WRD: A Web-Based, Crowdsourced Approach to Sentiment Lexicon Annotation

Pierce, Haylee

WRD: A Web-Based, Crowdsourced Approach to Sentiment Lexicon Annotation

Author(s)

Pierce, Haylee

Date Issued

May 1, 2025

Abstract

This paper introduces the Word Rating Database (WRD), a web-based platform designed to crowdsource sentiment ratings for individual words, in an effort to address the limitations of current linguistic datasets. As artificial intelligence (AI) models become more advanced, the datasets that are being used to train them remain largely static, outdated, and unrepresentative of the diversity that can be found in real-world language use. WRD proposes a more dynamic and inclusive alternative by allowing the public to participate in the process of labeling the data; subsequently, building a lexicon that can better reflect the collective and evolving nature of language. Existing sentiment lexicons are often labeled by a small group of annotators, which can lead to dataset that is both unrepresentative and biased. In contrast, WRD invites a broader, more diverse user base to contribute to the ratings, with the goal of capturing sentiments that are able to be applied to a larger population. This study explores the shortcomings of linguistic datasets, critiques current labeling practices, and argues for a shift toward more community- and human-driven methods of data collection and labeling. It also considers the ethical and practical implications of maintaining a dataset in a world with generative AI and raises concerns about data authenticity, ownership, and accessibility. WRD acts as an example for how linguistic datasets might be able to remain relevant, transparent, and representative in the future, and offers insight into how a more collaborative approach to data collection is capable of improving both the inclusivity and usefulness of linguistic data.

Major

Computer Science

First Reader(s)

Luman, Douglas J.