Wikipedia talk:WikiProject Portals/Quantum portals

Menu item or special: link that generates a portal when you click on it

Taking things to their logical conclusion, how feasible is it to have a portal link on the sidebar menu, or in one of the tab menus at the top of every page, that generates a portal for the current article's subject when you click on it?

So, if you are reading the article on chimpanzees, and you click on "Portal", a "Special:" page called "Special:Portal:Chimpanzees" is generated and displayed right then on chimpanzees.

I look forward to your replies. Sincerely,    — The Transhumanist   21:15, 12 August 2018 (UTC)

Quantum portals that exist only as a probability function (algorithm) until you collapse the wave form by observing through the portal button (run the script), and disappear again after use? · · · Peter (Southwood) ^(talk): 15:28, 14 August 2018 (UTC)

More seriously, what do you envisage as the structural foundation for such a portal? I can see an instant portal generator possibly working using a navbox or an outline list or a category as the source of the selected article set. The root article would presumeably be the one from which the portal is called, so how would it work if there is no similarly named category or outline list? How would it deal with orphan articles and others which don't have many associated articles? Would one just accept that instant portals would often be relatively low level and sparse of content, or would it only work for articles with the required associated structures? Some features commonly used in regular portals may be difficult to produce, or could be quite run time intensive to build. I like the idea in principle as a very user friendly introduction to a topic, but is it technically plausible for current technology within the Wikipedia environment, or would this be more appropriate as an external tool drawing on Wikipedia for the content? (just brainstorming here) · · · Peter (Southwood) ^(talk): 15:28, 14 August 2018 (UTC)

I believe it is feasible to build such a system, over time, but it is not a short-term project (maybe a year or three). However, I can't help but notice that we are heading in this direction, and the distance we've traveled in just the past 4 months.

With the right source pages (or other data sources), portal creation goes quick. Prior to the portals project reboot, portal creation typically took 6 hours or more, each. This morning, from 12:16am to 7:34am I created 51 portals. That's about 8 minutes each, on average. Now I'm thinking that portal creation could eventually be nearly instantaneous, via a universal gadget or Mediawiki extension, with the right data foundation.

But, we want human effort and computer power to intermingle, and so there would need to be a way for human contributions to be included. For example, if you wanted to add a particular section to a portal, or do most of it yourself. The tool could go off of existing portals, where they exist, augmenting those, and generating portals from scratch for subjects for which they don't exist. Or, users could control portals by editing the underlying data sources, perhaps with some special ones reserved for users.

Also, it's not feasible to do a portal for a stub article's title. So, there would be some formula for deciding whether a topic is extensive enough to even have a portal. See Portal:Quidditch for the lower end of the spectrum.

I'm not sure that initially, all of the instructions would be in the program, or if much of it would be in templates that were called, or other components. For example, in Template:Box portal skeleton, "In the news", "Did you know?", and "Get involved" are all conditional, appearing only if there is underlying data to populate them. So the portals that have a correspondingly named WikiProject, get a "Get involved" section automatically.

Focus might shift from building portals to standardizing the other navigation systems (navboxes, outlines, categories, etc.) from which portals would draw their data. We've yet to crack the categories egg, but so far, outlines and navigation footer templates have proven to be highly useful for populating portals. So, it doesn't have to be a single data source type. Tables look especially juicy and ready to harvest. :)

But, we're overlooking Wikipedia's database itself. I'm sure there are ways to mine that directly, to build portals. There are a myriad of data tables in there, for example.

The point is, that it's coming. We just need to remain alert as to how we can make it happen as we go along, and so this conversation should continue. Eventually, all of the resources will be available to click into place. But, what resources?

We've cut human input time on portal creation by a factor of about 36. If we do that again, in our ongoing development, portal creation will be down to around 13 seconds of human effort, for each portal. And from there, we'll probably see how to drop it by a factor of 36 again. :)

The fun part is, that we haven't even scratched the surface on the features that can be provided in this medium/venue. The future appears very bright for portals. (And for the other navigation systems, as much of this technology will be applicable or adaptable to them, as well. Self-building outlines, indexes, glossaries, books, navigation boxes, lists, tables, and categories, would be nice, for example).    — The Transhumanist   22:31, 14 August 2018 (UTC)

The instant portal and creation criteria

One concept we have discussed, toward which we are approaching fairly quickly, is the capability to provide a button that generates a portal at the time you press the button. No stored page required. Pbsouthwood referred to them as "quantum portals". My guess is that this isn't more than a few years away, maybe as little as one. To allow editor contributions, we would need to have a parameter page though, or some other way of storing editor input. Like if you wanted to add pictures to an image slideshow. And differences in style could be handled on a settings page somewhere, like on the preferences page, including randomizing styles. I wonder, where would the bar be for quantum portals?    — The Transhumanist   21:36, 27 September 2018 (UTC)

The current external pressure may force some accelerated development. I was hoping for a less forced development process where more options could be explored, but reality is, and we must live with it. We might already have the tools to do it, but not as well as I was hoping for.

