inspire-next
inspire-next copied to clipboard
inspirehep
Author facet missing results
Current Behavior
When I search for the author Mangano, Michelangelo L. , I get his papers, but his name doesn't appear in the Author name facet
Screenshots (if appropriate):
This is a general issue of author facets. Actually during usability tests it came out it's useless in the domain of experimental papers (see in this case all authors have similar order of magnitude of papers). While it can be useful in case of theoretical papers.
What about enabling this facet only when no experimental paper do match?
Alternatively what about displaying it only when the number of authors in the facet is actually small?
Also to assume "Mangano, Michelangelo L." to appear in the facet, means this literal string is actually there in the facet index. But already it might be in the form of "Mangano, M. L" or "Mangano, M."... so something smart would need to be done for implementing the concept of "the author I am searching for should appear in the facet".
several issues:
- we should normalize author names when facetting, so all possible name variations of the same author name are facetted together
- Mangano is a phenomenologist, not an experimentalist. Problem is that here we are including also all references in the results and facets, so if one paper of his happens to be cited a lot by big collaboration, his name gets drowned. So for it to ever be useful, references should not count towards author facet counts.
we should normalize author names when facetting, so all possible name variations of the same author name are facetted together
What do you propose as a good normalization? Also something based solely on the author name would facet together as usual omonims. In the past some proposals where made to use the BAI. As ugly as it is it identifies in an almost human-friendly way every author.
Mangano is a phenomenologist, not an experimentalist. Problem is that here we are including also all references in the results and facets, so if one paper of his happens to be cited a lot by big collaboration, his name gets drowned. So for it to ever be useful, references should not count towards author facet counts.
The thing is that: facets are faceting the results. And here results do match Mangano in any field (regardless of how well). So indeed what you say would imply that we detach the facets computation in ES from the search results computation. I fear this has quite some performance issue.
maybe the default search shouldn't include references. Anyhow, this is not a UI issue but a search tweaking issue, I am removing the UI tag and assigning to @iulianav.
But that is not enough to solve the general author facet problems: i.e. omonims and experimental papers. @StellaCh ?
could we do hierarchical facets? facet on normalized name (Mangano, M.), subfacet on BAI (M.Mangano.1, M.Mangano.2, etc.).
As long as we are super exact in normalizing. I.e. that all the possible signatures that an author might have are normalized in the same way. Otherwise you would end up having e.g.:
- Mangano M.
- M.Mangano.1
- M.Mangano.2
- Mangano M. L.
- M.Mangano.1
I believe that e.g. authors who changed surnames (spouses) fall into this trap (albeit a minority, but they would have some issues). And probably also Chinese/Russian users?
Just to see how other services have similar problem:
See the repetition of the name in the Author facet: https://www.semanticscholar.org/search?q=Michelangelo%20L.%20Mangano&sort=relevance
And here the author I am looking for is also not in the facets: https://www.semanticscholar.org/search?q=John%20Hardy&sort=relevance
Let's do this in an iterative way. The main concern here is that although from the results it's obvious that Mangano appears as an author, he doesn't appear in the facets. Why is that? Is it because other authors with more papers are higher? There needs to be a way to filter by "Mangano, Michelangelo L." (or any other variation) in the results.
(by showing more results or something similar) example: Click View All in Authors https://www.scopus.com/results/results.uri?numberOfFields=0&src=s&clickedLink=&edit=t&editSaveSearch=&origin=searchbasic&authorTab=&affiliationTab=&advancedTab=&scint=1&menu=search&tablin=&searchterm1=software&field1=TITLE_ABS_KEY&dateType=Publication_Date_Type&yearFrom=Before+1960&yearTo=Present&loadDate=7&documenttype=All&resetFormLink=&st1=software&st2=&sot=b&sdt=b&sl=23&s=TITLE-ABS-KEY%28software%29&sid=c0e981744a36ec8041923019833d1cd7&searchId=c0e981744a36ec8041923019833d1cd7&txGid=4b65c6351d365260f24721c56f78be93&sort=plf-f&originationType=b&rr=
After this is done, we'll investigate with the community:
- whether BAIs in facets make sense to them
- whether including references in the search results is a noise for everyone
- how to handle authors for experimental papers
Indeed the default search should ignore references,
- Annette
On 2 Oct 2017, at 15:36, Micha Moskovic <[email protected]mailto:[email protected]> wrote:
maybe the default search shouldn't include references.
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHubhttps://github.com/inspirehep/inspire-next/issues/2826#issuecomment-333536164, or mute the threadhttps://github.com/notifications/unsubscribe-auth/AM1-O9yW5xX6IPs7OP0H5sKMsn4S2KxVks5soObRgaJpZM4PqYoK.
