《Neural Symbolic Machines》Discussion: What are the input and target for the neural net ?

https://github.com/carpedm20/paper-notes/blob/master/notes/neural-symbolic-machine.md


From the picture of paper, we find that argmax operation is in the target.
But the WEBQUESTIONSSP dataset is like this:
```
{
"QuestionId": "WebQTrn-2632",
"RawQuestion": "what city has the largest population in mexico?",
"ProcessedQuestion": "what city has the largest population in mexico",
"Parses": [
{
"ParseId": "WebQTrn-2632.P0",
"AnnotatorId": 2,
"AnnotatorComment": {
"ParseQuality": "Complete",
"QuestionQuality": "Good",
"Confidence": "Normal",
"FreeFormComment": "can't do max after selecting \"city\""
},
"Sparql": "PREFIX ns: \nSELECT DISTINCT ?x\nWHERE {\nFILTER (?x != ns:m.0b90_r)\nFILTER (!isLiteral(?x) OR lang(?x) = '' OR langMatches(lang(?x), 'en'))\nns:m.0b90_r ns:location.location.contains ?x .\n?x ns:common.topic.notable_types ns:m.01m9 .\n?x ns:topic_server.population_number ?sk0 .\n}\nORDER BY DESC(xsd:float(?sk0))\nLIMIT 1\n",
"PotentialTopicEntityMention": "mexico",
"TopicEntityName": "Mexico",
"TopicEntityMid": "m.0b90_r",
"InferentialChain": [
"location.location.contains"
],
"Constraints": [
{
"Operator": "Equal",
"ArgumentType": "Entity",
"Argument": "m.01m9",
"EntityName": "City/Town/Village",
"SourceNodeIndex": 0,
"NodePredicate": "common.topic.notable_types",
"ValueType": "String"
}
],
"Time": null,
"Order": {
"SortOrder": "Descending",
"Start": 0,
"Count": 1,
"SourceNodeIndex": 0,
"NodePredicate": "topic_server.population_number",
"ValueType": "Number"
},
"Answers": [
{
"AnswerType": "Entity",
"AnswerArgument": "m.04sqj",
"EntityName": "Mexico City"
}
]
}
]
},
```
@carpedm20 Thank you.