Citizendium Forums
November 24, 2009, 11:16:36 AM *
Welcome, Guest. Please login or register.
Did you miss your activation email?

Login with username, password and session length
News: POSTING RULES FOR MAIN CZ BOARDS: (1) The CZ Forums are Citizens-only (a "Citizen" is a Citizendium member). Non-Citizens may use only the "Non-member discussion" and "General help" boards, but still must register before posting (it's easy!). Non-Citizen posts elsewhere will be summarily deleted. (2) All must now use their own real names. To edit your displayed name, click on Profile > Account Related Settings. (3) Citizens must now link to their CZ user pages. To edit your signature, click on Profile > Forum Profile Information.
Click here to return to the wiki
 
   Home   Help Search Login Register  
Pages: [1]
  Print  
Author Topic: Generating and displaying information about relationships  (Read 980 times)
Howard C. Berkowitz
Forum Regular
****
Posts: 1763


« on: September 15, 2009, 07:50:08 AM »

As some of you may have noticed, I have been creating many lemma articles, typically under political/ideological/influence groups, think thanks, etc., about individuals. The entries, unless they grow to full articles, emphasize the group affiliations and perhaps individual mentors/proteges.

Underlying this is the idea of incorporating, into CZ, the sort of information that is available from certain political network databases. While, for example, we cannot make a sourced statement about someone's beliefs, I like the idea of a user having the information available that the individual belongs to five groups that have the flat earth as a position, studied at six institutions with a Flat Earth Department, and has mentored seven Professors of Flat Earthism.

One of the hardest issues of this sort of thing is physically presenting the information. Many of the political science ones use a graphic model, but this gets extremely complex for influential, highly connected people. Still, assuming they are up-to-date, Related Articles pages have the raw data for much of this.

The Semantic Web and related technologies are relevant here, and the neuroscientists/connectionists think about this a lot. Perhaps there might be a way to designate certain articles as candidates for semantic/connectionist mapping.

Another issue is maintenance,  for which we have a partial solution, which I've discussed briefly with Daniel. The Related Article bot now runs once per article.  For certain expliciitly designated articles/links, it may be useful for rerunning it and adding suggested maintenance updates to Related Articles pages. In some cases, the lists may grow so large (e.g., a political party) that a Catalog may be more appropriate than a related article R-template.

As one example beloved of the conspiracy theorists, consider membership in the Council on Foreign Relations, or attendance at Davos/World Economic Forum, etc.  Consider membership in political parties.

I see no easy way to handle resignations, most end of elected terms, etc.

I would, however, like to start discussion. Shamira and her students may have particular insight, because I think these linkages are valuable data for people studying the sorts of policy-influencing groups she has assigned as Eduzendium projects.

An at least minimally useful linkage mapper would be a huge differentiator from WP.  In no way do I underestimate the maintenance.  Just as an example, I added a lemma article for Al-Shabab‎, an Islamist group about which I had never heard before, but was the target of a U.S. counterterrorism raid yesterday (or perhaps that is when it was announced) in Somalia. Since I didn't think there was anything about it, I did go ahead and create Related Articles, in which al-Qaeda is listed. Should I have to manually add it to al-Qaeda's subtopics? Arguments both ways...
« Last Edit: September 15, 2009, 08:38:30 AM by Howard C. Berkowitz » Logged

http://en.citizendium.org/wiki/User:Howard_C._Berkowitz

Prime Minister, you can't take the bull by the horns if you're grasping the nettle. I mean, if you grasped the nettle with one hand, you could take the bull by one horn with the other hand, but not by both horns because your hand wouldn't be big enough, and if you took a bull by only one horn it would be rather dangerous because...' (Yes Prime Minister II, pp. 221-2)
Peter Schmitt
Forum Regular
****
Posts: 373


« Reply #1 on: September 15, 2009, 11:00:15 AM »

Just as an example, I added a lemma article for Al-Shabab‎, an Islamist group about which I had never heard before, but was the target of a U.S. counterterrorism raid yesterday (or perhaps that is when it was announced) in Somalia. Since I didn't think there was anything about it, I did go ahead and create Related Articles, in which al-Qaeda is listed. Should I have to manually add it to al-Qaeda's subtopics? Arguments both ways...

If I understand correctly what you ask, my answer is: Yes.

The group is linked to al-Qaeda in the text.

