sandeepsoni/contextual-leadership

title
HP style models for modeling semantic innovation cascades in NLP venues.

Modeling pipeline:

• make_venue_cascades.py
  • python make_venue_cascades.py --innovs-file data/innovs.pkl --cascades-file data/cascades.pkl --offset 0.2 --start-year 1989 --num-cascades 200 --num-events 100 --data-file data/hp-data-v1-001.pkl --summary-file data/hp-data-summary.json
• collapse_and_filter_venues.py (optional: needed when venues need to be filtered or collapsed)
  • python collapse_and_filter_venues.py --input-data-file data/hp-data-v1-001.pkl --venues-file data/venues.tsv --keep-venues-file data/keep_venues.txt --output-data-file data/hp-data-v2-001.pkl --log-file data/collapse_and_filter_venues.log
• sem_influence.py
  • python sem_influence.py --data-file data/hp-data-v2-001.pkl --num-cd-iterations 50 --vanilla-hp-params-file data/v2_001_vanilla_hp_params.pkl --init-params-file data/v2_001_init_params.pkl --coarsened-hp-params-file data/v2_001_coarsened_hp_params.pkl --log-file data/sem_influence.log
• citation_analysis.py
  • python citation_analysis.py --data-file data/hp-data-v2-001.pkl --params-file data/v2_001_coarsened_hp_params.pkl --citations-file data/venue_citations.pkl --venues-file data/venues.tsv --keep-venues-file data/keep_venues.txt --output-file data/v2_001_influence_relations.csv --log-file data/citation_analysis.log

contextual-leadership's Issues

Is it possible for you to release a demo (e.g. a colab page) to illustrate the method?

Hi @sandeepsoni

I just read the EMNLP paper and felt the methodology is cool. However, as I found this repo, I have not yet found demo colab that provides some illustrative guide or playground for your new method. I checked the notebooks but they look like some log of your early experimentation.

Is `sem_influence.py` described in readme the same with `coarse_sem_influence.py`

Hi @sandeepsoni

Your paper is pretty cool and I tried to go through the codebase but a bit confused by how to replicate the results of your cool paper. This is the issue about pipeline description.

Does the sem_influence.py in the pipeline described by readme is the one called coarse_sem_influence.py in your model folder?

Pipeline description:

coarse_sem_influence.py`:
https://github.com/sandeepsoni/contextual-leadership/blob/main/scripts/modeling/coarse_sem_influence.py