Moving forward with Amir's "Internal Links in #Wikipedia" presentation

At Wikimania 2019, my friend Amir presented "Internal Links in Wikipedia". It provides a wonderful expose of what is problematic with the existing functionality with blue and red links in all our Wikipedias. At the end of 2024, technically things have moved forward, this blog post's intention is to provide arguments what a local Wikibase for wiki links will bring to both editors and readers and why it does not need to be controversial. By definition, changes made to Wikipedia are controversial. 

Functionally, every link red or blue should remain exactly as is. Technically, every blue link refers to one article and every article SHOULD have an item at Wikidata. Every link, blue or red, may be referred to from many places and SHOULD be about only one concept. For every destination there MAY be a link to an item at Wikidata. At this time we have no way of knowing if there is only one concept and if there is an item at Wikidata for that concept.

Many years ago Wikidata solved a similar problem. Wikidata was an instant success because it replaced the interwiki functionality. The solution proposed today is similar and only possible now that Wikidata can be "federated" with many instances of a Wikibase. 

All destinations for both red and blue links will be known in a local Wikibase federated with Wikidata. Any destination may be linked to a Wikidata item but the name of the local article/destination will remain unique. Thanks to this federation, disambiguation support may be provided based on what is known both locally and globally when a new link is created. It will know about the synonymy for each subject.

This change does not need to be controversial because like with the interwiki links, people can opt out of this new functionality. When only a subset of the editor community becomes involved, the quality of all links will improve quickly. With the interwiki links fixed, Wikidata was ready to become a knowledge base. As the wiki links in the local Wikibases get in shape, the Wikidata knowledge base may be used to signal that articles should be in specific categories, or that red links could be added in summation articles like in articles about an award.

Our dependence on Wikipedia editors will remain key but tools like the Wikidata knowledge base are available to bring us the data that enables us with information that is up to date and improves the connections between all our articles. Manually checking wiki links is a Sisyphean task, with tooling it becomes manageable and worthwhile.

Thanks,

      GerardM

#Wikimedia and its #BLP approach

There is a huge controversy about the policies about the "Biographies of Living People". Central in all this is that there is no such policy at Wikidata. Many seasoned Wikipedians are of the opinion that using data in Wikipedia is a violation of its BLP policy as a consequence. At the same time there are seasoned Wikidatans who oppose a BLP policy similar to the one at Wikipedia. The problem is that Wikidata does need a BLP policy but it needs to be different for various reasons.

• An item in Wikidata can be really rudimentary; Marian Latour, a Dutch author, was created because she won an award. This is allowed in Wikidata but the limited information is probably a violation of the English BLP policy. This information came from the Dutch Wikipedia. 
• The initial data of Wikidata were the interwiki links. This was a huge improvement for the Wikipedias and there are still many items that have no statements. This is used as an argument not to accept information from Wikidata.
• Wikidata data is retrieved from a Wikipedia, information like "who won an award". Given the BLP policy of that Wikipedia is should be faultless but it often is not due to disambiguation issues. 

The first issue refers to a red link on the Dutch Wikipedia. When the red link is associated with the Wikidata item, there will not be a new disambiguation issue when a different Marian Latour is introduced. Currently there is only one Marian Latour known to Wikidata.

The second issue is one where Wikidata statistics indicate that slowly but surely is adding statements. They also prove that there is still so much to do...

The third issue is the main one. When an article is linked to Wikidata, articles in other languages should link to the same item or to a red link. Solving these issues requires coexistence and preferably collaboration. 

What we need in a Wikipedia is the ability to link a blue or red link to a Wikidata item. Obviously changing links is either blatantly obvious like for Manuel Echeverria or it requires a source. Technically the necessary change in the MediaWiki software may be "opt in" so that only people who care about this approach to quality make use of it. 

As far as I am concerned, when some Wikipedians find fault elsewhere and do not reflect on this proposal and the improvements it brings them, that is fine. What is relevant is that this approach allows for the best Wikidata practices and at the same time improves the BLP quality in all Wikimedia projects.

Thanks,

       GerardM

#Quality for #Wikidata and for external #sources

There are always arguments to find why not to accept Wikidata as a quality resource. Many Wikipedians ignore Wikidata because they do not trust the quality of its data. They require sources because that is why they trust a fact in Wikipedia to be good.

The practical problem is that Wikidata has some 15 million items and most have one or multiple statements. Each statement should be sourced given the notion of sources as a requirement. Given the speed of new information in Wikidata, sourcing for all statements is not going to happen anytime soon and consequently an alternative that demonstrates quality is needed.

One best practice of Wikidata is publishing external sources for our items. It already adds a feeling of quality because it allows a person to see what those external sources have to say. It takes some software and a workflow to leverage this sense of quality and solidify it as a measurable quality improvement.

Obviously both Wikidata and external sources have their issues. Where they all agree, there is the least need to work on improving quality. Where Wikidata has no data, it is obvious to add data and use the external source as a reference. It becomes interesting when there is a difference.

The first thing to do is flag a differing statement as suspicious. It signals to software and people that there is a need for attention. People can research the issue and come to the conclusion that

• Wikidata is correct
• the external source is correct
• both are incorrect

In all these circumstances, the flag for the statement will be changed, the statement may be changed and in every case a source is to be provided. This is when true sources make the biggest difference because the flag does not go away and with quality sources where there is this obvious need, the quality of Wikidata is easier to appreciate.

Thanks,

     GerardM

#Wikidata - collaborating on data #sources

x

#Freebase's data found a home at :BaseKB and that is a great opener to an alternate approach to the Freebase data for Wikidata. This proposal has many parts that makes for great cooperation between multiple sources.

• items always have a link to its source(s)
• statements always state its source
• an indicator for the status of a statement is added

Linking to external sources is something Wikidata does a lot. It allows us to have a look at the source itself. When it is the origin of the data, it follows that the statements from that origin have to be exactly the same. When they are not, it should be indicated with a status.

When a statement differs from a source, we have identified something that needs work. In essence. such statuses have a function in a workflow as well. Because this is where intervention makes a difference. 

• we set the flag for people to investigate the difference
• we can flag the source itself that we found an issue 
• we can find alternate sources to find what is likely correct
• we can find corroboration in literature
• we change the statement when needed and set a flag to indicate the status

In this way, we spend our human capital wisely. We do not blindly spend time on "approving" any and all statements. We do it only where we know some research is needed. 

As far as I am aware, you can not copyright facts. We compare our facts against facts known elsewhere. We include our missing data where we may and we always investigate differences. We signal the work we have done and, in this way we not only improve our quality, we also provide a path to the sources we work with for them to work on their quality.

In this way we spend the time and effort of our community wisely and, we optimise the amount of information available in Wikidata. In this way everyone is a winner.

Thanks,

       GerardM