Reffer madness

From WikiDotMako
From a 10-6-2008 discussion on #openlibrary on
Reffer madness part I-II. Prefaced by a blog post on the longest now.



citations lack specificity.
as a result they mislead, they conflict, they obfuscate
they don't aggregate, they don't prioritize, they don't elaborate
slach lanu.
you have to track down the original to know how and in what sense and at what depth a source was cited and whether the source is relevant, a flimsy self-published rag, or not evey traceable.
and who has time or access to do that?

mako suggests: draft a standard for identifying types of refs, advocate for interest, start using it somewhere high-profile

sj suggests: blogging about this process, drawing in people who have though about this, defining it in a way that it can improve Wikipedia citation style and scalability of authority


all we can ask is that authors explain their perception of the reliability
i don't want to cite things opaquely
and i don't want others to do so as well
one nice thing i want is a way to do it that is at least quasi-compatible with exciting style guidelines
<brassratgirl> exciting!
so that i can argue in favor of doing it in things that i actually publish in other people journals
<greg-g> that seems like a must have
<mako> bemasc (not here) suggested treating citations as kind of like nouns

Actively attending[edit]

brassratgirl : Phoeb. E
mako         : B. M. Hill
jgay         : J.  Gay
_sj_         : SJ. Klein
solrize      : S. Olrize
stargirl     : H Wollach

Papers cited[edit]

  • Example : from Andrew McCallum
  • An annotation scheme for citation function by Simone Teufel, Advaith Siddharthan, Dan Tidhar (2006)
    • John M. Ziman. 1968. Public Knowledge: An Essay Concerning the Social Dimensions of Science. Cambridge University Press, Cambridge, UK.
      the only book I ever borrowed from the Harvard Physics Library without returning Sj
    • Ina Spiegel-Ruesing. 1977. Bibliometric and content analysis.

Social Studies of Science, 7:97-113.


on classifying academic papers[edit]

<jgay> another piece of software, Cora Research Paper Classification [relational document classification] - Research papers classified into a topic hierarchy with 73 leaves. We call this a relational data set, because the citations provide relations among papers.

err, rather, those are data sets we can use with the osftware

McCallum and Wollach[edit]

Andrew McCallum
He did Rexa and then here two short descriptions of papers he published last year
The two essays are "Learning to Predict the Quality of Contributions to Wikipedia" and "Topic Models Conditioned on Arbitrary Features with Dirichlet-multinomial Regression". I think if he can automate 90% accuracy rates with his programs, then he'll know what kinds of citations are good ones.
Also with Hannah Wallach he did a great paper entitled "Community-based Link Prediction with Text."
<mako> jgay: hanna mentioned tihs
<jgay> mako, his more recent work is more relevant, though

on reffing[edit]

Different ways of saying the same thing: revisit until all is self-similar and beautiful.

classes of refs[edit]

explicitly noting a dependency on a source's legitimacy (usually implying the reference is viewed positively and as a source of accuracy/legitimacy, save in satire or proof by counterexample):

  1. 'based (in some part) on',
  2. 'uses as positive reference/proof',
  3. 'uses as negative reference/proof'

explicitly stating legitimacy:

  1. 'promotes/supports',
  2. 'attempts to prove',

exlpicitly stating illegitimacy:

  1. 'discounts/criticizes',
  2. 'attempts to disprove'


  1. 'cites as transmitter of fundamental cite'
    <sj> there's actually a lot of conflation of proximal reference with original source that goes on when one is lazy or pressed for time leading at times to the wrong people being recognized for discoveries when this was not their intent
    <jgay> _sj_, yeah, that is really common.


  1. 'presents a different and possibly incompatible perspective'
  2. 'referred to for research but provided no inspiration for any section'

uses of sources[edit]

"I am relying on this source"
"I am refuting this source"
"I found this a source of amusement"
"this source was in my pile of library books at the end of the day, like the extra screws left over when you're done putting your whatsit back together"

types of cites[edit]

  1. nocite - influential work is used but not referenced or cited.
  2. noncite - incluential work is referenced in text but not in a cite
  3. anticite - citing a work to indicate it was read or reviewed as a potential reference, but could not be used anywhere in the work
  4. fauxcite - a random cite to make a section look better reffed than it is, not related
  5. selfcite - citing self's work as prior art; one can cite all of one's prior publications if one is godo at this, in each new work
  6. bibliocite - a cite to indicate a work was part of the reading/background
  7. middlecite - an intermediary who is citing the underlying original source, but was the work directly read by the author. there can be many layers of middleciting
  8. poison cite - intended to reframe the real meaning of the cited work; cite doesn't really say what it's imputed to say
  9. misleading cite - intended to confuse the course of a discussion; cite doesn't affect the argument the way it's implied to


<brassratgirl> _sj_: any citation scheme doesn't fix the having access problem

on why this is useful[edit]

<solrize> is this something anyone has cared about in the past N centuries of academic publishing? <solrize> i mean usually one just explains in words how the ref was used

<brassratgirl> solrize: sure, but there arent' that many people who are actually interested in citation styles
<_sj_> solrize, it has generally been considered charming that being a good scribe requires massive amounts of time and unique access to exotic journals

<_sj_> brg, I would say that this style standard shoulde xplicitly push for clarity in the cite for two reasons.

1) aggregation : you want to be able to combine a number of cites together, or combine cites through a chain of documents this is less possible if your aggregator has to parse natural language to make outall of the possible meanings your citing sentence may have
2) for parallelism : if you rely on natural language to define what you mean by a cite it will be more different among citers than if people have to explicitly pick a style. a style that says "I cite this to indicate I think it is wrong" is explicitly different from a "I cite this because it influenced me [negatively]"

on irc[edit]

For discussions like this we really need a tangents/talk channel and a get-shit-done channel.

every chan should have a get-shit-done channel. what to name it? sj
## tends to mean off-topic -jgay
so what means more-on-topic? sj


to come...