I wrote a Python script to scrape and convert to JSON the most common words in German and French from the legal codex of the Canton Bern (Belex). Adding this content for a lawmakers remix of the FastType game by Solomon Shalom Lijo. Try it above! Click on the yellow tab to switch modes.

from collections import Counter
import string
import re

# Load text file
file_path = "belex-fr.txt"
with open(file_path, "r") as file:
    text = file.read()

# Remove punctuation
text = re.sub(f"[{string.punctuation}]", "", text)

# Tokenize the text
tokens = text.split()

# Remove short words
tokens = [word for word in tokens if len(word)>6]

tokens1w = []
tokens1l = []
for word in tokens:
    if not word.lower() in tokens1l:
        tokens1w.append(word)
        tokens1l.append(word.lower)


# Create a dictionary with word frequencies
word_freq = Counter(tokens1w)

# Print the dictionary
print(word_freq)

Counter({'autorités': 106, 'communes': 89, 'Conseil': 81, 'Conseilexécutif': 71, 'données': 44, 'peuvent': 39, 'Constitution': 36, 'bernoises': 36, 'Recueil': 35, 'personne': 34, 'personnes': 33, 'prestations': 31, 'membres': 31, 'officiel': 31, 'cantonales': 30, 'présente': 27, 'législation': 26, 'dautres': 26, 'publication': 26, 'vigueur': 24, 'cantonale': 22, 'ressources': 22, 'électoral': 21, 'publiques': 21, 'liberté': 20, 'compétences': 20, 'protection': 20, 'populaire': 19, 'législatifs': 19, 'mesures': 18, 'surveillance': 18, 'dispositions': 18, 'Confédération': 17, 'cantonal': 17, 'manière': 17, 'législatif': 17, 'particulier': 16, 'notamment': 16, 'traitement': 16, 'ladministration': 16, 'larticle': 16, 'électronique': 16, ...

del word_freq.most_common()[0]
del word_freq.most_common()[1]
del word_freq.most_common()[2]

words = [k for k,i in word_freq.most_common() if i<50]
words

['données',
 'peuvent',
 'Constitution',
 'bernoises',
 'Recueil',
 'personne',

...

import json
output_json = "franz.json"
with open(output_json, "w", encoding="utf-8") as json_file:
    json.dump(words, json_file, indent=4)

The challenge is to create a pool of open-source contract templates for developers, starting with a Swiss-adapted version of an OSBA template. The goal is to make these templates accessible and customizable through an assistant-like interface on a website, using tools like Jinja2 or large language models. The project also involves comparing clauses from different contracts to develop criteria for explaining and evaluating contract sections, with the potential for future automation.

Challenge

An essential part of a comprehensive resource for open source developers in Switzerland is to address the challenges of having a solid set of clauses for basic legal compliance at an early stage. We propose creating a new resource to fill the gap in existing tools.

Several groups have already worked on templates for software contracts, as we have learned in feedback to our DINAcon workshop last year. This includes several schematic standards, premium contract libraries, as well as intermediary digital platforms that are commonly used in negotiation. Our next goal here is to publish a microsite with this information. Help us with your ideas on how to address the audience in an engaging and pragmatic way.

We have set up an initial contract based on a translation to Swiss terms of a template from the Open Source Business Alliance (OSBA). Create a version of this that can be processed with a templating engine (like Jinja2), or analysed by Large Language Models. Our goal here will simply be to make the template easily accessible in a wizard-like interface embedded on the website. A quick demo using Proxeus has been shared, something like Docassemble could work well too.

Comparing the clauses of our example contract with others found online, we can start building a set of criteria to better explain the individual sections and positions of the contract. This could in the future be used to do a quick automatic evaluation. Several ideas in this direction were discussed in our research and past editions of the Open Legal Lab (Legal Instruct, Textbausteine). A stretch goal of the challenge is to create a mock-up of such a process, clearly indicating at which points expert legal advice will be recommended.

🅰️ℹ️ Translated and analyzed with Qwen2.5-72B

OpenJustice is an open-source initiative by Queen’s University in collaboration with the University of Lausanne, aimed at making legal AI more transparent, fair, and accessible. The goal is to make legal information and legal assistance easily available to everyone. The challenge focuses on further developing OpenJustice specifically for Switzerland and contributing to a public good in the legal tech sector.

The key tasks for the challenge include:

1. Data Selection for OpenJustice:

   • Decide which data sources will be used for a Switzerland-specific Retrieval-Augmented Generation (RAG) solution.

2. Case Structure Development for Swiss Legal Cases:

   • Identify specific legal cases that are particularly relevant for individuals seeking legal help in Switzerland.
   • Analyze the typical legal decision-making logics (individual elements of the case) for these cases.
   • Translate these logics into structured dialog flows (modular prompt components for the OpenJustice platform).
   • Test and evaluate the developed case structures using real or fictional cases to ensure accuracy.

Evaluation

• Data Selection: This task is feasible within a 2-day hackathon if the team has a clear understanding of the available data sources and can quickly agree on the most relevant ones. Key data sources might include Swiss legal databases, court decisions, and legal statutes.
• Case Structure Development: Developing and testing case structures is a more complex task but can be managed within the time frame if the team is well-organized and has a mix of legal and technical expertise. The process involves identifying relevant cases, analyzing legal logics, and creating dialog flows, which can be time-consuming but is achievable with focused effort.

Risks and Constraints:

• Data Availability: Access to comprehensive and up-to-date legal data is crucial. The team may face challenges if the data is not readily available or is not in a format that is easy to process.
• Legal Accuracy: Ensuring that the case structures and dialog flows are legally accurate and compliant is essential. This will require close collaboration with legal experts to validate the content.
• Testing and Evaluation: Rigorously testing the developed case structures with real or fictional cases is necessary to ensure accuracy and usability. The team will need to manage this process efficiently to fit within the hackathon timeline.

Benefits:

• Accessibility: Making legal information and assistance more accessible to the general public aligns with the mission of OpenJustice and can have a significant positive impact.
• Transparency and Fairness: By developing a transparent and fair AI solution, the project can help build trust in legal technology and ensure that it serves the public interest.
• Innovation: This project can set a new standard for legal tech in Switzerland and serve as a model for similar initiatives in other regions.

Additional Considerations:

• User Feedback: Gathering feedback from potential users (e.g., legal professionals, individuals seeking legal help) can provide valuable insights and improve the final product.
• Scalability: Ensuring that the developed structures and dialog flows are scalable and can be easily adapted to new cases and legal contexts will enhance the long-term value of the project.

Conclusion

This challenge is realistic and has the potential to make a significant impact on the accessibility and transparency of legal information in Switzerland. The success of the project will depend on the team's ability to manage data selection, ensure legal accuracy, and efficiently develop and test case structures. Collaboration with legal experts and a well-organized approach will be crucial. Contributing to a public good in the legal tech sector is a valuable and meaningful endeavor.