UNFAIR-ToS Dataset

A UNFAIR-ToS Dataset is a legal text annotated dataset that can support unfair legal contract clause identification tasks through terms of service clauses annotated for consumer fairness.

• AKA: Unfair Terms of Service Dataset, UNFAIR-ToS Corpus, Terms of Service Fairness Dataset.
• Context:
  • It can typically provide Annotated ToS Clauses labeled with unfairness categories for unfair clause classifier training.
  • It can typically include Terms of Service Documents from online platforms with clause-level annotations.
  • It can typically contain Eight Unfairness Types including liability limitation, unilateral change, and jurisdiction restrictions.
  • It can typically offer Binary Fairness Labels distinguishing fair clauses from unfair clauses.
  • It can typically support Multi-Label Classification with multiple unfairness types per ToS clause.
  • ...
  • It can often serve as part of LexGLUE Benchmark for legal NLP evaluation.
  • It can often enable Consumer Protection Research through systematic unfairness analysis.
  • It can often facilitate Cross-Platform Comparisons of terms of service practices.
  • It can often provide Legal Expert Annotations based on consumer protection guidelines.
  • ...
  • It can range from being a Small UNFAIR-ToS Dataset to being a Large UNFAIR-ToS Dataset, depending on its ToS document count.
  • It can range from being a Single-Language UNFAIR-ToS Dataset to being a Multilingual UNFAIR-ToS Dataset, depending on its ToS language coverage.
  • It can range from being a Coarse-Grained UNFAIR-ToS Dataset to being a Fine-Grained UNFAIR-ToS Dataset, depending on its unfairness annotation granularity.
  • It can range from being a Static UNFAIR-ToS Dataset to being a Dynamic UNFAIR-ToS Dataset, depending on its ToS update frequency.
  • ...
  • It can enable Unfair Clause Detection Models through supervised learning approaches.
  • It can support Terms of Service Analysis via unfairness pattern mining.
  • It can facilitate Legal Technology Development for automated ToS review.
  • It can contribute to Consumer Protection Tools through unfairness detection training.
  • It can advance Legal NLP Research in fairness assessment domains.
  • ...
• Examples:
  • UNFAIR-ToS Unfairness Categories, such as:
    • Liability Limitation Category for excessive liability waivers.
    • Unilateral Change Category for arbitrary modification rights.
    • Content Removal Category for discretionary deletion clauses.
    • Jurisdiction Category for inconvenient forum selections.
    • Choice of Law Category for unfavorable law selections.
    • Arbitration Category for mandatory arbitration clauses.
    • Unilateral Termination Category for one-sided termination rights.
    • Contract by Using Category for browse-wrap agreements.
  • UNFAIR-ToS Application Domains, such as:
    • Social Media Platform ToS for user agreement analysis.
    • E-Commerce Platform ToS for seller agreement evaluation.
    • Cloud Service ToS for service agreement assessment.
  • ...
• Counter-Examples:
  • LEDGAR Dataset, which contains SEC filing provisions rather than consumer-facing ToS clauses.
  • ContractNLI Dataset, which focuses on contract inference rather than fairness classification.
  • CUAD Dataset, which targets commercial contract clauses rather than ToS unfairness.
• See: LexGLUE Benchmark, Unfair Legal Contract Clause Identification Task, Legal Text Dataset, Consumer Protection Law, Terms of Service Agreement, Annotated Dataset, Multi-Label Classification Dataset.

---

---

References

2024

• (Sai et al., 2024) => Bathini Sai Akash, Akshara Kupireddy, Lalita Bhanu Murthy. (2024). "Unfair TOS: An Automated Approach using Customized BERT." https://doi.org/10.48550/arXiv.2401.11207
  • ABSTRACT: Terms of Service (ToS)] form an integral part of any agreement as it defines the legal relationship between a service provider and an end-user. Not only do they establish and delineate reciprocal rights and responsibilities, but they also provide users with information on essential aspects of contracts that pertain to the use of digital spaces. These aspects include a wide range of topics, including limitation of liability, data protection, etc. Users tend to accept the ToS without going through it before using any application or service. Such ignorance puts them in a potentially weaker situation in case any action is required. Existing methodologies for the detection or classification of unfair clauses are however obsolete and show modest performance. In this research paper, we present SOTA(State of The Art) results on unfair clause detection from ToS documents based on unprecedented custom BERT Fine-tuning in conjunction with SVC(Support Vector Classifier). The study shows proficient performance with a macro F1-score of 0.922 at unfair clause detection, and superior performance is also shown in the classification of unfair clauses by each tag. Further, a comparative analysis is performed by answering research questions on the Transformer models utilized. In order to further research and experimentation the code and results are made available on this https URL.

---