Covalent Data, first impressions
Thu Dec 8, 2016
Covalent Data is a tool to search over research topics. It seems to have the following features:
- Is a DB of grants, papers, people and institutions.
- claims to use machine learning to tie these entities together.
- search results don’t seem to have a way to be exported, so for example though the grant awards results do list the amount of each grant, to get a total amount against a search term, you would need to do the work of extracting each data point manually.
- is hard to determine what their sources are, specifically what data bases are not being covered?
- for multi word searches the engine prefers to return “near results” than exact matches, it was quite fiddly to force exact matching in the search interface, and I’m not convinced I was able to actually get it to work.
Some example searches:
search term “Digital Humanities”.
16 results found
All results are from NSF.
search term “computational social science”.
Again, all NSF funded grants
“social science” + “big data”. 90 grants. 24 publications
search term “stem cells”. 76k results.
In contrast The NSF grant search tool does allow a download of results, it found the following:
search term “computational social science”. 13 grants
search term “digital humanities”. 23 grants.
search term “stem cells”. 632 grants.
I was puzzeled as to how covalent data found so many more grants for stem cells, when both NSF and NIH reported far fewer.
Overall I’d like to see more grant agency coverage, more clarity around how results are generated, and an export ability. I could see covalent data becoming a useful tool at some point in the future, expecially if I felt that it was taking away the pain of having to go to many different sources to find grant funding information. Right now I’m not sure I trust it, and the results, as returned, are a bit hard to work with for onward analysis.