Improve coverage in filters

[andrew-morrison]

Detailed Description

Not all Solr records (manuscripts, people, places, etc) provide at least one value to all the search facets (aka filters). Complete coverage is not possible in all cases, but the more records that are represented in each relevant filter the better the user experience. See discussion in #100.

Context

Under-populated filters can mislead people, if they think it's an accurate summary of what's available, when in reality it's more of a reflection of cataloguing priorities. It might also help future efforts to improve the cataloguing, by making it easy to find records which could be enhanced.

Link to Page

For example, on the People browse page, there are currently 4142 people. But if you use the Role filter on the left-hand side, you may not realise that there are 480 people with no role, who you are excluding. Presumably these people played no part in creating any of the manuscripts, so it made no sense to catalogue them with role attributes, but that subtlety might be lost on end users.

Possible Implementation

I'm just opening this issue for discussion, but there are a few possibilities:

If there are alternative sources of data in the TEI that the XQuery isn't picking up and feeding into Solr then combining these could be considered. For example, the Places browse page will soon have a roles facet. But only a subset of a subset of records (those places that come from orgName elements which have role attribute) will be represented. placeNames have type attributes (in places.xml) but that's something slightly different (e.g. "Region" or "Country" rather than "Owner" or "Binder"). @holfordm would prefer two separate filters, but the way Blacklight works means every new filter will be added not just to the browse page but also the general search, which already has 10 of them.

If things can be inferred from their position in the TEI documents, that might be worth trying. For example, if a persName is in the title of an msItem, is it reasonable to say that person's role is as a subject of the work? Maybe not with 100% confidence, but the filters are only an aid to help find manuscripts, and any such associations won't be displayed on the records themselves.

Another option is to give everything that doesn't have a value derived from the TEI catalogue a value of "Unknown" or "Other".

Or tooltips could be used to explain to users that not everything is represented in every facet. See #101.