Improve tokenizer.py to have er_link_style parameter

bart/pred.json

bart/qald-10-test-predictions.csv

bart/qald-9-test-linked-tokenized.csv

bart/qald-9-test-predictions.csv

bart/qald-9-train-linked-tokenized.csv

bart/sparql.ipynb

data/qald-9-test-linked-tokenized.csv

data/qald-9-test-linked.json

data/qald-9-train-linked-tokenized.csv

data/qald-9-train-linked.json

data/tokenizer.py