Summary of Voices in a Crowd: Searching For Clusters Of Unique Perspectives, by Nikolas Vitsakis et al.
Voices in a Crowd: Searching for Clusters of Unique Perspectives
by Nikolas Vitsakis, Amit Parekh, Ioannis Konstas
First submitted to arxiv on: 19 Jul 2024
Categories
- Main: Computation and Language (cs.CL)
- Secondary: 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 proposes a novel approach to training language models that captures minority perspectives without encoding annotator metadata. The proposed framework extracts latent embeddings informed by annotator behaviour and creates clusters of similar opinions, referred to as “voices”. This framework is validated post-hoc via internal and external quantitative metrics, as well as qualitative analysis to identify the type of voice each cluster represents. The results demonstrate strong generalisation capability and robustness across two distinct datasets. |
| Low | GrooveSquid.com (original content) | Low Difficulty Summary The paper aims to address the issue of language models reproducing biases in their training data by default. Previous solutions have tried to capture minority perspectives by modelling annotator disagreements or grouping annotators based on shared metadata, but these approaches face significant challenges. The proposed framework trains models without encoding annotator metadata and extracts latent embeddings that capture similar opinions, creating “voices” that are validated post-hoc. |
