Main Page

Welcome to QAWiki

QAWiki is a repository of questions and queries that anyone can contribute to!

QAWiki collects questions in different languages, and queries over knowledge bases like Wikidata that can answer them.

Questions are typically factual so that they are answerable by knowledge bases. As some examples:

• "Which city has the most cathedrals?" (Answer on Wikidata)
• "What diseases are associated with the gene FGF14?" (Answer on Wikidata)
• "In which museums can I see paintings by El Greco?" (Answer on Wikidata)
• "Who is the youngest head of state in the world?" (Answer on Wikidata)
• "What was the last black and white movie to win an Academy Award for Best Picture?" (Answer on Wikidata)
• "How many dwarf planets orbit the Sun?" (Answer on Wikidata)

Here you can query for all the questions (and queries) collected thus far by QAWiki.

Why QAWiki?

Knowledge bases like Wikidata offer a wealth of knowledge that users (in theory) can use to answer all sorts of interesting questions/queries. But in order to take advantage of this power, users must typically have expertise on query languages like SPARQL, which excludes a lot of people. If we could build good Question Answering (QA) systems that directly answer a user's question expressed in the (natural) language they speak, this could bring the benefits of Wikidata to a much wider audience. And with advances in Deep Learning, such a system is within reach. The main obstacle is the need for data to train such systems. While some public QA datasets do exist, they are static, may have errors, may lack diversity (being automatically generated), etc. (Probably the best QA datasets are proprietary, used internally by companies to build applications like Alexa, Siri, etc.)

The main goal of QAWiki is then to offer a platform for a rich and open QA dataset to be built collaboratively: a dataset that captures the interests of diverse users, being vetted, extended and improved by the community, and being made openly available for the community to use. (QAWiki data are licensed under CC0.)

Systems using QAWiki

A key ingredient in order to bootstrap QAWiki is to have applications using QAWiki. This will hopefully encourage users to improve such systems by adding questions and curating QAWiki, seeing their work in action. Here's a list of systems currently using QAWiki:

• Templet: A system supporting template-based question answering over QAWiki (see example). If you add or modify a query in QAWiki, you can load it in Templet by copying https://templet.dcc.uchile.cl/update_template/Q169, replacing Q169 with the ID of the question on QAWiki. (Queries are refreshed from QAWiki on an hourly basis.)

New systems using QAWiki are very welcome!

How can I contribute?

There are many ways in which you can contribute to QAWiki (whether or not you are familiar with SPARQL, Wikidata, etc.)

Suggest new questions

QAWiki welcomes new questions relating to your interests! Other editors can then hopefully suggest a query, add mentions, translations, etc.

1. Create a new item.
2. Add the question in the language(s) you prefer as the label of the new item.
3. Optionally add other ways to ask the same question as aliases.
4. Add the claim that the new item is an instance of question.

If you're looking for some tips, see the following guide.

Translate questions

QAWiki aims to be a multilingual resource, so translations of questions and other elements into the languages you speak is very welcome!

1. Use the following query to find questions without a translation (the example is for French, but you can change the two letter "fr" code to any language).
2. Visit the questions you like and translate the question to your language.

Add mentions for entities

Adding entity mentions is much appreciated, and will help systems using QAWiki to be more precise.

1. Use the following query to find questions ordered by the number of entity mentions annotated for a language (the example is English, but you can change the two letter "en" code to any language).
2. Visit the question(s) you like.
3. Add new values for the mention property in the language of your choice. These should be substrings of a question or alias in the same language.
4. Add one or more qualifier values for the Wikidata item ID property indicating the item on Wikidata that the mention(s) refer(s) to in the specific context of the question.

If you're looking for tips, see the following guide.

Add mentions for properties

Again, adding property mentions is very welcome, and will help systems using QAWiki to be more precise.

1. Use the following query to find questions ordered by the number of property mentions annotated for a language (the example is English, but you can change the two letter "en" code to any language).
2. Visit the question(s) you like.
3. Add new values for the mention property in the language of your choice. These should be substrings of a question or alias in the same language.
4. Add one or more qualifier values for the Wikidata property ID property indicating the item on Wikidata that the mention(s) refer(s) to in the specific context of the question.

If you're looking for extra tips, see the following guide.

Add queries to answer questions

If you who are familiar with query languages like SPARQL you can also provide queries to answer questions on the Wikidata Query Service.

1. Use the following query to find questions without a corresponding query.
2. Visit the question(s) you like.
3. Add the query to solve the question via the property query.
4. Add the qualifier dataset to indicate Wikidata and the qualifier query language to indicate SPARQL.

If you're looking for extra tips, see the following guide.

Other ways to help out

• You can also build tools and systems using QAWiki and add them to the list. (You can also consider adding a link back to QAWiki for users to add questions that your system struggled with.)
• Try out some of the queries, and if you see some dubious or missing answers, edit the query or Wikidata to improve those answers.
• You can help to curate QAWiki. See this list of quality-control queries that help to detect potential issues to correct.
• ...