Summary of The Right Model For the Job: An Evaluation Of Legal Multi-label Classification Baselines, by Martina Forster et al.
The Right Model for the Job: An Evaluation of Legal Multi-Label Classification Baselines
by Martina Forster, Claudia Schulz, Prudhvi Nokku, Melicaalsadat Mirsafian, Jaykumar Kasundra, Stavroula Skylaki
First submitted to arxiv on: 22 Jan 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Artificial Intelligence (cs.AI)
GrooveSquid.com Paper Summaries
GrooveSquid.com’s goal is to make artificial intelligence research accessible by summarizing AI papers in simpler terms. Each summary below covers the same AI paper, written at different levels of difficulty. The medium difficulty and low difficulty versions are original summaries written by GrooveSquid.com, while the high difficulty version is the paper’s original abstract. Feel free to learn from the version that suits you best!
| Summary difficulty | Written by | Summary |
|---|---|---|
| High | Paper authors | High Difficulty Summary Read the original abstract here |
| Medium | GrooveSquid.com (original content) | Medium Difficulty Summary A novel evaluation is conducted to assess various Multi-Label Classification (MLC) methods on two public legal datasets, POSTURE50K and EURLEX57K. The study explores how different approaches perform in relation to dataset properties by varying the training data amount and number of labels. Notably, DistilRoBERTa and LegalBERT consistently demonstrate strong performance in legal MLC with reasonable computational demands, while T5 shows comparable results as a generative model when label sets change. Additionally, CrossEncoder exhibits potential for significant macro-F1 score improvements at the cost of increased computational costs. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary Legal documents often have multiple labels assigned to them. Researchers used two public datasets to test different methods for this task called Multi-Label Classification (MLC). They wanted to see how well each method worked depending on how much data was used and how many labels were given. The results showed that certain models like DistilRoBERTa and LegalBERT do well in a reasonable amount of time, while others like T5 can also work well when the labels change. Finally, another model called CrossEncoder could improve its performance but at a higher cost. |
Keywords
* Artificial intelligence * Classification * F1 score * Generative model * T5
