Wikimedia wants to make it easier for you and artificial intelligence developers to search through its data

The late English writer Douglas Adams is known as the 1979 book author Hitchhiker guide to the galaxy. But there is much more for Adams than what is written in Wikipedia entry. Whether you are or not Need Let’s know that it has Birth mark It is the whale or those libraries all over the world. Its books are stored under the same series of numbers – 13230702 – You He can If you go to The angle that was overlooked From the Wikimedia Foundation called Wikidata.

There, pictures, texts, major words and other information related to Adams are both in A. web page For robots between us, in coordination designed for machines such as Jason.

Now, Wikidata gets a new database friendly to Amnesty International that makes it easy for large language models to eat information. The database comes from Wikipedia included project From the German chapter of the Wikimedia Foundation, Wikimedia Deutschend, which oversees Wikidata. The Berlin -based team spent its headquarters in the past year using a large language model to convert the 19 million posts into Wikidata from the fully organized data into vectors that capture the context and meaning on the introduction of Wikidata.

In this veiled coordination, it is best to imagine information such as the graph with points and interconnected lines – Adams will be linked to “man” as well as the addresses of his books, Lydia Benchcher, and Wikidata. freedom.

While the user experience in the front end will remain the same as it is, no, Wikipedia no The project leaders say that the back end will become easier for artificial intelligence developers to reach when building, for example, their Chatbots using data.

Benchcher said that the goal of the project is to settle the stadium for the Monied Big Tech. Companies such as Openai and Anthropic have the resources to overcome Wikidata, just like Pintscher and their team. They are smaller clothes that benefit more than new access to coordinated data stored in Wikidata’s cellar. “Really, for me, it comes to giving them this edge and giving them at least an opportunity, right?” Benter said.

It indicates Referees For example, the wide Wikidata data project sponsored the volunteers for good. The statute allows users to find social media handles and email messages for government officials worldwide.

Most AI Chatbots gives priority to common words and topics online. In addition to giving a little technology, the team hopes to access Wikidata to artificial intelligence systems that better reflect specialized topics that are not widely represented online, Benter said. “This may be a better way to enter information in ChatGPT, for example, from” generating a lot of content and then waiting to return Chatgpt, and perhaps, or perhaps, taking into account what it contributed to. “

In practice, artificial intelligence systems will allow better access to the context about information as well as the information itself. freedom.

The team used a model from Ai Jina Ai to convert structural Wikidata data, which was captured until September 18, 2024, into vectors. IBM Company Datastax currently provides infrastructure to store the database for the project for free.

The team is awaiting notes from developers who use the database before updating it with the information that was added during the past year. Although the current database does not include completely new information that was added last year, Saadé says small modifications or modifications to the current Wikidata will not reduce the benefit of the database. He said: “At the end of the day, it is similar to the vector in which we calculate is a general idea of ​​an element, so if some small liberations are made on Wikadata, the matter will not be highly related.”