1. single word verbs(lexical units) data files:

swv_T.pkl: contains 79584 records, there are 8 columns in it.
column 'masked_sent' contains the sentence where targetword is highlighted using underscore like:  __targetword__

swv_gold_dataset.pkl: file with similar format as swv_T, but contains gold clusters.
Column 'gold_clusters_processed' should be used for evaluation.

Following two files conatin indices to split swv datafile into dev and test set (as it was used in original ELMo experiments)
swv_gold_dataset_dev_split.json
swv_gold_dataset_test_split.json

2. single word frame roles data files:
swr_T.pkl: file for experiment, contains 191252 records, there are 8 columns in it.
column 'masked_sent'  contains the sentence where target role is highlighted as: __targetrole__

swr_gold_dataset.pkl: contains gold clusters, column 'gold_cluster_patternlemmatized' should be used for evaluation                               

Following two files conatin indices to split swr datafile into dev and test set (as it was used in original ELMo experiments)
swr_gold_dataset_dev_split.json
swr_gold_dataset_test_split.json

3. vocabulary file for verbs:
verbs_list.txt: This file contains list of verbs used to filter predictions of lexical_unit during postprocess step.should be used for vocabulary_path variable in run_postprocessing_predictions.py