4 ms·
I tried this with a question for an area I know well. It's pretty impressive but missed some key references. I'd love to see limitations like this quantified a
by mnode 2y ago
I tried this with a question for an area I know well. It's pretty impressive but missed some key references.
I'd love to see limitations like this quantified and clearly flagged. Otherwise there's a danger that people may the assume results are definitive, and this could have the opposite outcome to that intended (much time spent working on something only to disocver it's been done already).
- jramette 2y agoYes, this is one of the most important aspects of the tool; in cases where you care about getting everything, make sure to take a look at the estimated percent of papers found at the bottom of the summary section. That gives you a sense of how complete the set of references likely are. We've tuned it to get around half of the total papers for the "median" user query on a first pass. If users desire, they can "extend" the search to have Undermind look for more papers. Additional caveat to remember with the current system is that it only accesses abstracts, so if you'd need to look in the full text to know a work is relevant, we wouldn't be able to catch it.
- mnode 2y agoThanks for clarifying. I didn't appreciate that it only searches abstracts. That might explain some of the missing references. Anyway, great work, will look forward to using it more.
- jramette 2y agoYep, will add in full texts as we can in the future. Let me know if the percent of papers found as an indication of an exhaustiveness measure was clear? Reach out to us at support@undermind.ai if you'd be willing to provide more feedback on your experience.