5 ms·

That seems more a criticism of search engines than of anonymous publishing. What you need is a search engine that can find similar work. Being written by the a

by an_ko 11y ago

That seems more a criticism of search engines than of anonymous publishing. What you need is a search engine that can find similar work. Being written by the author is just a heuristic in a search for something different, even if probably helpful.

rsfern 11y ago
What I would like is to be able to explore the full citation graph, both interactively and programmatically. I think it would make identifying the key literature in a new topic much faster.
fixermark 11y ago
If we're going to group papers by author, we should probably assign some kind of unique, convenient identifier to such individual groupings, to make it easier for researchers to communicate between each other regarding groupings of interest to them. Oops, I just re-invented names.
dalke 11y ago
"A sufficiently advanced search engine" should be the new "sufficiently advanced compiler." I'll give an example of how difficult that is. I am researching the historical development of graph canonicalization for chemical information. I have bits and pieces from published papers, but did those people work together or independently? The papers include institutions, so I know two of the people (Gluck and Morgan) worked together at DuPont before Morgan went to the Chemical Abstract Service. The latter then cites a presentation Mooers gave at an American Mathematical Society meeting in 1958. But what was the relationship or key insight from Mooers? A key detail came from an autobiographical essay from 2001 at http://www.asis.org/History/12-lynch.pdf http://www.asis.org/History/12-lynch.pdf , which contains the following: > I was party to a conversation in which Harry Morgan [CAS] and Calvin Mooers discussed the issue of a canonical connection table. Mooers, in the light of the counter-example raised against the work of Dave Gluck, suggested a permutational approach to the design of an algorithm, which would overcome the objections. Harry Morgan (1965) then devised the method that bears his name and that has been a keystone in chemoinformatics since that time. That reveals a closer connection than seen in the papers. But what was this counter-example? I then read http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=html&identifier=AD0460819 http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=ht... which says: > The presentation of counterexamples, such as the arbitrarily numbered planar graph in Figure 5 devised by Dr. Lehman of the Walter Reed Army Institute of Research, have necessitated revisions to canonical ordering techniques such as those used by DuPont and the Chemical Abstracts Service. And there is it - the counter-example. But notice the use of institutional names, not published papers? How is a search engine supposed to make those connections? I can totally understand the argument that people should be referring to unique names, instead of a choice of names, to make this sort of thing easier for mechanized search. But that's not how people work. Why are they using institutional names? Because those components, like canonicalization, are part of a much larger system, which is more coupled to the institution than a single person. Hence why there's this marvelous publication (for a historian) which is a survey of the chemical information systems in the US in the early 1960s. https://books.google.com/books?id=B0ArAAAAYAAJ&lpg=PA172&ots=TP5X6xlSJM&dq=survey%20of%20chemical%20notation%20systems&hl=de&pg=PR1#v=onepage&q&f=false https://books.google.com/books?id=B0ArAAAAYAAJ&lpg=PA172&ots... . Indexed by institutions and people. Take away the latter, anonymize everything to just published papers, and the structure for understanding things collapses.
- dalke 11y ago
  "American Mathematical Society meeting in 1958" should be "American Mathematical Society meeting in 1950".
- phreeza 11y ago
  > How is a search engine supposed to make those connections? I can't tell you how exactly, but if current word embeddings can do stuff like "Paris - France + England = London", I don't see why future systems shouldn't be able to resolve the relationships between institutions, people and publications you mention, in a similar fashion. I am not saying that it will be easy, your example requires much more "one-shot learning", but I could in principle see how an automated system might do it.

by an_ko 11y ago

rsfern 11y ago

What I would like is to be able to explore the full citation graph, both interactively and programmatically. I think it would make identifying the key literature in a new topic much faster.

fixermark 11y ago

If we're going to group papers by author, we should probably assign some kind of unique, convenient identifier to such individual groupings, to make it easier for researchers to communicate between each other regarding groupings of interest to them. Oops, I just re-invented names.

dalke 11y ago

"A sufficiently advanced search engine" should be the new "sufficiently advanced compiler." I'll give an example of how difficult that is. I am researching the historical development of graph canonicalization for chemical information. I have bits and pieces from published papers, but did those people work together or independently? The papers include institutions, so I know two of the people (Gluck and Morgan) worked together at DuPont before Morgan went to the Chemical Abstract Service. The latter then cites a presentation Mooers gave at an American Mathematical Society meeting in 1958. But what was the relationship or key insight from Mooers? A key detail came from an autobiographical essay from 2001 at http://www.asis.org/History/12-lynch.pdf http://www.asis.org/History/12-lynch.pdf , which contains the following: > I was party to a conversation in which Harry Morgan [CAS] and Calvin Mooers discussed the issue of a canonical connection table. Mooers, in the light of the counter-example raised against the work of Dave Gluck, suggested a permutational approach to the design of an algorithm, which would overcome the objections. Harry Morgan (1965) then devised the method that bears his name and that has been a keystone in chemoinformatics since that time. That reveals a closer connection than seen in the papers. But what was this counter-example? I then read http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=html&identifier=AD0460819 http://oai.dtic.mil/oai/oai?verb=getRecord&metadataPrefix=ht... which says: > The presentation of counterexamples, such as the arbitrarily numbered planar graph in Figure 5 devised by Dr. Lehman of the Walter Reed Army Institute of Research, have necessitated revisions to canonical ordering techniques such as those used by DuPont and the Chemical Abstracts Service. And there is it - the counter-example. But notice the use of institutional names, not published papers? How is a search engine supposed to make those connections? I can totally understand the argument that people should be referring to unique names, instead of a choice of names, to make this sort of thing easier for mechanized search. But that's not how people work. Why are they using institutional names? Because those components, like canonicalization, are part of a much larger system, which is more coupled to the institution than a single person. Hence why there's this marvelous publication (for a historian) which is a survey of the chemical information systems in the US in the early 1960s. https://books.google.com/books?id=B0ArAAAAYAAJ&lpg=PA172&ots=TP5X6xlSJM&dq=survey%20of%20chemical%20notation%20systems&hl=de&pg=PR1#v=onepage&q&f=false https://books.google.com/books?id=B0ArAAAAYAAJ&lpg=PA172&ots... . Indexed by institutions and people. Take away the latter, anonymize everything to just published papers, and the structure for understanding things collapses.

dalke 11y ago
"American Mathematical Society meeting in 1958" should be "American Mathematical Society meeting in 1950".
phreeza 11y ago
> How is a search engine supposed to make those connections? I can't tell you how exactly, but if current word embeddings can do stuff like "Paris - France + England = London", I don't see why future systems shouldn't be able to resolve the relationships between institutions, people and publications you mention, in a similar fashion. I am not saying that it will be easy, your example requires much more "one-shot learning", but I could in principle see how an automated system might do it.