The Mexican Gayze: A Computational Analysis of the Attitudes towards the LGBT+ Population in Mexico on Social Media Across a Decade

Scott Andersen, Segio-Luis Ojeda-Trueba, Juan Vásquez, Gemma Bel-EnguixNAACL 2024

TL;DR

This paper analyzes a decade of Twitter data to computationally track the evolution of language used to refer to the LGBT+ community in Mexico. By examining term usage, sentiment, and semantic shifts, the authors reveal how societal attitudes, as reflected in online discourse, have changed over time, highlighting both progress and persistent challenges.

What Problem Does This Paper Address?

The increasing prevalence of social media has created vast digital archives of human communication, offering unprecedented opportunities to study sociolinguistic phenomena. However, analyzing this data requires sophisticated computational tools. This paper addresses the challenge of understanding how language used to refer to the LGBT+ community in Mexico has evolved on Twitter over a ten-year period. Specifically, it investigates changes in the denotation and connotation of these terms, exploring how they are used both within and outside the LGBT+ community, and how these usages reflect broader societal attitudes.

Methods

The researchers employed a multi-faceted computational approach. They collected a large corpus of tweets from Mexico between 2012 and 2022, focusing on terms related to the LGBT+ community (known as "formas nominales de tratamiento" or FNOMTs). To analyze sentiment and track changes over time, they utilized the Pysentimiento sentiment analysis model, which was pre-trained on Spanish and English. Statistical trend analysis was performed using the Mann-Kendall test to determine if usage and polarity trends were increasing or decreasing. To understand semantic shifts, they employed Word2Vec to create vector representations of the terms and analyzed their nearest semantic neighbors across different time periods, revealing how the contextual meaning of these terms evolved.

Datasets

The primary dataset consists of 730,178 unique tweets collected from Twitter between January 2012 and October 2022, specifically from users in Mexico (tagged with "es" and "mx"). The data was collected prior to Elon Musk's acquisition of Twitter. The authors also compiled a list of FNOMTs indicative of the LGBT+ community from social media platforms, which was then used as search terms for data collection.

Key Insights and Ideas

  • Language Evolution is Accelerated Online: Social media's rapid dissemination of information accelerates linguistic change, leading to faster shifts in connotation and denotation than previously observed.
  • Shifting Terminology: There's a noticeable decrease in the usage of older, more general derogatory terms, while more specific terms for various LGBT+ subgroups are gaining traction.
  • Persistent Negative Sentiment: Despite apparent advances in acceptance, the overall sentiment surrounding LGBT+ related terms on Mexican Twitter remains predominantly negative, reflecting societal challenges.
  • "Queer" as an Inclusive Umbrella Term: The term "queer" is increasing in usage, likely due to its inclusivity and ability to encompass diverse identities without revealing specific sexual orientations or gender expressions.
  • Trans-Related Terms on the Rise: Terms related to the trans community (e.g., "trans," "transexual," "transgénero") have seen a significant increase in usage, potentially linked to increased public discourse around trans rights and visibility in media.
  • Semantic Contexts Evolve: The semantic neighbors of terms like "gay" and "homosexual" have shifted from political and social issues to more colloquial and context-specific terms used within the community.

Key Results

  • Usage Trends: Many terms show a decrease in usage over time (e.g., "homosexual," "lencha," "lesbiana," "marica," "maricón"), while others show an increase (e.g., "trans," "transexual," "transgénero," "drag," "queer").
  • Polarity: While positive usage exists, negative and neutral polarities dominate the discourse for most terms.
  • Semantic Shifts: Terms like "asexual" have shifted from discussing internal group matters to social context and rights. "Drag" has evolved from general terms to specific references to reality shows. "Gay" has moved towards more colloquial terms used within the LGBT+ community.
  • Specific Term Analysis: "Puto" maintained consistent usage but with a predominantly negative polarity. "Lesbian" usage declined, possibly due to the rise of more specific terms like "bisexual," "demisexual," and "pansexual."

Limitations and Future Work

The authors acknowledge that their list of FNOMTs might be biased towards terms used by university students and that the sentiment analysis model, while effective, may not perfectly capture all nuances of polarity. They also note the difficulty in determining the author's background, assuming a diverse range of social and economic contexts. Future work could involve collecting more data, conducting a more fine-grained analysis of specific LGBT+ subgroups, and exploring the impact of regional variations within Mexico.

Summary updated

BibTeX
@inproceedings{andersen2024mexican,
  author = {Scott Andersen and Segio-Luis Ojeda-Trueba and Juan Vásquez and Gemma Bel-Enguix},
  title = {{The Mexican Gayze: A Computational Analysis of the Attitudes towards the LGBT+ Population in Mexico on Social Media Across a Decade}},
  booktitle = {Conference of the North American Chapter of the Association for Computational Linguistics},
  year = {2024},
}
Figure