Summary of Calpric: Inclusive and Fine-grain Labeling Of Privacy Policies with Crowdsourcing and Active Learning, by Wenjun Qiu et al.
Calpric: Inclusive and Fine-grain Labeling of Privacy Policies with Crowdsourcing and Active Learning
by Wenjun Qiu, David Lie, Lisa Austin
First submitted to arxiv on: 16 Jan 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: Cryptography and Security (cs.CR); Human-Computer Interaction (cs.HC); Machine Learning (cs.LG)
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 This paper presents Calpric, a novel approach to generating a large, comprehensive training set for deep learning models on privacy policies at a low cost. By combining automatic text selection and segmentation, active learning, and crowdsourced annotators, Calpric simplifies the labeling task and reduces inter-annotator agreement. This enables untrained annotators from crowdsourcing platforms to be competitive with trained annotators, such as law students. The paper also explores the use of active learning, which uses fewer training samples to efficiently cover the input space, further reducing cost and improving class and data category balance in the data set. Calpric’s training process generates a labeled dataset of 16K privacy policy text segments across 9 categories with balanced positive and negative samples. The approach allows for the production of models that are accurate over a wider range of data categories, providing more detailed, fine-grain labels than previous work. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary This paper helps us create better deep learning models for understanding privacy policies by making it cheaper to get lots of training data. They do this by using special computer programs to pick out the most important parts of the text, then getting people who aren’t experts to help label those parts. This makes it easier and cheaper to get accurate labels. The researchers also use a clever way to make sure they’re not wasting time or money on unnecessary training data. As a result, their models can understand more types of privacy policies and provide better details than previous attempts. |
Keywords
* Artificial intelligence * Active learning * Deep learning
