Hi @ratishsp,

Thanks a lot for the great paper & code. The experiments for RotoWire are fine, but with MLB I think some file is missing here. Pls see the error below:

(macro_plan directory is the main branch's code)

Traceback (most recent call last):
  File "./macro_plan/scripts/create_mlb_target_data.py", line 160, in <module>
    process(args.json_root, args.dataset_type, args.ordinal_adjective_map_file, args.output_folder)
  File "./macro_plan/scripts/create_mlb_target_data.py", line 20, in process
    ordinal_adjective_map = get_ordinal_adjective_map(ordinal_adjective_map_file)
  File "/check/data2text-macro/macro_plan/scripts/mlb_utils.py", line 556, in get_ordinal_adjective_map
    ordinal_adjective_map_file = open(ordinal_adjective_map_file_name, mode="r", encoding="utf-8")
FileNotFoundError: [Errno 2] No such file or directory: './macro_plan/data/traintokens-ordinaladjective-inning-identifier'

Is there something obvious I'm missing here?

Thanks!

Hi @ratishsp,

Thanks for providing the code for your paper. I am trying to learn more about this method and run the code, but I do not see many instructions like the other repos such as https://github.com/ratishsp/data2text-plan-py.

I just wonder how can I run the inference pipeline?

I can see the structure is quite similar with https://github.com/ratishsp/data2text-plan-py.

So if I want to run the code, shall I follow the same instructions?

Many thanks

Hi,
How exactly did you apply best-worst scaling? And how did you obtain Table 5 in the paper from binary comparisons (Macro against RBF- 2020,4 ED+CC, Gold, and Templ)?

Why do you think best-worst scaling is more reliable? I've never seen this method in NLP papers before.