search

monarch-initiative / semsimian

Simple rust implementation of semantic similarity
BSD 3-Clause "New" or "Revised" License
9 stars 5 forks source link

create a set of unit/integration tests that can used in semsimian and also Monarch App #118

Open justaddcoffee opened 9 months ago

justaddcoffee commented 9 months ago

Per convo in A&A Monarch Huddle on Feb 13, 2024

@cmungall says: we should have a bank of test profiles (e.g. tracked in yaml in github) with annotations of what we expect the results to be. Test the framework with this. The UI group can just use the same set

@hrshdhgd can you comment about whether our current semsimian tests can be shared with the Monarch App/UI team via a set of YAML files or some such?

justaddcoffee commented 9 months ago

also @kevinschaper thoughts on this? would this be useful, and where would the test artifacts (YAML or otherwise) live?

cmungall commented 9 months ago

Should we coordinate with pheval?

On Wed, Feb 14, 2024 at 8:37 AM Justin Reese @.***> wrote:

also @kevinschaper https://github.com/kevinschaper thoughts on this? would this be useful, and where would the test artifacts (YAML or otherwise) live?

— Reply to this email directly, view it on GitHub https://github.com/monarch-initiative/semsimian/issues/118#issuecomment-1944196783, or unsubscribe https://github.com/notifications/unsubscribe-auth/AAAMMOP4VWJSQVJR3BUZX63YTTR5FAVCNFSM6AAAAABDIS5KGOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTSNBUGE4TMNZYGM . You are receiving this because you were mentioned.Message ID: @.***>

justaddcoffee commented 9 months ago

Per discussion today on the MWF hackathon, could we refine/clarify here what this test suite should be testing, and what the intention is

If we are testing that semsimian and the Monarch App are getting the correct semantic similarity results given some simple test fixtures, then it seems like this suite belongs in semsimian, and possibly the existing semsimian tests are sufficient.

Possibly though this ticket is about testing edge cases that we observed in the Monarch App when doing an association search: e.g. pairs of terms with fairly generic most informative common ancestors like "Abnormality of the Head" were showing fairly high IC scores. This kind of testing possibly is more to do with the data itself (Phenio) or how we are generating frequencies?

What sorts of things do we want to test here?

caufieldjh commented 9 months ago

See also https://github.com/monarch-initiative/phenio/issues/58