RENDER/Berlin summit 2012/frontend

Notes taken during the Wikidata/RENDER summit 2012 concerning frontend issues.

frontend analysis toolkit

edit

https://toolserver.org/~render/toolkit/index.php

(Brandon Harris, Gerrit Holz, Angelika Adam, Kai Nissen, Stephen LaPorte, later Dario Taraborelli & Erik Moeller) (notes by Sumana Harihareswara)

edit

http://toolserver.org/~render/toolkit/LEA

This is the LEA toolkit, part of the RENDER toolkit.

Need to figure out the lack of usability

What is the link extractor? Anselm authored this tool; Source code is available on toolserver. BSD license. Enter an article title & a language version of the article, and it looks up all linked lang versions, scans the content of the same articles on multiple wikis, determines which of these are the 3 articles that have the most links, compares the reference lang (En here) to those three. This is about fact coverage. Green: there is a link in this version of this article. Yellow: there's an article, but no links Red: no article. We will need to surface what these colors mean.

Brandon says: only nerds will use it.

This is a complex concept, hard to make usable. But you don't have to make it usable by novices. The audience: people who are already familiar with Wikipedia & its systems, so they are probably somewhat technically savvy

Question re expectation: this is a showcase. We don't know where to put the tools in the long run. Can reside on toolserver for the next month

We should make this data available to all the wikis. This is a microtask system! We want microtasks! Editor engagement -- the best way to become a good, long-time editor is to have something to do. These are better than redlinks. "We've noticed that on 4 wikis, this word is linked to this page; can you ...?"

If we could install this software on the clusters, we could just run a cronjob every few days to generate the corpus of this data

Good example article: Georg Friedrich Hegel?

Ideas from Brandon: clarify the sys navigation vs tool buttons

"these could be better as tabs" - flipping to different tools. but not a big problem

Have the buttons go directly to the functionality, don't do this tab thing call out the feedback thing separately vs having the icon link to it

"kill the words here and just use color"

hard to hit all 200 lang versions. Can this scale up to thousands?

Idea to have a better overview - kick out those that haven't been changed... show which lang versions have been changed?

Brandon notes: this doesn't show me an outlying pattern. How do we surface the patterns? It's ok for there to be no Eurovision Contest 2013 article in Malayalam. It won't be in 250 of the wikis. It's a problem, though, that English isn't represented

"Misfits (TV series)" - can we suss out the pattern?

This... the audience for the data -- microtask data. Excellent. But hard to figure out how to use this in a microtask system. also researchers also article readers (signal: this article is out of date!)

So we need to be able to get this data via JSON API or similar export ton of uses for this

But then this has to be cached or very fast. Right now it's run 1ce a day, cached.

Then again, in 20 weeks, the 2012 Hurricane Season article will be done forever, except for bot categorisation.

So, on this screen, de & fr .... use those codes, instead of the full names of langs, to save real estate. Brandon wants the whole span to be clickable. Just do color Whole row should subtly change color at rollover, in both directions (row & span) to help user see where she is.

Scrollable thing,

here - use the browser, let it go off the screen forever.

This - you'll get a giant table, it's acceptable.

Nice to have: AJAX/jQuery lazy loading

We don't want minor or bot edits to be analyzed.

always label the thing you're talking about via vertical stacking

Look at the MediaWiki style guide - https://www.mediawiki.org/wiki/Style_guide

Assume that at least one person from each wiki wll come and use this tool to look at this data

Very simple form! not much more you can do to improve. Eventually, a pulldown for project? pulldown for project might not work, 280 items. Maybe a predictor box that, as you type, helps autocomplete language or title. But that means you have to search on every wiki! Lang is easier. Nice to have.

Compliment on choice of article (Jerry Siegel, inventor of Superman)

This table will get huge...

in instances where everything in kosher..... (how does that look in the tool?) We probably don't need the line, because it's not actionable

Brandon: Weirdness: my brain wants to interpret what I'm seeing backwards! want to interpret that the problem is with the wikis that DON'T have an issue! because the rest of the chart is uncolored, looks like the colored part is the key! so just do green all the way across! if exists & is linked, green! then what you have is, good is green, prob is yellow

Red/yellow/green: (culturally & accessibility-wise) red is classic problem color bio-attuned to enjoy green but there's a percentage of pop that cannot distinguish between those 2. Yellow is poorly done on monitors - no yellow pixel! it's actually a very light brown You can do this ok by changing the intensity of the color. In Photoshop... color should not be your only signifier.

Green is a go color.

(more of this is in the MediaWiki style guide) Example: look at look in MediaWiki style guide for more)

For example, New Page Triage/Patrol page


info in pie chart: should be a hover

explanatory text: clarify. outbound links.

