nodalpoint.org - Unique identifiers for authors in scientific publications - Comments

all good points

Neil — Wed, 27 Apr 2005 04:16:45 -0400

Yes, I was in "devil's advocate" mood today (server hardware failure will do that to you). Of course, I think machine identification is A Good Thing (TM).

Now, don't get me started on data integration :-)

Implicit in all this discussi

Greg — Wed, 27 Apr 2005 02:40:49 -0400

Implicit in all this discussion (IMO) is that we wish to provide a means of identifying scientists to machines. Clearly if there was a problem for humans to identify scientists in the literature then the current system simply would not work. For machines this is not the case.

At present none of these issues seem like major problems. Sure there are probably numerous C. Lees in the literature, but the one you're interested in publishes on archaea and not cancer. So the same name confusion doesn't really exist, for you.

The premise for all this discussion is that the scientific literature is getting out of hand: too many papers, too much information and it's getting worse. I am inclined to believe these assertions, based on my personal experience as a scientist. One possible solution to the overwhelming amount of information being produced is to use software (i.e. machines) to filter, cluster, link and search for us. Now assuming that your software can distinguish between all the different C. Lees (see previous posts on identifying properties) then you open up some interesting possibilities...

First, how often do we use author names as the main criterion for tracking literature?

At the moment it is technically difficult (for both machines and humans) as noted in the original post to track author names. But if it was easier for machines, then the range of searches would be significantly expanded.

Assume for the moment that I'm interested in archaea, and hell, why shouldn't I be they are interesting organisms. What I need to do is become knowledgeable in that area, currently this task is non-trivial. I need to read reviews, track down the papers referenced in those reviews, become knowledgeable about the authors of those papers etc. The end result is inefficient. I need a more efficient way of doing this. But why make it complex, just stick to your field of interest ?

Everything is becoming more cross-disciplinary, the boundaries that existed 10 to 20 years ago in science are no longer there. So often I find myself needing to head into new fields to essentially cherry pick knowledge. I need the best information possible and I need to access it efficiently.

And this is why keyword searches are a dead end. You as an expert in your field will have no problems with key word searches. But when it comes to archaea I have no idea. But let's say I find one paper that is interesting. What I want to do is seed a knowledge agent (machine) with that paper and let it do the work of tracking down relevant information (no this won't substitute for actually reading the papers, it is just a more efficient way of gathering the information). People who work together publish together, people in the same field will propagate similar ideas. Tracking authors and the links between them may in fact be more relevant than using keywords (think professional scientific networking).

The ego driven concern is not relevant to my argument (i.e. I want machines to identify you) nor is the fraud issue (this was just a consequence of using real world identity as an example).

Is it reasonable to expect that someone else should build a complex system just so that you can be identified?

If the question is to be identified by a machine then the answer is yes. In the future you won't be able to survive with out it...

Overkill?

Neil — Tue, 26 Apr 2005 19:14:14 -0400

This is a fascinating thread, but I'm beginning to wonder if this is an instance where we are creating an unnecessarily complex solution for a problem that may not really be a problem.

First, how often do we use author names as the main criterion for tracking literature? More commonly, we start our searches by topic - I'm far more likely to type "archaea" into PubMed than "Woese". True, we may be interested in the work of a particular person, but if we lose track of them as we go back through time due to their name changing, it's still more likely that keywords in the title or abstract are the things that will catch our eye.

Second, it seems some of these concerns are ego-driven - in other words, "how will people find me and my work?" I think when most people publish a paper, they keep a record of it for their C.V. and many people will keep a list of their publications on a website. In other words - people will find you if you promote yourself. Is it reasonable to expect that someone else should build a complex system just so that you can be identified?

The fraud issue is interesting, but I feel not very relevant in science. You could claim to be an author on a paper when you are not, but anyone who wanted to check would quickly discover the truth. I suspect that most people don't bother to confirm authorships via PubMed when presented with a candidate's publication list, but the fact that they could do so is sufficient to prevent this kind of fraud. That and the fact that scientists are on the whole rather honest people who take pride in their real achievements rather than concocting false ones.

