Performance analysis of large language models in the domain of legal argument mining

Accompanying code for the paper "Performance analysis of large language models in the domain of legal argument mining" by the authors:

  Abdullah Al Zubaer[1], Granitzer Michael[1], Mitrović Jelena[1,2]

  Affiliation:
  [1] Faculty of Computer Science and Mathematics, Chair of Data Science, University of Passau, Passau, Germany
  [2] Group for Human Computer Interaction, Institute for Artificial Intelligence Research and Development of Serbia, Novi Sad, Serbia

Prerequisites

Before you begin, ensure you have met the following requirements:

• You have installed Anaconda or Miniconda.
• You have a machine with Windows, Linux, or macOS.

Installation

Tested on Ubuntu 22.04.4 LTS using Python 3.8.19

Follow these steps to get your development environment running:

1. Clone the repository:

   git clone https://github.com/yourusername/yourprojectname.git
   cd yourprojectname

2. Create a Conda environment: Replace frontier with the name of the environment you want to use:

   conda create --name frontier python=3.8  

3. Activate the Conda environment:

   conda activate frontier

4. Install required packages: This command installs all the dependencies listed in requirements.txt:

   pip install -r requirements.txt

Usage

Step 1

Dataset Extraction

1. Please download the dataset and place it in the ./data/original folder. We only need ECHR_Corpus.json. The dataset can be downloaded from http://www.di.uevora.pt/~pq/echr/.

2. Extract the dataset using the following command. This will extract the json file from the dataset and save it in the ./data/experiment_data/ folder as pickle files.

It will also save supplementary data in the ./data/supplementary_data/ folder. This data can be used for further analysis and understanding of the dataset. Currently this one is not used in our work. But it has good potential for further analysis.

python extraction-echr.py --json_file_path ./data/original/ECHR_Corpus.json

Step 2

Argument Mining

Please follow the configurations outlined in Table 6 of the paper to provide the appropriate keywords as arguments to run the code using GPT-3.5 and GPT-4 as our LLMs. The embeddings, model outputs, and results will be automatically saved in the current directory with appropriate names and formats.

Embeddings will be saved in ./embeddings directory, model outputs will be saved in ./model_classification directory, and results will be saved in ./results directory.

To get help on the command line arguments, run:

python classification.py -h

sample command to run the code is as follows:

python classification.py -e text-embedding-ada-002 -m gpt-3.5-turbo -pt twoshot_with_instruction -mode similar -c premise

python classification.py -e text-embedding-ada-002 -m gpt-3.5-turbo -pt twoshot_with_instruction -mode similar -c conclusion

Note: Certain configurations is not allowed by nature of our experiment. For example, using --embedding_model multi-qa-mpnet-base-dot-v1 but with--mode random.

Concrete example: python classification.py -e multi-qa-mpnet-base-dot-v1 -m gpt-3.5-turbo -pt twoshot_with_instruction -mode random -c premise This is because we do not perform any embedding operation when we are choosing few shot examples randomly. This will throw an error. Therfore, please follow the configurations outlined in Table 6 of the paper to provide the appropriate keywords as arguments to run the code.

License

This work is licensed under a Creative Commons Attribution 4.0 International License.

Citation

If you use our code and paper, please cite our paper as follows:

@ARTICLE{10.3389/frai.2023.1278796,
  
AUTHOR={Al Zubaer, Abdullah and Granitzer, Michael and Mitrović, Jelena},   
	 
TITLE={Performance analysis of large language models in the domain of legal argument mining},      
	
JOURNAL={Frontiers in Artificial Intelligence},      
	
VOLUME={6},           
	
YEAR={2023},      
	  
URL={https://www.frontiersin.org/articles/10.3389/frai.2023.1278796},       
	
DOI={10.3389/frai.2023.1278796},      
	
ISSN={2624-8212},   
   
ABSTRACT={Generative pre-trained transformers (GPT) have recently demonstrated excellent performance in various natural language tasks. The development of ChatGPT and the recently released GPT-4 model has shown competence in solving complex and higher-order reasoning tasks without further training or fine-tuning. However, the applicability and strength of these models in classifying legal texts in the context of argument mining are yet to be realized and have not been tested thoroughly. In this study, we investigate the effectiveness of GPT-like models, specifically GPT-3.5 and GPT-4, for argument mining via prompting. We closely study the model's performance considering diverse prompt formulation and example selection in the prompt via semantic search using state-of-the-art embedding models from OpenAI and sentence transformers. We primarily concentrate on the argument component classification task on the legal corpus from the European Court of Human Rights. To address these models' inherent non-deterministic nature and make our result statistically sound, we conducted 5-fold cross-validation on the test set. Our experiments demonstrate, quite surprisingly, that relatively small domain-specific models outperform GPT 3.5 and GPT-4 in the F1-score for premise and conclusion classes, with 1.9% and 12% improvements, respectively. We hypothesize that the performance drop indirectly reflects the complexity of the structure in the dataset, which we verify through prompt and data analysis. Nevertheless, our results demonstrate a noteworthy variation in the performance of GPT models based on prompt formulation. We observe comparable performance between the two embedding models, with a slight improvement in the local model's ability for prompt selection. This suggests that local models are as semantically rich as the embeddings from the OpenAI model. Our results indicate that the structure of prompts significantly impacts the performance of GPT models and should be considered when designing them.}
}

And please cite the following paper if you use their dataset in your research:

@inproceedings{poudyal-etal-2020-echr,
    title = "{ECHR}: Legal Corpus for Argument Mining",
    author = "Poudyal, Prakash  and
      Savelka, Jaromir  and
      Ieven, Aagje  and
      Moens, Marie Francine  and
      Goncalves, Teresa  and
      Quaresma, Paulo",
    editor = "Cabrio, Elena  and
      Villata, Serena",
    booktitle = "Proceedings of the 7th Workshop on Argument Mining",
    month = dec,
    year = "2020",
    address = "Online",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2020.argmining-1.8",
    pages = "67--75",
    abstract = "In this paper, we publicly release an annotated corpus of 42 decisions of the European Court of Human Rights (ECHR). The corpus is annotated in terms of three types of clauses useful in argument mining: premise, conclusion, and non-argument parts of the text. Furthermore, relationships among the premises and conclusions are mapped. We present baselines for three tasks that lead from unstructured texts to structured arguments. The tasks are argument clause recognition, clause relation prediction, and premise/conclusion recognition. Despite a straightforward application of the bidirectional encoders from Transformers (BERT), we obtained very promising results F1 0.765 on argument recognition, 0.511 on relation prediction, and 0.859/0.628 on premise/conclusion recognition). The results suggest the usefulness of pre-trained language models based on deep neural network architectures in argument mining. Because of the simplicity of the baselines, there is ample space for improvement in future work based on the released corpus.",
}