Related Articles (by the way: the link is not shown on the main page: is this typical for lemma articles?)
should point to it as Parent (if it is a subgroup) or as Related topic, (if it is independent - then "Islamist group" would be parent).
If I follow your description, it is not a subtopic (as in your list).

In Related Articles of al-Qaeda it should be listed either as subtopic or as related, depending on the relation.
In Islamist groups it would be a subtopic, but could be omitted if it is a subgroup of al-Qaeda (or indicated as such).

These decisions cannot be automated.
Logged

Howard C. Berkowitz
Forum Regular
****
Posts: 1763


« Reply #2 on: September 15, 2009, 11:22:00 AM »


Related Articles (by the way: the link is not shown on the main page: is this typical for lemma articles?)
Yes. No subpages, even if defined, show on the main page of a lemma article.

Thanks for the thoughts; they are helpful. Perhaps automatically adding the relationships is unwise, but a report of the potential additions as a result of new article creation might be very useful.
Logged

http://en.citizendium.org/wiki/User:Howard_C._Berkowitz

Prime Minister, you can't take the bull by the horns if you're grasping the nettle. I mean, if you grasped the nettle with one hand, you could take the bull by one horn with the other hand, but not by both horns because your hand wouldn't be big enough, and if you took a bull by only one horn it would be rather dangerous because...' (Yes Prime Minister II, pp. 221-2)
Daniel Mietchen
Forum Regular
****
Posts: 674



WWW
« Reply #3 on: September 16, 2009, 01:42:48 AM »

Just as an example, I added a lemma article for Al-Shabab‎, an Islamist group about which I had never heard before, but was the target of a U.S. counterterrorism raid yesterday (or perhaps that is when it was announced) in Somalia. Since I didn't think there was anything about it, I did go ahead and create Related Articles, in which al-Qaeda is listed. Should I have to manually add it to al-Qaeda's subtopics? Arguments both ways...

If I understand correctly what you ask, my answer is: Yes.

The group is linked to al-Qaeda in the text.

Related Articles (by the way: the link is not shown on the main page: is this typical for lemma articles?)
should point to it as Parent (if it is a subgroup) or as Related topic, (if it is independent - then "Islamist group" would be parent).
If I follow your description, it is not a subtopic (as in your list).

In Related Articles of al-Qaeda it should be listed either as subtopic or as related, depending on the relation.
In Islamist groups it would be a subtopic, but could be omitted if it is a subgroup of al-Qaeda (or indicated as such).

These decisions cannot be automated.
They cannot be automated in a general way across the site, though the Semantic Mediawiki extension goes in this direction (has anyone played with it recently?). It may be possible, however, to automate this for particular fields. A biomedical ontology, for instance, can be found via the MeSH terms (example: dehydration). The problem with this is that it will be hard to find to couple their taxonomy to that of other fields, even if comparable ontologies exist there. And even in the entries of a well-curated and comprehensive field-specific ontology like MeSH, certain aspects will always be missing. The dehydration entry above, for instance, neglects dehydration due to freezing of water inside or near the organism. I thus think that although we should regard these decisions as not automatable for the moment (because we lack resources to put into that), we should keep the option in view for the long run.
Logged

Daniel Mietchen
Forum Regular
****
Posts: 674



WWW
« Reply #4 on: September 16, 2009, 01:57:15 AM »

Another issue is maintenance,  for which we have a partial solution, which I've discussed briefly with Daniel. The Related Article bot now runs once per article.  For certain expliciitly designated articles/links, it may be useful for rerunning it and adding suggested maintenance updates to Related Articles pages. In some cases, the lists may grow so large (e.g., a political party) that a Catalog may be more appropriate than a related article R-template.
Small correction: The bot currently does not edit existing Related Articles, but it may run multiple times over an article as long as nothing links to it from the main namespace. I am indeed working on a version that provides updates, either automatically (for pages in the "bot-created" category), or optionally upon manual request. I have most of the algorithmic part sorted out but my Python is too rudimentary to really put those ideas into working code all by myself. If any one who reads this is an experienced Python programmer, please leave me a note, and this updating bot may soon be up and running.
Logged

Howard C. Berkowitz
Forum Regular
****
Posts: 1763


« Reply #5 on: September 16, 2009, 09:26:48 AM »

Another issue is maintenance,  for which we have a partial solution, which I've discussed briefly with Daniel. The Related Article bot now runs once per article.  For certain expliciitly designated articles/links, it may be useful for rerunning it and adding suggested maintenance updates to Related Articles pages. In some cases, the lists may grow so large (e.g., a political party) that a Catalog may be more appropriate than a related article R-template.
Small correction: The bot currently does not edit existing Related Articles, but it may run multiple times over an article as long as nothing links to it from the main namespace. I am indeed working on a version that provides updates, either automatically (for pages in the "bot-created" category), or optionally upon manual request. I have most of the algorithmic part sorted out but my Python is too rudimentary to really put those ideas into working code all by myself. If any one who reads this is an experienced Python programmer, please leave me a note, and this updating bot may soon be up and running.

