2

I am a little bit confused why I get such low confidences with the REST API of Spotlight.

I run the following code/call:

curl http://spotlight.dbpedia.org/rest/annotate \
 --data-urlencode "text=Hallo, i am looking for a new Microsoft or Apple tablet." \
 --data "confidence=0.2" \
 --data "support=20"

then I just get similarityScores between 0.130 and 0.135.

When I try this text in the demo (http://dbpedia-spotlight.github.io/demo/) then I can get the results until to a confidence of 0.60. What cause my low confidence values? How can I obtain correct values?

p.s. Is there anywhere a more comprehensive documentation?

Gassa90
  • 80
  • 7

1 Answers1

1

First similarityScores does not correspond to the values of Confidence. But more importantly http://spotlight.dbpedia.org/rest/annotate is not the endpoint being used by the demo.

The endpoint being used by the demo is: http://spotlight.sztaki.hu:2222/rest

David Przybilla
  • 830
  • 6
  • 16
  • thanks for sharing this information. Is their some more documentation available to get an understanding how the level of Confidence is calculated in spotlight, especially in the direction of disambiguation? – rochbenritter Mar 21 '15 at 09:59
  • 1
    Probably this mail will help you clear it out: http://sourceforge.net/p/dbp-spotlight/mailman/dbp-spotlight-users/?page=2 – David Przybilla Mar 21 '15 at 14:53