Guidelines

• Please submit a QID CSV file containing a list of seed entities you are interested in. Each line should consist of a single QID, which is the identifier of an entity on Wikidata. For instance, Q142 represents France on Wikidata.
• Please also submit a PID CSV file containing a list of properties you are interested in. Each line should consist of a single PID or (-)PID, where PID is the identifier of a property on Wikidata. For instance, P31 represents the property 'instance of'. A PID (e.g., P31) indicates that direct edges labeled with the property will be traversed. A (-)PID (e.g., (-)P31) indicates that inverse edges labeled with the property will be traversed. You can indicate both a PID and (-)PID in a file if you need to traverse direct and inverse edges labeled with that property.
• Our model will traverse the neighboring entities of the input seed entities by following (direct or inverse) edges labeled with the properties indicated in the submitted PID CSV file. These entities will be retained or removed based on our model's assessment of their relevance.
• If you want to see the results with all the properties available in Wikidata, you can populate your PID CSV file with "*" to traverse all direct edges labeled with the properties, or use "(-)*" to display all indirect edges labeled with the properties. Please keep in mind that we enforce limited time for queries. Extensive traversals could thus lead to query timing out. To avoid such limitations, we recommend that you deploy your own instance of KGPrune. Please contact kgprune@inria.fr for assistance in finding a customized solution.