"the article .... exists in further language versions" - No way to improve the structure of the data -- once you get it, you get it


Use case: looking for discrepancies, or nondiscrepancies? only show the green? never show the green?

data analyst cares about greens; microtask, no.


What if the word "fanzine" doesn't appear in other languages?..... what I want to analyze: Jerry Siegel across all wikis.

I am working on French Wikipedia, want to see how good our coverage is, compared to other wikis.

Here's a way to get your article up to Good or Featured article status?

option - filter option, show or don't show me this.

Report: you'll get everything. AJAX to modify display of the form.

(Dario & Erik come in)

disambiguation.... What if there is no word for fanzine in German Wikipedia?

You get a result within 20-30 seconds, checking every lang version. heavyweight when it runs.

Young wikis don't have strict reference policies. Things change over time

Editorial decisions re linking generic words


Change detector

edit

https://toolserver.org/~render/toolkit/ChangeDetector/

shows activity on a certain date, compared across wikis Example: a news event -- but no one has edited the relevant article on one of the smaller wikis

Purpose: There are 10 or 12 ....

Capitalisation issue with MediaWiki - case sensitivity issue


Change detector: give people a mechanism to subscribe to a series of articles, get pinged if I need to edit it based on other wikis having updated relevant articles Tie it to the watchlist Brandon will be working on a "this page was last modified n days ago" (hotlink to history). Surface fact that encyclopedia is a living entity. "maybe I can contribute by helping keep this up to date."

See my change reflected in other langs! Brandon interested in Athena

interwiki service - powerful if done well!

Watch: do you want to watch IN ALL LANGUAGES? or just en.wp? or select: en + de but don't put too much burden on user to select a reasonable config

surface trends (sparklines?)

Gregor: from editor's perspective, important to have more than one day. Few editors check every day. 7 day range? Some measure of editing intensity

Example: Atlantic hurricane season 2012. As younger wikipedias mature, they'll create the article, move it in. English & Spanish speakers are affected.

Where's all the code? MMP under Toolserver, user "render" Changedetector code written in Python, not in public_html folder, but we could make it so

Could you put it in Gerrit?

frontend supporting tool integration (JS)

edit

(Thursday 16:15-17:45)

We will install some fancy supporting tools for the community Talking about implementation of supporting tools that have showcase roles

https://commons.wikimedia.org/wiki/File:Hackathon2012-RENDER.pdf

Article Statistics Quality Monitor & task lists.


Brandon wants to present at Wikimania about Athena, "Wikipedia 2015" - what the Wikipedia lifestyle will be in 3 years wants to include this


Article Stats Quality Monitor

edit

Bundle of things that are currently scattered on stats.se, toolserver, etc. Provide assessment analysis tools, scores Good idea, Brandon says!

We need to dumb down or de-emphasize some things for this audience. Which of these are most important? Order them accordingly.


Part of Global Profile: "I worked on these articles and 14 million people have seen my edits!"

It'd be nice if you could slide by # of changes since a date.

"RENDER" - name that others won't understand. Display name: human-readable.


"Quality monitor" - consumer, not a publisher, with the exception of generating stats.


In the future, we will de-emphasize tabs; we have a poor tab metaphor. 2 main tabs plus subtabs are all actions on the thing, vs separate actions.

Brandon like stats that imply this is alive.

How do we make this useful for normal readers?

Task lists

edit

"enter all your categories in the same line, separate by commas" - decreases the complexity of the task Ideal solution - how WordPress adds tags. Hit Enter, and it creates links you can click X next to.


Don't click "go" for the text. Buttons should always be a verb. Verb should be what you're doing. "Search"

Experienced editors have opened the toolbox on the side and it stays open forever for them, but inexperienced ones don't know what's in there.


Generate this list all the time, let other things consume it Brandon wants to see a suggestion bar



stats display vs. list of flawed articles - different audiences, diff displays

stats: reader & editor-oriented


We want to build a global profile of affiliations tool, like "I'm a member of WikiProject History" - we'll have structured data around those projects. Membership. We'll hopefully eventually have a tool, software that manages these things, so each WikiProject can subscribe to a specific list of categories (mention of Aaron Pranama's GSoC project https://www.mediawiki.org/wiki/User:Blackjack48/GSOC_proposal_for_watchlist_improvements ) Some high-traffic WikiProjects might find this more contentious, but that is ok


Brandon will have more ideas later.

Without further context, it's challenging to provide specific information about RENDER/Berlin summit 2012/frontend. However, frontend development typically refers to the creation and maintenance of the user interface (UI) and user experience (UX) aspects of websites, web applications, and software. This involves writing code using technologies such as HTML, CSS, and JavaScript to design and implement the visual elements and interactive features that users interact with.

If you have specific questions or need further information about frontend development, the RENDER/Berlin summit, or any related topics, please feel free to provide additional context, and I'll do my best to assist you.