple on the related tickets on the helpdesk tracker.
Free text searching should definitely be less prominent, and it should be much clearer when you are performing a free text search (and what is being searched).

[cmungall]

See also: http://jira.geneontology.org/browse/GO-688

[cmungall]

As a simple initial compromise, the textbox could be hidden at first, opened after the user clicks "filter by free text"

[ValWood]

On 07/01/2015 19:33, Chris Mungall wrote:

See also: http://jira.geneontology.org/browse/GO-688

—
Reply to this email directly or view it on GitHub
#158 (comment).

ah was I using a free text search?...I wassure that I selected a GO ID
from the options which popped up....will try that again

[ValWood]

On 07/01/2015 19:35, Chris Mungall wrote:

As a simple initial compromise, the textbox could be hidden at first,
opened after the user clicks "filter by free text"

—
Reply to this email directly or view it on GitHub
#158 (comment).

Hi Chris,

I definitely wasn't doing a free text search
I started from here
go.geneontology.org/amigo
typed cellular_component and selected
and selected the option(cellular_component GO:0005575)

Amigo then says "forwarding to GO:0005575"

Then I selected "link to all direct and indirect annotations to
"cellular component"

so I'm sure that I shouldn't have any MF or BP annotation in my results?

VAl

[kltm]

I'm still catching up, but http://jira.geneontology.org/browse/GO-688 seems to be a separate issue from this one. My druthers would be to delete this last set of comments. Moreover, the issue likely be upstream in the loader anyways, rather than the search itself (or the search box, etc.).

[cmungall]

Based on observations from @ValWood and comments from @dosumis I have categorized this as a bug that "affects usability"