Unfortunately, I am most experienced with Monty Python, but, for a variety of reasons, I need to become proficient again at programming, something I did for many years. Would it be useful to establish a Forum area for Python and for MediaWiki programming mutual learning assistance? 

Last night, I was considering trying to download the Windows version of gawk, since I was once very good in awk, and realized that is putting off the inevitable. Having taught programming, everyone has their own learning difficulties, but they tend to be different enough that learners can help one another.
Logged

http://en.citizendium.org/wiki/User:Howard_C._Berkowitz

Prime Minister, you can't take the bull by the horns if you're grasping the nettle. I mean, if you grasped the nettle with one hand, you could take the bull by one horn with the other hand, but not by both horns because your hand wouldn't be big enough, and if you took a bull by only one horn it would be rather dangerous because...' (Yes Prime Minister II, pp. 221-2)
Howard C. Berkowitz
Forum Regular
****
Posts: 1763


« Reply #6 on: September 16, 2009, 09:30:25 AM »

As some of you may have noticed, I have been creating many lemma articles, typically under political/ideological/influence groups, think thanks, etc., about individuals. The entries, unless they grow to full articles, emphasize the group affiliations and perhaps individual mentors/proteges.

Underlying this is the idea of incorporating, into CZ, the sort of information that is available from certain political network databases. While, for example, we cannot make a sourced statement about someone's beliefs, I like the idea of a user having the information available that the individual belongs to five groups that have the flat earth as a position, studied at six institutions with a Flat Earth Department, and has mentored seven Professors of Flat Earthism.

One of the hardest issues of this sort of thing is physically presenting the information. Many of the political science ones use a graphic model, but this gets extremely complex for influential, highly connected people. Still, assuming they are up-to-date, Related Articles pages have the raw data for much of this.

We have had some good thoughts on maintenance and full semantics. Could I ask for a focus on whether we could find a way to present relationship information that is already present in Related Articles? Essentially, it would start with one article and show things linked to it. The graphic (if that's the format) would then show the articles one more degree of separation away.

Right now, the "links here" and the Related Article bots don't show the degree of separation, which they probably know. Visual presentation is a separate challenge; there are things around and perhaps one could be adapted -- mindmappers, etc.
Logged

http://en.citizendium.org/wiki/User:Howard_C._Berkowitz

Prime Minister, you can't take the bull by the horns if you're grasping the nettle. I mean, if you grasped the nettle with one hand, you could take the bull by one horn with the other hand, but not by both horns because your hand wouldn't be big enough, and if you took a bull by only one horn it would be rather dangerous because...' (Yes Prime Minister II, pp. 221-2)
Peter Schmitt
Forum Regular
****
Posts: 373


« Reply #7 on: September 17, 2009, 04:14:58 PM »

Right now, the "links here" and the Related Article bots don't show the degree of separation, which they probably know. Visual presentation is a separate challenge; there are things around and perhaps one could be adapted -- mindmappers, etc.

"What links here" has (at least) two levels: Links to redirects are shown indented.
Perhaps this can be generalized?
Logged

Peter Schmitt
Forum Regular
****
Posts: 373


« Reply #8 on: September 17, 2009, 04:23:02 PM »

Would it be useful to establish a Forum area for Python and for MediaWiki programming mutual learning assistance? 

Having taught programming, everyone has their own learning difficulties, but they tend to be different enough that learners can help one another.

In principle, I am not opposed to learn (a) new language(s). In principle it means a new syntax and new reserved words.
And finding out the implemented procedures. (The thinking in algorithms is essentially the same.)

However, I did not expect this to be necessary when joining an encyclopedia project.  Smiley
Logged

Pages: [1]
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.7 | SMF © 2006-2008, Simple Machines LLC Valid XHTML 1.0! Valid CSS!