Wikipedia:Temporal Indexing

From Wikipedia, the free encyclopedia

This describes an effort to add time-based tagging articles as an additional method for indexing the contents of the encyclopedia so that events can be correlated by when they took place. This expands on the current year-linking procedures with the intention of adding more intelligence to how information is presented.

The purpose of this page is to:

  1. Identify whether temporal/time based tagging is something the community wants.
  2. Propose uses for the feature that support the ideals and goals of the project.
  3. Design the best method for implementing a supporting feature or discovering a way to replicate the needed functionality using features already present in MediaWiki.

Genesis[edit]

The initial proposal and discussion for this effort was made at this thread on Village Pump (proposals). Discussion points from it should be seeded to this page.

Overview[edit]

The proposed end-result would allow an end user to attach a sense of simultaneity to events documented in articles. Time has the potential to allow the creation of context where it currently doesn't exist. Articles about battles might be correlated with politics in a way that the current referencing system doesn't. While the means exist to cross-link related subjects together manually, there are limitations (in terms of maintenance and editors) in how many can be added to any article. Categorization has made great strides towards bridging this gap, adding time to the mix could finish the job.

The implementation of temporal indexing could be anything from a new type of date tag in MediaWiki that can specify ranges to a modification of existing Category logic or something else. The user-impact could be anything from the existence of a new tag that wikignomes would then use to add time awareness to articles all the way to logic that can extract dates already existant and treat them as (or convert them) into tags that are appropriately indexed.

Visually, a user might potentially be able to pivot an article to see other content in similar categories that are taking place during the same period. Or they might be able to expand the view to see all of the things happening at that time, something that might allow correlation with something that's not categorized as such but has a direct bearing on the events outlined. The politics of one country might affect the industrialization of another, for example. The patronage of one artist might have an effect on the career of another, the potential for this correlation is evident.

Discussion

Proposed methods[edit]

MediaWiki Feature[edit]

A feature enhancement for MediaWiki to support a new type of tag that allows text segments and/or pages to be identified and indexed to support time-based pivots/categorization.

Pros

  1. Potentially powerful
  2. Can be designed to do exactly what's needed programatically instead of relying on humans to build the structure by hand.

Cons

  1. Requires development effort, a rare commodity in a project this active.

Discussion

We could put tags around text, with the text showing as normal within the article, but not the tags:

<timeline date=[[16 March 1952]] Description of the event </timeline>


and when the edit is saved, the enclosed text is transcluded into an mainspace article (say, 1952 - March - 16) or even into an article in a new namespace [[Timeline:1952 March 16]]?

One concern regarding this approach is that in articles, full names of people typically aren't used after the first mention; similarly, sentences in the middle of an article may lack context (e.g., "Hitchcock began shooting the film in March 1952" is lacking Hitchcock's full name, the name of the film - presumably mentioned in a prior sentence, not necessarily the one just prior to this one - and an exact date in March, because that's normally uninteresting to the reader). I'm not sure how one might add detail to a transclusion but not the actual text; I suppose that the text to be included in the timeline could be written out in full as part of the tag, e.g. as a "Text=" function/argument, rather than trying to make the text do double duty (within the article and within the timeline). -- John Broughton (♫♫) 21:20, 18 May 2007 (UTC)[reply]

Use of existing Functionality[edit]

Maybe a category transclusion structure that takes the current year based categorization and linking (like 1993, 1886, etc) and adds more granularity (for example, individual days of these years).

Pros

  1. Can be done without modifying MediaWiki

Cons

  1. Potentially creates a large maintenance burden.
  2. Potential high cost for specifying ranges ranges. If a hundred date tags need to be individually transcluded to accurately reflect a date range, then it becomes unwieldy.

Discussion

Using existing information[edit]

It might be possile to make use of the link structure already present in articles to implement something like this.

Pros

  1. Doesn't require any extra work from editors
  2. Might be possible to implement on the toolserver or from a DB-dump without affecting anything else (and could be hosted by a third party to save work for the developers)

Cons

  1. Could be difficult to implement
  2. Wouldn't be as flexible as desired
  3. Might misinterpret the information already present

Discussion

Proposed idea[edit]

Idea text

Pros

Cons

Discussion

Uses[edit]

Time based browsing[edit]

Starting with an article, a user might instruct Wikipedia to show them everything else that directly relates to this (as ascertained with categories) that was happening at the same time (or within a week). Alternately, they might expand the view to show everything within a thousand kilometers that was happening at the same time (as a result of a similar geographic tagging effort or effort that indexes based on geotags).

Discussion:

Without some kind of discriminating factor (categories, geocoding, whatever), temporal indexing of information within articles could end up being fairly worthless. If a thousand things are coded as happening on July 1, 1986, how does one pick out the ones of interest (or, to be more exact, build a timeline of events that includes only the one or two events of that date - if any - that are relevant to that timeline (say, the history of Argentina)?

The problem, of course, is that categories that apply to the article as a whole don't necessarily apply to a particular event within the article. If a famous person was injured in (say) Paris, France, it's unlikely that the article about that person will have a category that ties the person to Paris (and if the article has localities as categories, there will be ones for the places of birth and death; how does one not tie those categories to events unrelated to them?)

In short, temporal indexing is fairly pointless unless there is some way to categorize the event; categories for the article are likely to be irrelevant or misleading in most, perhaps almost all, cases. And that, in turn, implies even more work might be necessary for temporal indexing - coding individual events (location=, type=, ... ). -- John Broughton (♫♫) 21:29, 18 May 2007 (UTC)[reply]

Machine generated timelines[edit]

While viewing a page about the rocket, a time-indexed database might be able to present an outline that brings together related subjects starting with Greek steam-wheel experiments, early alchemy, Chinese imperialism under early dynasties, introduction of gunpowder to Europe, German refinement during the 1930s, etc. While these are all subjects that diligent WikiBrowsing might eventually uncover, temporal categorization could allow new historical progression to be put together of individual disparate projects without requiring enough interest in a subject for someone to write an omnibus article.

Discussion:

New use idea[edit]

Lorem dang ipsum and whatnot

Discussion: