Unrelated interactions being returned for a specific CSDL


#1

Hi. We are having issues with a CSDL which is returning non related tweets. Specifically, the CSDL is this:

( (((interaction.content contains_any "twister,#twister")) or ((interaction.content contains_any "cyclone,#cyclone")) or ((interaction.content contains_any "circulation,#circulation")) or ((interaction.content contains_any "condensation funnel,#condensation funnel")) or ((interaction.content contains_any "multiple vortex,#multiple vortex")) or ((interaction.content contains_any "non-supercellular,#non-supercellular"))) ) and interaction.content CONTAINS_ANY "Tornado, #Tornado" and interaction.type contains_any "facebook,twitter"

As you can see, the CSDL requires that every interaction contains the word "Tornado" or "#Tornado". Nevertheless, we received the following interaction:

{ "demographic":{ "gender":"female" }, "interaction":{ "source":"twitterfeed", "author":{ "username":"CarliPerilloZbo", "name":"Carli Perillo", "id":540267571, "avatar":"http://a0.twimg.com/profile_images/1990317837/1292778333kv1111_normal.jpg", "link":"http://twitter.com/CarliPerilloZbo" }, "type":"twitter", "created_at":"Tue, 17 Apr 2012 12:22:34 +0000", "content":"Sunbrella 39051 View 60 In Silver Outdoor Fabric: Sunbrella outdoor fabric is the premium outdoor fabric for awn... http://t.co/dwXoqstr", "id":"1e1888802b26a100e074bb3ff12a9c6c", "link":"http://twitter.com/CarliPerilloZbo/statuses/192226544339529728", "tags":[ "402489347" ] }, "klout":{ "score":15 }, "language":{ "tag":"en" }, "salience":{ "content":{ "sentiment":4 } }, "twitter":{ "created_at":"Tue, 17 Apr 2012 12:22:34 +0000", "domains":[ "amzn.to" ], "id":"192226544339529728", "links":[ "http://amzn.to/nWlGjH" ], "source":"twitterfeed", "text":"Sunbrella 39051 View 60 In Silver Outdoor Fabric: Sunbrella outdoor fabric is the premium outdoor fabric for awn... http://t.co/dwXoqstr", "user":{ "name":"Carli Perillo", "description":"Are you off your rocker?", "location":"Amarillo, TX, United States", "statuses_count":15719, "followers_count":274, "friends_count":3, "screen_name":"CarliPerilloZbo", "lang":"en", "listed_count":3, "id":540267571, "id_str":"540267571", "created_at":"Thu, 29 Mar 2012 20:49:11 +0000" } } }

Seems like the issue is somewhat related to tweets containing links to Amazon because we've noticed that those we found to have the issue have a link to amazon (could be a coincidence though).

Best regards.

Rodrigo


#2

This seems a little odd. If you could create a ticket on the DataSift Support site, and include your DataSift username in the ticket, I can look into this further for you.