Legalpha

Chatbot for Tenancy Law Questions 🤖 ⚖️

Problem

For equal rights, everyone should have equal access to legal information.

Solution

A chatbot giving easy to digest answers to law questions

• Free for everyone to use
• With paragraph references to prevent disinformation

Scope

The scope of this project is limited to tenancy law in Germany for now.

Current topics covered:

1. Contract termination
2. Deposit

How to contribute?

Contribute to the question dataset

Submit sample questions about tenancy law to train Legalpha using this form.

Contribute to tasks

You can also contribute to the tasks regarding the following topics:

• Data collection
• Legal research
• ML / NLP
• Frontend Development
• Backend Development

To contribute, please contact me here.

How to run the code?

Prerequisites

• Docker
• Docker Compose

Run the server

1. Clone the repository
2. Run docker compose up in the root directory
3. Open localhost:8000 in your browser

Run ML jobs

docker compose up with respective commands.

1. Tune hyperparameters: python main.py --tune --model-name <MODEL_NAME> --tune-iters <ITERS> --tune-cv-folds <FOLDS>
2. Test the model with default parameters: python main.py --test --model-name <MODEL_NAME> --test-size <TEST_FRACTION> --test-sampling <SAMPLING_STRATEGY>
3. Train and save the model with default parameters: python main.py --train

Test Results

Random Sampling

Stratified Sampling