PersonaLex: Validated short-text classifier for identities, interests, and issues

IP Status: Copyrighted

Applications

• Computational social science research
• Public-interest and non-profit organizational research
• Methodological training and education in text analysis
• Foundation for building new open-source analysis tools

Key Benefits & Differentiators

• Transparent & Interpretable: Every category, subcategory, and term is documented, allowing researchers to understand, adapt, or critique classifications, unlike opaque "black-box" or proprietary models.
• Scalable & Efficient: Automates the identification and classification of markers in large datasets of short-text, providing a robust alternative to time-consuming and inconsistent manual coding.
• Empirically Validated: Grounded in a rigorous human coding process and informed by communication and social science theories, ensuring relevance and accuracy for focused research on identities, interests, and/or issues.
• Purpose-Built: Designed specifically to detect self-expressed social identity, interest areas, and issue concerns in short text, offering greater specificity than general-purpose linguistic tools.

Technology Overview

Studying social identity expression in digital media presents significant challenges. Researchers often rely on labor-intensive manual coding, which is difficult to scale and prone to inconsistency. Conversely, machine learning and large language models (LLMs) are often proprietary "black-boxes," making their classifications difficult to interpret, replicate, audit, or justify in research. Existing off-the-shelf text analysis tools are typically built for narrow purposes like sentiment analysis and lack the nuanced insights of theoretically grounded research on digital self-expression.

Researchers at the University of Minnesota, University of South Florida, and Rutgers University have developed a computational classifier database designed to be an open, transparent, and scalable tool for research on social identities, interest areas, and issue concerns . The technology is a structured and validated lexicon of terms associated with various identities, interests, and issues. PersonaLex (v1.0) includes categories for classifying self-expressed religion, race, skintone (emojis), ethnicity, military affiliation, issue concerns, political affiliations, and sports fandom. Unlike opaque models, this classifier is fully documented and grounded in an empirical, human-coded validation process. This provides an accessible, replicable, and methodologically sound solution that enables researchers to systematically extract self-expressed information from short-text data, like social media bios, at scale.

Associated Code Repository

The source code and supporting resources associated with PersonaLex are available on the PersonaLex GitHub repository for researchers and developers interested in implementation details, customization, and further development.

Phase of Development

TRL: 6-7
The PersonaLex (v1.0) classifier database is finalized, validated, and ready for use as a research tool. Implementation scripts in R and Python are in development for public release.

Desired Partnerships

This technology is now available for:

• License
• Sponsored research
• Co-development

Please contact our office to share your business’ needs and learn more.

Licensing Information

• Academic Use: A valid academic (.edu) email address is required for license approval.
• Research/Non-Profit Use: If you are with a governmental, non-governmental, or non-profit research organization, please contact Adam Saffer ([email protected]) for the appropriate license agreement link.
• Commercial Use: This product is generally not licensed for commercial use. For case-by-case consideration of commercial inquiries, please use the "Contact Us" button.

Researchers

• Adam Saffer, PhD Associate Professor, University of Minnesota
• Joshua Scacco, PhD Director, Center for Sustainable Democracy, University of South Florida
• Jianing Li, PhD Assistant Professor, Department of Communication, School of Communication and Information, Rutgers University