I would expect the bar for quantum portals to be reasonable. It must be appropriate to an automated process. The existence of a suitable navbox will allow a reasonable quantum portal to be autogenerated from a template. If the creation criteria are specified carefully it will help with refining the main template. Some of the box section templates will probably need tweaking. They probably need that tweaking anyway. · · · Peter (Southwood) ^(talk): 07:07, 28 September 2018 (UTC)

At this stage, I would guess that the existence of a navbox might be sufficient. A quantum portal would only be rendered for someone who is interested enough to press the button. On the other hand, I currently support the notion that for every navbox there should be an equivalent category (until I am shown reason why this would be a bad thing). However, navboxes are easy to use as portal frames, and categories are not, so I would be inclined to leave the category out of the quantum portal criteria if only for this reason. For semi-automated portals I would be quite happy to see a requirement for an equivalent category, if only because it would encourage their creation. There might be some conflicts where navboxes and categories are named inconveniently, and I have seen cases where the navbox could have been named better. · · · Peter (Southwood) ^(talk): 09:51, 28 September 2018 (UTC)

Activating a quantum portal

I'm thinking that a menu item could be used to 1) jump to the corresponding portal if there is one, and if there isn't, 2) generate one on the fly and display it. Or it could be a button automatically displayed on the current page. Or a hot key. Or all three. To solve the problem of pages without enough support material to generate a portal, a list (similar to the AWB checkuser page) could be built with subject names known to have the requisite support. Unfortunately, that wold only work for a few thousand subject names, but it would provide a start.    — The Transhumanist   09:10, 7 February 2019 (UTC)

• I like the option of using a hard portal if it exists. So will everyone else who builds them with care and attention to detail.
• Which menu? Problem with menus is they tend to disappear from view as you scroll.
• Buttons make it easy to use. Hotkeys not so much. How will the user know there is a hotkey, and what it is? Hotkey is good when you know the system, and does not scroll away. Combination may be best. It should not complicate the software much to activate by more than one method.
• Navboxes seem to work fairly well for the current system. Some pages have no navbox which is a problem. Others have more than one navbox, which is a different problem, possibly easier to work through. If there are two or more navboxes give the user the options of which to make the portal from. It would also be possible to just generate a portal using all the navboxes, but that could be a large portal, with lots of not very relevant articles, so my first impression would be a choice of one from a list based on the navboxes.
• Solving the no-navbox problem is labour intensive at this stage. I see no obvious way to get around it, however, navboxes are a useful alternative way to navigate - I use them a lot.
• Index and outline lists are potentially useful, but probably not enough of them to make much of a dent.
• Categories are a possible route, but I doubt if they are sufficiently focused and reliable.
• we need to consider possible scaling problems as Wikipedia grows. We already have to deal with 5.n million articles, many hardly categorised, and many with not very informative titles. It may be possible to use the not yet common short descriptions in some way, but maybe that is just wishful thinking.
• I will invoke the unspeakable, and suggest that there might be something on Wikidata that could be used. No idea what, but it does seem the sort of thing that WD might be useful for some day.
• What is an AWB checkuser page? · · · Peter Southwood ^(talk): 11:35, 7 February 2019 (UTC)

@Pbsouthwood: I was referring to the tools menu on the sidebar.

No navbox = menu item does not display.

Now that we have User:DannyS712 on the portals team, it might be possible to have a nav footer building script that populates the footer from a CatTree.

Wikidata is an amorphous mass. It seems like it would be ideal for recording what topic belongs to what topic, but I can see no rhyme or reason to it at this point. It's almost like a bunch of JavaScript objects jumbled together.    — The Transhumanist   01:14, 14 February 2019 (UTC)

P.S.: The AWB checkpage is the list of users approved to use WP:AWB. AWB won't edit WP unless you are logged in and are included on that page -- the origram checks that page before it executes any search/replaces. WP:JWB, which is written in JavaScript, also checks that page.    — The Transhumanist   01:21, 14 February 2019 (UTC)

The Transhumanist, Sidebar is a practical place from a wikipedian's point of view, but would the target audience know about it and what to do with it?

I look forward with great interest to what DannyS712 can extract and build, as I have no idea of how they would do it.

Wikidata has great potential, but as you suggest, may not be ready yet, or may require a fairly complex tool to be useful for this purpose. We walk the road of the adjacent possible. It has no long-distance forward-looking map, but sometimes we can see something close enough to build a bridge. · · · Peter Southwood ^(talk): 09:29, 14 February 2019 (UTC)

Would the portal checkpage be a list of topics for which a portal exists, for which a portal could be created because the framework exists, both, something else altogether? · · · Peter Southwood ^(talk): 09:32, 14 February 2019 (UTC)

The menu item could be named so that it is self-explanatory, like "Generate portal".

Danny is building category harvesting tools. In a script, you can harvest a category with https://www.mediawiki.org/wiki/API:Categorymembers .

