Open karafecho opened 1 year ago
See also #123
See also https://github.com/ranking-agent/reasoner-transpiler/issues/72 - do we want to handle this at the transpiler level?
I understand the issue now.
FWIW, my suggestion would be to provide users with control over subclass inferences, especially when a high-level node such as "asthma" or "cardiovascular disorder" is selected. For instance, it might be helpful if a popup appeared when users selected a non-specific node to alert them to the fact that they may want to select a more specific node and perhaps offer suggestions or link out to, e.g., MONDO's hierarchy. Alternatively, you could exclude subclass_of from the initial query results that are returned and then alert users to the option of choosing subclasses. I'm sure there are other options, but I think we need to identify the simplest fix and then fine-tune it over time.
This issue is to suggest that 'subclass of' answers are not returned to users, or are returned in entirety, or are optional. In the example below, the user (me) selected "asthma" as a node, but the top answer was for "occupational asthma", which was not selected. Note that there are many subclasses of asthma, so it is not clear why only one subclass is returned (presumably that was the only one connected to ammonia?). Either provide the user with all known subclasses, or restrict the results to only the selected node, or allow users to select for subclasses (probably the most appealing option).
Note that this is only one example. The issue appears to be general and likely related to the Answer Coalescer or perhaps Name Resolver/Node Normalizer, but it is confusing to users.