Sunday 21 September 2014

Exploring Office Graph and the Graph Query Language

So you must have heard about the Office Delve launch recently. I have been trying out my hands at the Office Graph which powers Delve and the Graph Query Language which can be used to get data from the Office Graph. To get a comprehensive idea of how it all works, follow these links:

Developing Apps against the Office Graph - Part 1
http://blogs.msdn.com/b/richard_dizeregas_blog/archive/2014/09/15/developing-apps-against-the-office-graph.aspx

Developing Apps against the Office Graph – Part 2
http://blogs.msdn.com/b/richard_dizeregas_blog/archive/2014/09/17/developing-apps-against-the-office-graph-part-2.aspx

Using Graph Query Language (GQL) with the SharePoint Online Search REST API to query Office graph:
http://msdn.microsoft.com/en-us/library/office/dn783218(v=office.15).aspx

So basically we can say that the Graph Query Language can be used to filter normal search queries to return content which is highly relevant to the user. I think the most important thing about the GQL is the EdgeWeight property which can be used to return content sorted by it's "closeness" to the user. This 'magic' property will also return a relevance score of how close the object (content) is to the actor (user).

For example, if  I make a query to get all the people with whom the current user closely works with, the person who will have the highest relevance score will be the person with whom the current user works with most closely. The next person will have a slightly lesser relevance score and so on. This will also be true for other queries such as documents which are most relevant to the current user. This data can be very useful and can be used in a variety of scenarios.

Here is an example of a query to get the users with whom the current user works closely. I have modified the code from Richard diZerega's blog which is linked above.

2 comments:

Geetanjali Arora said...

Great Article !!! Very informative. Thanks Vardhaman for sharing this information.

Brent Ellis said...

Is it possible to do this for colleagues that you are currently working with (updating over time) versus closeness (which I think is all time)?