Law Smells

Replication package for "Law Smells: Defining and Detecting Problematic Patterns in Legal Drafting" (Artificial Intelligence & Law 2022)

General

Below, we list the (derived) data and the code needed to reproduce our tables and figures.

For the file paths to match, the folder us (provided in the data release) is expected reside in the folder ../../legal-networks-data (when viewed from the notebooks folder), and ZIP archives are expected to be unpacked.

The code is tested in a virtual environment with a Python 3.8 interpreter and the requirements from requirements-dev.txt installed via pip install -r requirements-dev.txt.

Tables

• table-duplicates
  • data:
    • ../data/corpus_size_token_us.json (created using token_size_usc.ipynb)
    • ../data/dupex_mf-10000_results_all-years.csv
    • Dupex results (JSONs) with -mf 10000 for y in range(1998, 2020), examples for 2019 placed in ../dupex_mf-10000_results
  • code:
    • Dupex code: Zenodo Deposit
    • compute_compressions.py
    • dp_figures.ipynb (uses smell_helpers.py)
• table-long-elements
  • data: ../../legal-networks-data/us/4_crossreference_graph
  • code: longitem_analysis.ipynb (uses longitem_support.py), third code box under Example 2 heading

Figures

• dp-compression
  • data: see table-duplicates
  • code: see table-duplicates
• longitem_icicles
  • data: ../../legal-networks-data/us/4_crossreference_graph
  • code: longitem_analysis.ipynb (uses longitem_support.py), under Example 4 heading
• operator_patterns_abs_us_splitted
  • data:
    • ../../legal-networks-data/us/2_xml
    • ../../legal-networks-data/us/4_crossreference_graph/detailed
    • ../data/corpus_size_token_us.json
    • ../data/pattern_abs.csv
  • code:
    • operational-binding.ipynb
    • operational-binding-analysis.ipynb
• operator_patterns_rel_us_splitted
  • data: see previous figure
  • code: see previous figure
• reference_tree_size_ref_edges_2dhist
  • data:
    • ../../legal-networks-data/us/4_crossreference_graph/detailed
    • ../data/reference_sets_{year}.csv
  • code:
    • reference-set.ipynb
    • lrt_figures.ipynb
• diverse-reference-tree
  • data: input data only (traced manually)
  • code: none (traced manually)
• named_entities_per_thousand_tokens
  • data:
    • ../data/corpus_size_token_us.json
    • ../ner_counts data for 1998 and 2019
  • code:
    • token_size_usc.ipynb
    • nlo_figures.ipynb
    • smell_helpers.py
• committees_per_thousand_tokens
  • data: Dupex results (JSONs) with -mf 10000 for 2019, placed in ../dupex_mf-10000_results
  • code: see previous figure