data_format.dat

Each token should have the following columns, separated by whitespace or tab (non-mandatory fields may have any value)

1. Token number, starting from 1
2. The token itself
3. Lemma (not necessary, but nlpnet may be configured to use lemmas instead of surface forms)
4. Coarse POS tag (not necessary, but may be used as an additional attribute)
5. Fine POS tag / morphological information (not used, just a CoNLL convention)
6. Clause boundaries (not necessary)
7. Chunks (not necessary)
8. Parse tree (not necessary)
9. A dash (-) for non-predicates and anything else for predicates.
10. (and others) The argument labels. Each column starting from the 10th refers to an predicate, in the order they appear in the sentence. If a token is the only one in an argument, this field must contain (ARG-LABEL*). If it starts one, it must be (ARG-LABEL*. And if it ends one, it must be *). Others should have *.

Note: For our data, we essentially have only the 9th and 10th+ columns after the column nos. 1 and 2 as other information is not available for us/is not within the scope of the project.