The first step in creating a Wikipedia article is to search for information already there. The topic, I have chosen is “Learning from demonstration”, and former Wikipedia authors have uploaded some content to the subject. With the internal searchbox, the websites are identified and read carefully, because we do not want post the same information twice.
After reading the given information it is obvious that the existing article https://en.wikipedia.org/wiki/Apprenticeship_learning fits at best to the “Learning from demonstration” topic. It is focused not on program synthesis, but on robot-control and mentioned vocabulary like “inverse reinforcement learning” which is the term used in the literature too. Adding the new information to our notes makes sense:
A view on the pageview statistics shows, that the number of daily visits is very small (35 visits per day). Such a niche topic is a good starting point for the first edit in Wikipedia, because the chance for conflicts is small. That means, it is likely that the Wikipedia admin let us playing around with the text, because this article is not under fire.
The next question is, how to bring our own content into the article? Deleting all the existing information and write our own text from scratch is not a good idea, because previous authors have the article perhaps on their watchlist and will protest if we are doing so. What is always possible is to add new information in a soft-mode, that means we are creating a new section with the markdown syntax “# section” and write down our needs.
Before writing our text, some literature may be helpful. The existing references in the Wikipedia article are not enough, so we must add some new papers. In the best case, we have already a literature list in the bibtex-format, which we found useful. Here is the list:
In theory, the above cited literature must be formatted in a special way to match with the Wikipedia-syntax. In reality, this is one of the minor problems. It is explained somewhere in the help section, how a literature template is used right. So I can leave out this step and focus more on a content level. We have the above cited 5 sources, which are useful for describing the topic. Reading again the papers is useful and helps us, to make some notes about it, which we can later extend to a text for Wikipedia.
After reading the 5 papers again, we find that two of them a similar and one is boring. The list can be reduced to only 3 papers which are good, and our readers have less information overload. A second advantage of reading the information again, is that it is very clear what “Learning from demonstration” means, and we can write down a prototype text. But where to start, how to lowering the entry barrier? The best advice is to imagine that not Wikipedia is the target but we want to write a comment! for a Youtube! video. Everybody knows, that comments there are not very serious and it is more important that they are colloquial. That is exact the writing style a good academic paper should have.
Writing a first draft version was surprisingly easy. Because it is formulated as prose and not as a scientific paper. Making a complex academic article from it, is only a formal question. This is realized by adding references to external literature. That is the main difference between prose and science. From the text itself, there is no difference. According to the filesize, the short text in the screenshot is around 2400 bytes long, which is compared to the average edit in Wikipedia very long. Most edits there are not longer than 500 characters.
Until now, it is unclear if many small edits or one big edit is the better choice for interacting with the encyclopedia. But one advice is very clear, that making an edit to fast is wrong. Our above prototype text is not ready for uploading to the internet. It has some spelling mistakes, the cited literature is in the wrong format and some aspects are missing. In theory it is possible to upload even draft-content and edit it on-the-fly. But testing out the tolerance of the wikipedia admin is not the best idea. Instead the novice text-creater should do writing a nearly perfect text on his local harddrive and upload only the final version. So we must postpone the interaction with Wikipedia a bit, and improve first our text.
The good news is, that the article is not very often changing. The last edit was 3 weeks ago, and last year a period of 6 month was there, in which no edit took place. So it makes no matter, if we are uploading our text today, or in one week. Wikipedia can wait.
The next step is to reformat our prototype text into the wikisyntax. This has to be done with the text itself, which can be enhanced with so called Wiki-links for referencing to articles already there and with a citation template for making the reference list right. The result can be controlled in the Wikipedia sandbox.
Even our text is short, the formatting is surprisingly complex. Until all the keywords are referenced and a new created table is in place it takes some time. A nice sideeffect of the formatting is, that our prose text which was originally targeted to a youtube audience looks now more professional. It consists of the same words, but this time they have literature references and clickable links.
After doing some improvements in the literature list the version in the sandbox looks like in the following screenshot.
The article itself is ready. Until now, only the sandbox was aware of it. Now it is time to getting the real Wikipedia informed. That means, we are copy & paste the 4kb sourcefile into the article which is already there and wait, what will happen. The opponent known as “Wikipedia admin collective” has many options. He can delete our text in total, he can delete minor parts of it, or he can accept the text, that means it will happen nothing. It is hard to guess, what the Wikipedia admin will do. In my opinion, the admin will perhaps edit the English language a bit, because I’m a native speaker and it is very likely that some grammar mistakes are there. But it is also possible, that Wikipedia thinks, that my text is not scientific, and that the referenced literature makes no sense.
That is exactly the feeling which I’m calling in the beginning “fear”. That means, the author of new uploaded content doesn’t know, what the opposite will do. He has to loose a lot. On the other hand, it is payback time, and the text will go online under any conditions. Do not playing the game, is not an option!
After saving the changes in the real Wikipedia, the Revision history is updated and our edit is placed on top.
If we are trying to identify the edit in the global list of Recent changes, we will notice that our edit is only one of many hundreds. The updates are coming very fast and we must scroll a bit to find our posting.