combining scores with outlines

[eroux]

we can approach the score of an outline node using three different angles:

• the score of the node itself, as an independent entity
• the score of the root collection
• the score of the work (text) corresponding to the outline node (for instance we know some sutras are more popular than others, independently of the version we're looking at)

we currently have:

• db_score for the work
• popularity_score for the node
• popularity_score for the work (at least for now, in the new system we'll have to collect that in another way since there will be no independent page for works)
• popularity_score for the root collection

how should we represent that in the index?

[roopeux]

In my previous work I have most often ignored elements that list links to other pages, because they clutter search results with hits that nobody wants.

We don't want to index the leaves of outlines, because the texts will have their direct links. We could consider indexing some volume or section names unless they have their own pages too. If I understood correctly, outlines are embedded in instances which also have their own pages. All this leaves very little to index at all. Enlighten me, I may be missing something.

[eroux]

Well, I'm not sure what you're proposing, do you mean that for instance when people search a sutra they shouldn't find any result on BDRC because we don't index outline nodes? That won't do... but perhaps I don't really understand