Autocomplete/search results need to be tuned to be maximally intuitive

[kltm]

This covers a lot or ground, but the idea is that autocomplete and search results should be very intuitive for the majority of users.

The bulk of the solution may be connected to #131, and the simplified constructs around it. The rest will likely be tuning the new fields/parameters to get the desired "intuitive" results.

[kltm]

Once the engineering bits of #131 are handled, we'll need a consistent target list (all desired outcomes are possible) to work with and try to achieve it with tuning.

[kltm]

Val's cdc2 vs. cdc20 is another good example.

[cmungall]

Here's the main issue. There are dozens of genes called foo. The labels is unique (more or less...) within a species, but shared across species.

A user typing 'foo' sees all these in a random order. Sometimes you can guess what the species is by whether it's foo Foo or FOO. But this is rubbish. This will get worse as we add more organisms. The organism most people want (e.g. human) may be buried beyond the available list.

1. Add species in parentheses. Can be done upstream in Golr loading (I think my preference) or display/AC logic. Also with the upstream method I think savvy users can streamline their search (e.g. "foo mus")
2. There should be an ordering that prioritizes humans above e.g. armadillos. Not sure how best to do this. We could add a 'relevancy' field to the schema. Golr loading would take care of filling this in. This would be generally useful. E.g. in term AC GO would be more relevant than others.

[kltm]

1 is essentially #24.
2 will require some fairly low-level goosing, likely a form of #131 and #132.

[kltm]

This is fundamentally related to the search reworking in berkeleybop/bbop-js#16.

[kltm]

More good examples in:

http://jira.geneontology.org/browse/GO-633

[cmungall]

Should we split this ticket to get separate specs for bioentities vs classes? There are similarities and differences. For bioentities the species issue is paramount (see comments above, the Noctua AC is a massive improvement).

The class requirements have not been stated, first pass:

• obsolete classes should be ranked lower
• synonyms should probably be ranked higher than they are, this is partly a nudge to accommodate the fact that GO is overly formal in its nomenclature at times. E.g. try getting "apoptosis" in AC - you can't. (hint: GO calls this "apoptotic process" and uses "apoptosis" as a synonym, contrary to normal biologist usage)
• the primary ontology (GO for GO) should rank above others. Again, try "apoptosis" - none of the top ACs are in GO