Response can contain list as surface form value

[aolieman]

Likely something that should also be fixed in Spotlight itself, but it should be easier to fix here.

File "C:\Python27\lib\site-packages\pyspotlight-0.6.5-py2.7.egg\spotlight\__init__.py", line 168, in annotate
    return [_dict_cleanup(resource) for resource in pydict['Resources']]
  File "C:\Python27\lib\site-packages\pyspotlight-0.6.5-py2.7.egg\spotlight\__init__.py", line 80, in _dict_cleanup
    clean[key] = _convert_number(value)
  File "C:\Python27\lib\site-packages\pyspotlight-0.6.5-py2.7.egg\spotlight\__init__.py", line 44, in _convert_number
    return int(value)
TypeError: int() argument must be a string or a number, not 'list'

It was a pretty long chunk of text, so I had to split it up to do manually with curl. Here is the request that causes the error:

$ curl -H "accept: application/json" http://localhost:2222/rest/annotate --data
 "text=%0D%0AResearch+in+this+cycle+has+happened+in+different+ways.+Part+of+it+
was+done+through+analysis%2C+%0D%0Aothers+through+prototyping+and+observation.+
%0D%0A+%0D%0AAnalysis+of+flashmobs+%0D%0ATo+find+out+what+the+exact+interaction
+was+that+I+was+looking+for%2C+I+did+an+analysis+of+%0D%0Aflashmobs.+The+channe
l+of+%E2%80%98Improv+Everywhere%E2%80%99+I+used+as+inspiration.+%0D%0AI+looked+
at+many+of+these+examples+and+found+keywords+that+can+be+found+in+the+workbook+
%0D%0A%5B1%5D.+%0D%0A+%0D%0A+%0D%0AObservation+%0D%0ATo+see+how+people+would+re
spond+to+a+smile+I+went+in+the+train+from+Duivendrecht+to+Delft.++%0D%0ASome+of
+the+observations+can+be+found+in+the+workbook.+%5B3%5D+%0D%0A+%0D%0AFindings+%
0D%0APositivity%2C+such+as+a+smile%2C+has+a+good+influence+on+the+atmosphere+of
+the+train-%0D%0Acompartment.+A+happy+ticket-controller+creates+some+smiles+alr
eady.+Actually+just+like+in+%0D%0A&confidence=0.9&support=0"

The response (just the relevant part):

{
  "@text": "\r\nResearch in this cycle has happened in different ways. Part of i
t was done through analysis, \r\nothers through prototyping and observation. \r\
n \r\nAnalysis of flashmobs \r\nTo find out what the exact interaction was that
I was looking for, I did an analysis of \r\nflashmobs. The channel of ÔÇÿImprov
Everywhere' I used as inspiration. \r\nI looked at many of these examples and fo
und keywords that can be found in the workbook \r\n[1]. \r\n \r\n \r\nObservatio
n \r\nTo see how people would respond to a smile I went in the train from Duiven
drecht to Delft.  \r\nSome of the observations can be found in the workbook. [3]
 \r\n \r\nFindings \r\nPositivity, such as a smile, has a good influence on the
atmosphere of the train-\r\ncompartment. A happy ticket-controller creates some
smiles already. Actually just like in \r\n",
  "@confidence": "0.9",
  "@support": "0",
  "@types": "",
  "@sparql": "",
  "@policy": "whitelist",
  "Resources":   [
        ....
    {
      "@URI": "http://dbpedia.org/resource/Contub",
      "@support": "4",
      "@types": "",
      "@surfaceForm": [1],
      "@offset": "421",
      "@similarityScore": "0.8010543367647199",
      "@percentageOfSecondRank": "0.0"
    },

The issue is that "[1]" is present in the text, but is not encoded as a string by Spotlight. Could this be fixed by turning it into a string anyway? I'm guessing surfaceForm should always have a string value..

[originell]

I would feel better if you could first make a bug report in dbpedia-spotlight/dbpedia-spotlight :) Let's see what they have to say about that. I guess any JSON parser will convert "[1]" into the language's Array type.

I will try and introduce a workaround for this.. though that's really patch-y haha

[aolieman]

Okay, thanks for the workaround! Of course you are completely right about this being a dbpedia-spotlight issue. I reported it as issue #197 there.

[originell]

Great :) Let's see what they have to say about that :)