Abstract Wikipedia team/Ways you can help

The Abstract Wikipedia team is working towards a grand vision. We can't – and won't – be able to work on this on our own. We need help, be that from Wikimedia editing volunteers, from people outside the community keen to help with what we're building, from organisations that might want to support our research, and from dozens of other groups.

Here are some ideas about how you might get involved:

Abstract Wikipedia

Knowledge coverage

What is the most important knowledge that Abstract Wikipedia can tackle? As with every volunteer-lead project, we cannot tell our volunteers what to focus on. But given the already existing languages and articles in the different language Wikipedias, which articles would be the most important ones to cover in order to achieve an equitable common baseline of knowledge.

Which languages are most needed?

Not all languages have an equally large impact. Some languages are spoken by more people than others. Many people speak several languages very well. We want everyone to have access to all knowledge. People who already speak a language very well that has comprehensive articles will not benefit from making the same articles available in another language they speak. We need to gather data that tells us which languages would be most impactful to allow the biggest number of people access to knowledge.

Which articles are most needed? How complete are these articles?

What articles are considered the most crucial in order to provide a common baseline of knowledge? How well are these covered in the languages that we identified? Not just in terms of "it has an article", but how complete are these articles?

Constructor coverage

For Abstract Wikipedia, we will need a set of constructors that will be used to create the abstract content. What is a good set of constructors to start with? How can we even tackle this question? How much important knowledge can we cover with these constructors? How flexible is their use?

It would be great to build towards a set of constructors and compare with the existing articles how much of the knowledge in those articles these constructors would cover. This can be either done through Wizard of Oz style manual analyses, or through clever application of parsing methods and pattern detection.

Constructor universality

How universal are the constructors? Given a set of constructors, how well can we generate text in the relevant languages?

Wide-coverage simple articles

Besides the manual, sentence by sentence creation of abstract content, Abstract Wikipedia is also a way to support the mass creation of articles based on the type of the topic they represent, e.g. mass create articles for species, locations, historical people, etc. How many articles could we cover? How does this compare to existing efforts in various language editions of Wikipedias? How can we move from the current system of bots running on individual Wikipedia to a Wikifunctions-supported system that allows everyone to contribute to that workflow?

UX for abstract content creation and editing

How could a UX that allows a wide number of contributors to contribute content to Abstract Wikipedia look like? Can we imagine something like a ML-based parser that guesses the right constructors and allows for manual approval and refinement? Does this work across languages? Can we create a system that makes it easier than today to contribute to Wikipedia?

Pipeline for ingesting content

Can we use a pipeline to ingest Wikipedia articles wholesale? Point to an article and generate abstract content?