No

Greg — Tue, 26 Apr 2005 03:48:32 -0400

I am saying that a collection of "identifying properties" are, in most reasonable cases, sufficient to identify you in the scientific literature. These identifying properties maybe your name, email address, titles of previous publications, institute address etc. I say "in most reasonable cases" to avoid the objection that if you move, change your name, then change your name again how are we going to identify you ? The answer is we won't. Furthermore this is a more flexible system as it does not rely upon a single stable identifier for each author.

I think the first intuitive solution to this problem that everyone initially thinks of is: "Give every author a unique identifier". Which on the surface seems reasonable (i.e. there are no technical barriers). When you consider this solution in detail you begin to see the flaws. For example a single identifier must be maintained by some organization. It has already been stated that maintaining this kind of registry for all time is a non trivial task. Also if the system has a single identifier then the system also has a single point of failure.

To put this in more concrete terms, think of real world identity. In countries like Australia with no national identity card, when you are required to identify yourself, for example when opening a bank account, you do so by providing a number of different documents. The information in these documents are "identifying properties". And a majority of these identifying properties must be shared across documents e.g. your name, address etc. each document is also weighted. Thus your passport is given more weight than your most recent bank statement.

Now in countries like Taiwan where everyone must have a national ID, opening a bank account only requires your national ID card. On the surface this seems simpler, more efficient etc. But it is less secure and less flexible. To open a bank account in Australia using a false ID you must provide more than one fake document (the weight of evidence system), whereas in Taiwan you would only need a single fake national ID card. True, it may be more difficult to forge a national ID card but not impossible for those who really want to.

What I would like to see is the weight of evidence system translated into the electronic world. So each time you submit a paper you must supply a number of identifying properties. These are then made available in a machine readable format (FOAF) so that other (external) application can use that data (to link authors and publications).

Also of interest is PGP keys being used in this process. Again, not even your PGP key is proof of who you say you are, which is why the need for key signing parties.

Responsibility?

chris — Mon, 25 Apr 2005 08:48:45 -0400

So you're saying that an independent semantic web entity can uniquely identify me in the scientific literature for all time, providing I don't do anything silly like moving, changing address, getting fired, etc. And this is a good thing? I thought the point was to make the system more, rather than less, flexible...

Yes you are correct if we ass

Greg — Mon, 25 Apr 2005 06:27:36 -0400

Yes you are correct if we assume that your email is being used as the only identifier, but that is not really the point of "identifying properties". Think of being identified (or rather authenticated) as the weight of evidence to support the authentication. It is not absolute and it is never going to be perfect. What we lack is journals providing identifying properties in a machine readable format i.e. FOAF.

In fact using identifying properties like email (or a hash of your email) is a good reason not to change your email (which is entirely possible). In effect this puts the burden on you to maintain your online scientific identity.

In effect you could remain unidentifiable in the online scientific literature if you chose to (i.e. change your name and email address etc.). But what would be the point of that ? Think of the confusion, the lost fame, fortune, and glory... umm, this is science, what am I thinking...

Backups revisited

chris — Thu, 21 Apr 2005 22:52:38 -0400

Still, even 10 years ago we had nightly departmental backup...grrr.

My currect solution is to have external USB hard drives to dump to, which are then taken off-site for storage. My previous system, incremental backups to a separate drive in the system, didn't address the possibility of theft/destruction of boxen, which would have been, in a word, catastrophic.

Unfortunately, the current setup means that I have to rely on other people doing their backup concientiously - but that's their problem, not mine. Empower the user, and watch them stuff up!

Past times

Neil — Thu, 21 Apr 2005 22:15:35 -0400

Not to mention the inability to index pre-email literature...

Easy to forget that there was once no such thing as email, isn't it? Yesterday I was exploring the multitude of Google services and the fantastic Wikipedia, when it occurred to me that not only did these things not exist when I was a Ph.D. student, but the Web itself barely existed.

Still, even 10 years ago we had nightly departmental backup...grrr.

Email addresses not always stable

chris — Thu, 21 Apr 2005 20:51:59 -0400

For this case this would mean: use e-mail addresses, rather than names.

Until someone moves and gets a new email account, with the added headache that there is no direct mapping between past and present/new addresses (bob@somewhere.edu becomes bjbloggs@institute.edu.au). Not to mention the inability to index pre-email literature...

