The datasets are formatted in JSON and organized as one aspect term corresponding to multiple opinion terms:
[{
"id": "id of sentence 1",
"sentence": "context of sentence 1",
"triples": [{
"uid": "id of first aspect term of sentence 1",
"target_tags": "the first aspect term with BIO scheme",
"opinion_tags": "the corresponding multiple opinion terms of the aspect term with BIO scheme",
"sentiment": "the corresponding sentiment polarity of the aspect term"
},{
the second aspect term of sentence 1
...
}]
}, {
sentence 2
...
}]
So you can fetch one aspect term, its corresponding multiple opinion terms, and the corresponding sentiment polarity to form multiple opinion triplets of this aspect term.