5 ms·

Open Academic Graph

64 points by kercker 9y ago

Faaak 9y ago
The website is very long to load, its a bit of a shame. However, I've always wanted to do that. A revocation feature would be good: if someday a paper is proved false, all underlying papers citing this one should be too.
- _0ffh 9y ago
  "if someday a paper is proved false, all underlying papers citing this one should be too" Um, most certainly not? First, not all of the things you might cite in a paper are part of the support structure to your results. Citing other works one the same topic is quite common, for example. Second, even of those citations which are part of the support structure for your conclusions, usually just a few of them are absolutely crucial to your point. The rest is just a web of additional supporting evidence. Third, in view of the above points and the fact that every paper quotes a multitude of papers, just invalidating one paper would unnecessarily "invalidate" vast swathes in an explosive chain reaction.
  - flamedoge 9y ago
    papers that cited need to be re-examined at best. we certainly shouldn't rule out the event of recursive retraction of papers if they all were vitally dependent on false claims.
    - mattkrause 9y ago
      This seems incredibly unlikely. I just had a paper accepted (yay!). Although we cited sixty-odd papers, I don't think our conclusions critically depends on any one of them. A lot of the citations place the current work in context. If these papers turned out to be false, it might make the paper less "interesting"--the mechanism underlying a useless therapy is pretty boring, after all--but the factual content would be unchanged. Other citations compare our results with others' work. These discussions are silly if the other work turns out to be flawed, but again, the facts in our paper would not change. We had a lot of techniques-related citations. Many of these, however, are mostly about giving credit or pointing to extra information: we used a behavioral paradigm adapted from this paper, evaluation was done using the criteria established by that paper, etc. A few citations are used to "prove" something is (ir)relevant for the current experiments. An error in one of these papers could be damaging. However, we usually try to have multiple sources for these claims, so any one error would not be calamitous. People usually try to show that their own data matches the cited work, so we have several analyses like that. Finally, it's worth noting that an individual paper (including mine) does not provide iron-clad or no evidence for something. Instead, it should nudge your beliefs slightly one way or another.
- conjectures 9y ago
  So paper X is published. Paper Y says paper X is false. -> Paper Y is false. I think there is a clear reason that this is not satisfactory.
RossBencina 9y ago
For me, it took more than 30 seconds to "load," (spinner, progress bar) but then just showed a blank page with a grey bar at the top. Windows 10, Chrome, Ghostery.
paulpauper 9y ago
nothing loaded
ColinWright 9y ago
Nothing loaded. I'm looking forward to this working properly - is there anything I can do to help debug/investigate?
- anfractuosity 9y ago
  It looks like the request to https://api.aminer.org/api/article/alias/open-academic-graph https://api.aminer.org/api/article/alias/open-academic-graph hangs. Eventually I get '502 bad gateway'.
_Microft 9y ago
If the page isn't loading for you, go to the frontpage https://aminer.org/ https://aminer.org/ instead of the linked page.
nopinsight 9y ago
Their list of top venues for different fields of computer science [1] seems to miss several top conferences in each field. Also, a mistake like this does not help: Machine Learning, International Council for Machinery Lubrication. https://aminer.org/mostinfluentialscholar#awards_by_fields https://aminer.org/mostinfluentialscholar#awards_by_fields
anfractuosity 9y ago
Woo it finally loaded for me. "This data set is generated by linking two large academic graphs: Microsoft Academic Graph (MAG) and AMiner. The data set is used for research purpose only. This version includes 166,192,182 papers from MAG and 154,771,162 papers from AMiner." There's zip files: * 103GB MAG papers * 39GB AMiner papers * 1.6GB linkage

64 points by kercker 9y ago

The website is very long to load, its a bit of a shame. However, I've always wanted to do that. A revocation feature would be good: if someday a paper is proved false, all underlying papers citing this one should be too.

_0ffh 9y ago
"if someday a paper is proved false, all underlying papers citing this one should be too" Um, most certainly not? First, not all of the things you might cite in a paper are part of the support structure to your results. Citing other works one the same topic is quite common, for example. Second, even of those citations which are part of the support structure for your conclusions, usually just a few of them are absolutely crucial to your point. The rest is just a web of additional supporting evidence. Third, in view of the above points and the fact that every paper quotes a multitude of papers, just invalidating one paper would unnecessarily "invalidate" vast swathes in an explosive chain reaction.
- flamedoge 9y ago
  papers that cited need to be re-examined at best. we certainly shouldn't rule out the event of recursive retraction of papers if they all were vitally dependent on false claims.
  - mattkrause 9y ago
    This seems incredibly unlikely. I just had a paper accepted (yay!). Although we cited sixty-odd papers, I don't think our conclusions critically depends on any one of them. A lot of the citations place the current work in context. If these papers turned out to be false, it might make the paper less "interesting"--the mechanism underlying a useless therapy is pretty boring, after all--but the factual content would be unchanged. Other citations compare our results with others' work. These discussions are silly if the other work turns out to be flawed, but again, the facts in our paper would not change. We had a lot of techniques-related citations. Many of these, however, are mostly about giving credit or pointing to extra information: we used a behavioral paradigm adapted from this paper, evaluation was done using the criteria established by that paper, etc. A few citations are used to "prove" something is (ir)relevant for the current experiments. An error in one of these papers could be damaging. However, we usually try to have multiple sources for these claims, so any one error would not be calamitous. People usually try to show that their own data matches the cited work, so we have several analyses like that. Finally, it's worth noting that an individual paper (including mine) does not provide iron-clad or no evidence for something. Instead, it should nudge your beliefs slightly one way or another.
conjectures 9y ago
So paper X is published. Paper Y says paper X is false. -> Paper Y is false. I think there is a clear reason that this is not satisfactory.

RossBencina 9y ago

For me, it took more than 30 seconds to "load," (spinner, progress bar) but then just showed a blank page with a grey bar at the top. Windows 10, Chrome, Ghostery.

paulpauper 9y ago

nothing loaded

ColinWright 9y ago

Nothing loaded. I'm looking forward to this working properly - is there anything I can do to help debug/investigate?

anfractuosity 9y ago
It looks like the request to https://api.aminer.org/api/article/alias/open-academic-graph https://api.aminer.org/api/article/alias/open-academic-graph hangs. Eventually I get '502 bad gateway'.

_Microft 9y ago

If the page isn't loading for you, go to the frontpage https://aminer.org/ https://aminer.org/ instead of the linked page.

nopinsight 9y ago

Their list of top venues for different fields of computer science [1] seems to miss several top conferences in each field. Also, a mistake like this does not help: Machine Learning, International Council for Machinery Lubrication. https://aminer.org/mostinfluentialscholar#awards_by_fields https://aminer.org/mostinfluentialscholar#awards_by_fields

anfractuosity 9y ago

Woo it finally loaded for me. "This data set is generated by linking two large academic graphs: Microsoft Academic Graph (MAG) and AMiner. The data set is used for research purpose only. This version includes 166,192,182 papers from MAG and 154,771,162 papers from AMiner." There's zip files: * 103GB MAG papers * 39GB AMiner papers * 1.6GB linkage