0

I am looking to search for influential people from Freebase database. I have tried to look at this but I can't get my head around developing the query to get a bug-free result. Here's the link: www.freebase.com/influence/influence_node.

Brian Tompsett - 汤莱恩
  • 5,753
  • 72
  • 57
  • 129
Ali Gajani
  • 14,762
  • 12
  • 59
  • 100

2 Answers2

1

A good start would be to click on the "Instances" tab at the URL you gave, then click on "Build Query" from the gear dropdown at the far right of the resulting page.

That will give you access to the network of influencers and influences, but you'll need to figure out how to weight the raw results because not all influencers are created equal. Some are master's thesis advisor's, others "influenced," in some undefined way, an artist's style, and I'm sure you can imagine many more variants -- each having more or less weight depending on the values which are important to your particular study.

Another factor you may want to include (or not) is what Google things about the ranking of specific entities in terms of importance or relevance. The Freebase search, with the appropriate option settings, would be one way to incorporate this, albeit in a black box sort of way without any access to why they thing something is more or less important.

Tom Morris
  • 10,490
  • 32
  • 53
  • Yes, I agree, but this query only gives a list in JSON, I would like a table, two column table with influenced and influenced. How would I go around doing that? Thanks Tom for your help. Moreover, the result is very small, 30-40, I expected thousands. http://www.freebase.com/query?autorun=1&q=%5B%7B%22id%22:null,%22name%22:null,%22type%22:%22/influence/influence_node%22%7D%5D – Ali Gajani Dec 21 '13 at 09:24
  • Looking at the Instances tab shows over 50,000 topics, so I suspect there's a problem with your query. JSON is the APIs native (and only) format, but it's trivial to turn the results into a table. https://www.freebase.com/influence/influence_node?instances= – Tom Morris Dec 21 '13 at 18:10
  • My query is the one you instructed me to do earlier, by going to the Build Query feature. What is the query that you used? Thanks and sorry if I come across ignorant, I am new here and learning. – Ali Gajani Dec 21 '13 at 20:44
  • @AliGajani if I understand correctly, you want the edges on the influence graph to be the rows in your table. MQL AFAIK only returns entities (nodes) as result, not edges. So you will have to transform the result yourself. – Falcon Dec 29 '13 at 17:37
1

To get started, you can get the count of all influence nodes on Freebase via this query: 

{
  "type": "/influence/influence_node",
  "return": "count"
}

(which returns 50856 for me just now)

To get the graph structure, you will need to first query the influence nodes and their neighbors like this: 

[{
  "mid": null,
  "name": null,
  "type": "/influence/influence_node",
  "influenced": [{
    "mid": null,
    "name": null,
    "optional": true
  }],
  "influenced_by": [{
    "mid": null,
    "name": null,
    "optional": true
  }]
}]

then transform the resulting list of nodes into edges by using the mid field (machine id, one of several unique ids on Freebase). Note while Freebase returns only the first 100 nodes, you can page through the result by using the cursor parameter in the mqlread API. Note also that "option": true keep nodes with no influencers or influencees around in the result.

Falcon
  • 1,317
  • 1
  • 13
  • 30
  • This is good but the problem is Freebase's difficulty in getting the data prepared for Analysis. DBPedia is much easier to query and returns the data in the correct format, prime for data analysis. Freebase returns JSON, and then its confusing to see who is influencing who, in addition to getting JSON converted to table format. – Ali Gajani Dec 29 '13 at 23:23
  • yeah, Freebase API's response format is rather limited right now. I think `influenced` is the named inverse relation of `influenced_by` (i.e. A `influenced` B iff B is `influenced_by` A, so the response to my sample query contains redundant information). You might also want to consider to work directly with Freebase's RDF dump: http://stackoverflow.com/questions/17760747/import-freebase-to-triplestore – Falcon Dec 30 '13 at 02:58
  • That's correct, the two properties are inverses of each other. You can see in the schema one labeled as "master" and one as "reverse" https://www.freebase.com/influence/influence_node/influenced_by?schema= If JSON is a "confusing," I'm not sure RDF will be an improvement. :-) – Tom Morris Dec 30 '13 at 15:56