For task 1, the files in /DeepBank1.1 have been modified and the associated FrameNet information has been added to the DeepBank EDS node.
For example, in /DeepBank1.1/20001001, the node e3 becomes
e3:_join_v_1**-fn.Cause_to_amalgamate**<34:38>{e SF prop, TENSE fut, MOOD indicative, PROG -, PERF -}[ARG1**-fn.Agent** x6, ARG2**-fn.Part_1** x23]
The FrameNet frame Cause_to_amalgamate was added to the predicate and the FrameNet roles Agent and Part_1 were added to its arguments.
The predicates that were properly labeled were extracted and saved in train.csv, this becomes the training data in task 2. The predicates that were labeled as IN/NF were also extracted and saved in predict.csv, this is the data to be predicted in tsak 2.
For task 2, a Convolutional Neural Network was trained on the data in train.csv, and the model is then used to predict the FrameNet frames for the nodes in predict.csv. The outcome of the prediction is saved in pred_output.csv.