Search with free text instead of DOI

greenelab / preprint-similarity-search

A web app that uses machine learning to recommend the most suitable journals based on the text content of your preprint

https://greenelab.github.io/preprint-similarity-search/

Other

20 stars 4 forks source link

Search with free text instead of DOI #42

Open cgreene opened 4 years ago

cgreene commented 4 years ago

I got some feedback from @jakelever and permission to share it here

[x] It'd be nice to show the title of the preprint that's being searched for as some positive feedback that I'm being shown the right search results. When I used the example DOI, I didn't know what the paper was I was searching with, so didn't know if the results looked reasonable. I then dug out a few papers and the results look pretty good.
[x] It'd also be nice to not have the titles of the Related Papers shortened, as I had to use the Tooltip to read a lot of them
[x] The down arrow that appears after a while to take you down to the PMC Map is just a bit confusing. I wasn't sure where it would take me when I first clicked it.
[ ] I think someone else mentioned it, but it'd be awesome to be able to search with free text, if you don't even have your preprint up yet.

cgreene commented 4 years ago

The first two suggestions look like solid UI suggestions that we should implement. The other two might be a little more involved with respect to discussion (more below).

The down arrow is something we've been trying to figure out. Without it, we were getting feedback that it wasn't apparent that there was "more" below the list of journals for some users. Is there a way to incorporate an element into the design that suggests continuation? This might be worth breaking out into its own issue.

Also, on the free text input, we will need to think about that. It's technically quite possible but it does raise our potential attack surface a bit (right now, someone would have had to get a malicious PDF by bioRxiv as well). Also, having to post on bioRxiv to get the benefit seems helpful to the broader community, but this is a bit of a lesser concern to me. This also might be worth breaking out into its own issue for discussion.

vincerubinetti commented 4 years ago

Without it, we were getting feedback that it wasn't apparent that there was "more" below the list of journals for some users

How much feedback were we getting for this. I can't think of another case where a page has to do something special to indicate that the user should scroll down for more content, unless it is there is a perfectly fullscreen-sized element (like with the new greenelab website).

cgreene commented 4 years ago

A couple people said they stopped at the journals and didn't see the article or maps. I think we could restructure to:

Most similar articles
Most similar journals
Where this preprint fits in the overall landscape

That feels like it goes from most local to most global. Right now it feels like it starts at a middle level of resolution, then fine grained, then zooms all the way out.

vincerubinetti commented 4 years ago

A couple people said they stopped at the journals and didn't see the article or maps.

I don't understand this though. The journals and papers take quite a bit of time to load, and in the meantime the map should be clearly visible on most screens.

Take a look at the recent PR, where as soon as you search, it shows the "recommended" and "related" headings with loading spinners under each. Maybe that will help the issue somehow?

cgreene commented 4 years ago

Ok - I definitely think that helps! We can nixthe arrownow if we want to :) I do think we should reorder, and also change the text from "recommended journals" and "related papers" to "Most Similar Papers" and "Most Similar Journals"