Greg — Thu, 21 Apr 2005 00:51:17 -0400

Here are some more references to identity managemnt. I belive this post is an oblique reference to IFPs:

Take for example my name. When I am among my immediate family, "John" is sufficient to identify me absolutely unambiguously. At work, where there are many "John"s, it is sometimes necessary to refer to me as "John Black". Occasionally, even that fails, as there are more than one "John Black" in the context, then it is necessary to add additional properties, such as, "The John Black from New Mexico."

This very RDF specific post on identity from Phil Dawes:

This blurryness has lead to me consider putting the final responsibility of identity on the shoulders of the ‘aggregator’ (i.e. the person doing the aggregating) rather than the author of the data. It’s a compelling solution since the person doing the aggregating is collecting and merging data for a particular purpose under a particular context.

Good point

Greg — Wed, 20 Apr 2005 09:05:27 -0400

This comment remined me of the FOAF specification. FOAF is an RDF vocabulary for describing people and by definition identifying people. However during the design of the FOAF vocabulary it was decided that giving each person a unique identifier was a bit too 1984. So the solution was to use "inverse functional properties" or IFPs. Basically these are "identifying properties" (i.e. properties which only one individual is likely to have), for example your email address. In fact FOAF uses the SHA1 hash of you email address as the main IFP. See this article for more on IFPs in FOAF.

This seems like an appealing solution to the problem of a unique author ID being associated with academic publications. For example your IFPs would be your institution, the SHA1 of your email address, maybe the SHA1 hash of your other papers' titles ?

A simple solution exists ...

boris — Tue, 19 Apr 2005 20:52:58 -0400

IMHO the crux with all unique identifiers is the overhead of maintaining the required registries. Convincing solutions I have seen build on existing registries, rather than add new ones. For this case this would mean: use e-mail addresses, rather than names.

Uh-huh

chris — Sun, 10 Apr 2005 07:20:43 -0400

After all, the publications are just a vehicle to get scientific discovery to other scientists. It does not matter at all which journal it appears in or who writes it because it's all peer reviewed. No, wait..

Wow! Can I come and play in your universe? :-)

Changing the system

Spitshine — Sun, 10 Apr 2005 03:16:29 -0400

Progress is slow these days - we should really get such a system in place but I don't think we will. People want to abolish the silly single file of the author list or better the peer review process for a long while. PLoS could have changed a lot and while they improved on several aspects of scientific publishing beyond the Open Access model, they have not introduced many enhancements. I guess we have to live with the Park Ks, Chou Xs and the Smiths Js.
After all, the publications are just a vehicle to get scientific discovery to other scientists. It does not matter at all which journal it appears in or who writes it because it's all peer reviewed. No, wait..

A good idea

Neil — Sat, 09 Apr 2005 02:11:19 -0400

I support this. My advice to people whose name might be going on a paper - make sure the writer is aware of your full name, all initials and correct spelling before submission. I have appeared variously as Saunders N, Saunders NF and Saunders NFW and it can cause confusion later.

Side note - why does online PubMed limit initials to 2? Everyone knows that the best scientists have at least 2 middle names :-)

Unique identifiers for authors in scientific publications

toniher — Fri, 08 Apr 2005 09:27:38 -0400

In Biomedical Digital Libraries journal, I have read an interesting paper which proposes unique identifiers for the authors of publications in Biomedicine.
This need comes from the growing amount of scientific publications, and the ever increasing difficult task of identifying the authors of those works.
Because of women incorporation in research and worldwide scientific contribution, the Anglo-Saxon inspired system of annotation based on a first name and one surname is not currently suitable for following authors trajectory. Things can get more difficult when authors may change the names they use along their lives. For instance, in some cultures, women usually adopt the surname of their husbands when they get married.
In order to show this, we can search in Pubmed Hernández A. and Lee C. These are not two very prolific researchers in different Biosciences fields, but simply, there are many people who may respond to this query.
Using a unique, unequivocal and personal identifier for each author, which could be centralized and registered in servers such as Pubmed (but preferably shared with other sites), could be a possible solution for this kind of problems.