Wikidata is convoluted. The items are not structured uniformly enough for us to easily construct a nav footer building script using them. See the entry for car, for example: https://www.wikidata.org/wiki/Q1420 .

A portal checkpage could be a list of portals that meet whatever criteria we decide. But, I was thinking "creatable portals", that is, subjects that would result in a quantum portal when you push the "Quantum portal" button.    — The Transhumanist   09:55, 14 February 2019 (UTC)

First thing to do is list what makes a portal createable.

We know that a big enough navbox is good.

An index list should be workable, and a category might be workable. What else? · · · Peter Southwood ^(talk): 10:51, 16 February 2019 (UTC)

@Pbsouthwood: Trying out the prototype that is posted on your talk page would be a good next step. It might provide further insight.    — The Transhumanist   02:51, 17 February 2019 (UTC)

The Transhumanist, Been there, done that. Seems to do what is expected. Only ran a few runs, but noticed that the image slideshow module does not like an empty set of images (gives a lua error message). Try making that box only display if there is content. That would be a good feature for it for one page portals anyway, as some types of article seldom have an image (law for example), and this is likely to come up a lot with QPs · · · Peter Southwood ^(talk): 04:35, 17 February 2019 (UTC)

Working on that: an automated version of placing a "2" at the end of "bpsp". Try it manually, to get a feel for what I'm after. A "3" ditches the subtopics section, a "4" removes the category tree section, and "5" adds "topics" to the end of the template name in the Selected general articles" section. I've just started adding combos; "2-3" removes the image slideshow and subtopics sections. To see the selection of templates, do a search for "Template:bpsp", and see what comes up. I think I can further develop the script to adjust "bpsp" by adding the appropriate number based on scraping the HTML for lua error messages. The main problem I foresee with this approach is the multiple 10-second build times to display the portal.

Thoughts?    — The Transhumanist   04:50, 17 February 2019 (UTC)

Potential implementation approaches

@Pbsouthwood: I can think of two. Feel free to add others...

JavaScript rendering on a blank page

WP:JWB does this.

Thoughts?    — The Transhumanist   16:00, 15 February 2019 (UTC)

Custom special page

See https://www.mediawiki.org/wiki/Manual:Special_pages#Custom_special_pages

Thoughts?    — The Transhumanist   16:00, 15 February 2019 (UTC)

Seems a reasonable option as special pages tend to be volatile, as quantum portals would be. — Preceding unsigned comment added by Pbsouthwood (talk • contribs) 03:05, 16 February 2019 (UTC)

Toolforge tool

A tool on Toolforge would give a lot more flexibility (choice of programming languages, using libraries, html structure and css), and allow links to virtual portals to be shared. Something sort of like Reasonator, but based off Wikipedia articles/categories/navboxes, and perhaps formatted more like a wikipedia portal. - Evad37 [talk] 02:47, 10 March 2019 (UTC)

Navbox builder script

<brainstorming> A script that pulls out the content of a category and buids a navbox from it could be useful. Could also be tricky, and would have to be told how many levels to go down the rabbit hole or it might not get back this week. The product would have to be human inspected and probably in most cases have most of the content deleted as not relevant, but worth a look, I think. Maybe each time it is ready to look at a subccategory it should ask for confirmation [include/skip], and maybe after it has listed a subcategory there shoud be another retrospective [include/skip] option.</brainstorming> · · · Peter Southwood ^(talk): 04:47, 17 February 2019 (UTC)

One possible approach here is to build a tool to place hidden tags pointing to categories. Then the category harvester comes along and reads those tags, replacing the current content in a particular section (the previous fetch) with the current membership of the category specified in the tag in that section. JL-Bot uses a similar approach with maintaining Recognized content sections.

Thoughts?    — The Transhumanist   04:58, 17 February 2019 (UTC)

Hidden tags where? · · · Peter Southwood ^(talk): 12:36, 17 February 2019 (UTC)

You place a tag in the form of a standard-formatted hidden comment before and after the location you want the material to go. JL-Bot tags look like this:

<!-- Start of content generated by JL-Bot -->
<!-- End of content generated by JL-Bot -->

Naturally, tags for placing category members would include the category name, so that the bot or script would know where to get the content.    — The Transhumanist   23:27, 17 February 2019 (UTC)

OK, I have used that before. It seems to work well enough for some applications. Categories in general tend to be a bit less predictable regarding content than some of the more specialised maintenance categories. I don't know if this would be sufficient filtering to get generally usable lists. I suspect that most of the time some human oversight will be necessary. Could still be a useful tool. · · · Peter Southwood ^(talk): 09:29, 21 February 2019 (UTC)

Shouldn't this be "Virtual portals", rather than "Quantum portals"

As a quantum mechanic, I think this task, even if it were well-defined, more resembles virtual particles than quantum mechanics. — Arthur Rubin (talk) 22:30, 9 March 2019 (UTC)