improve search

[tgbugs]

Compilation of a number of different comments on search functionality.

1. prioritize entities returned based on their category (eg anatomical entity, disease, protein, ..., biological process, resource (do we even want resources listed here, it is an artifact of using a single scigraph server for scicrunch and ks, this can be fixed)).
2. how to deal with duplicate terms, esp in cases where a parcellation scheme has a similar label (the way I'm planning to ingest parcellation terminologies into the ontology should make it possible to assign a category label to them so those are less of a concern). A related issue is how/whether to allow users to filter results to specific curie prefixes (from Mathew).
3. General layout of the search results page, what should we include?
4. When do we redirect a search query directly to a single page? (I'm feeling lucky style) This is related to point 1 and 2, when do we ignore alternate ids and direct people to the primary entry. Check how wikipedia search/disambiguation does this. (Satra)

[tgbugs]

An additional point is that some search re-direct incorrectly, for example:
https://knowledge-space.org/NeuroKS/index.php/pages/view/elephant redirects to
https://knowledge-space.org/NeuroKS/index.php/pages/view/SCR:003833 because we are not searching synonyms and we are ranking resources too highly in these cases. This also occurs from the search box.

[tgbugs]

Mathew notes that most of his students HATE the layout of the results page for entity search (the main search box). The identifier is especially confusing for them. As a note: Expert users find the identifier one of the most useful features for copy/paste to other programs. Maybe we can put it as the last column?

[tgbugs]

We need a consistent way to allow users to write their own queries extending our automatically generated queries for, entities, literature, and data. @cfitz

[jgrethe]

We have that already implemented for other ES endpoints (e.g. terms and data). James has already implemented the appropriate pass throughs. We will need to move the data to the production ES endpoint though as this has ES X-Pack running that allows us to limit the types of actions users can take (i.e. on production the key that the portal has for the pass through doesn't allow adding or deleting documents and indices.)

[tgbugs]

We need the categories for all the top level to search. Then the trees beneath that then walk the tree back to the defining category.

The hierarchies we need are.

• subClassOf
• part of
• delineates
• possibly 'derived sequence' for gene families
• has phenotype (really phenotype of) for neurons

[tgbugs]

Common field types normalized across datasets should also include the curation status, the data modality, etc. #29