-
The data is serialized in jasonline format, with each line a sample.
-
Each sample has three fields: "meta", "words", and "oia".
-
The "meta" field contains the data source and the Uniform Resource Identifier(URI)
-
The "words" field contains the words in the sentence.
-
The "oia" field contains two list: "nodes" and "edges".
- each node in "nodes" has two fields
- the 'spans' field contains the context in the node, which could be
- pre-defined predicate/function
- the list of placeholders, additional (be) words, or start-end index pairs
- 'type' field is the fine-grained node type
- the 'spans' field contains the context in the node, which could be
- each edge in "edges" has three fields
- start node index of the edge
- end node index of the edge
- the label of the edge
- each node in "nodes" has two fields
-