Welcome to UCLA Undergraduate Research Week 2025!

Thank you for visiting the 2025 Undergraduate Research and Creativity Showcase. This Showcase features student research and creative projects across all disciplines. As a university campus, free expression is encouraged, and some content may not be appropriate for all ages. Visitors under the age of 18 are encouraged to explore these presentations with a parent or guardian. The views and opinions expressed here are those of the participants and do not necessarily reflect UCLA or any policy or position of UCLA. As a visitor, you agree not to record, copy, or reproduce any of the material featured here. By clicking on the "Agree" button below, you understand and agree to these terms.

Humanities: Session A: 12:30-2pm - Panel 1

Tuesday, May 20 12:30PM – 1:50PM

Location: Online - Live

The Zoom link will be available here 1 hour before the event.

Presenter 1
EMILY FOSTER, Tamar Kremer-Sadlik
Sentiment Analysis in Naturalistic Multiparty Speech
Introduction Sentiment analysis using NLP offers a scalable alternative to manual affect coding in social science. While BERT models perform well on structured text, their effectiveness on naturalistic speech—where sentiment is often ambiguous—remains underexplored. We test RoBERTa-base on a dataset of multiparty, real-life interactions. Background Previous work links speech sentiment to psychological well-being. We examine correlations between model-generated sentiment and participants’ depression and marital satisfaction scores to assess the model’s ability to capture emotional tone in speech. Methods Data includes depression (CES-D) and marital satisfaction (MAT) assessment scores, as well as 62 hours of video data from interviews conducted by UCLA Sloan CELF project on 32 U.S. dual-earner families. Transcripts from these interviews were split into 54,901 sentences to be analyzed by RoBERTa-base which was fine-tuned on Switchboard Sentiment for 3 epochs. Results Model outputs significantly correlated with psychological scores: more depressive symptoms correlated to less positive (r=-0.27) and more negative speech (r=0.27); greater marital satisfaction correlated to less negative speech (r=-0.33). Findings support BERT’s ability to analyze naturalistic speech sentiment, with further validation underway through manual coding.
Presenter 2
WYNTER HALE, Laila Wheeler, Kyah Gaines, Sierra Stewart, Meelah Rodriguez, Brailyn Page, Sakari Dunlap, and Domonique Henderson.
Online Gendered Racism: The Lived Experiences of Black Girls Online
Online Gendered Racism explores the intersection of racism and sexism experienced by Black girls through their interactions on social media. Our study aims to answer the following: In what ways do Black girls experience online gendered racism on social media and online platforms such as Instagram, X (formerly known as Twitter), Facebook, Snapchat, Tiktok, or gaming chat rooms? Along with, in what ways are Black girls’ mental health (anxiety and depression), academic performance, gendered racial identity, and resistance impacted by gendered racial comments and discourse online? This qualitative study employed Constructivist Grounded Theory through participatory visual methods such as semi-structured interviews, photovoice, and a quantitative survey. Our participants were 41 Black girls and non-binary individuals ranging from the ages 14-18 across the US. All data was analyzed through reflective memos, line-by-line coding, focused coding, and monthly team workshops. Our participants shared that they have witnessed and encountered instances of racism, colorism, and hyper-sexualization on social media and gaming platforms. Our participants also revealed that their mental health, self-confidence, and in-person peer relationships were impacted following these encounters. By highlighting Black girls’ experiences, this uncovers the crucial need for policymakers, educators, and social media platforms to actively combat online gendered racism.
Presenter 3
KEYA TANNA
Ways of Seeing Women in Indian Art: A Decolonial Feminist Inquiry into Visual Tradition and Cultural Memory
This project investigates the gendered visual language of Indian folk and miniature art through the lens of John Berger’s Ways of Seeing, centering on how women are depicted as symbolic, stylized, and often passive figures. Rather than solely critiquing these portrayals, I approach them through a decolonial feminist lens—seeking to understand how visual storytelling in traditional Indian art encodes gender roles while also serving as a vehicle of cultural identity. My interest in this intersection was sparked while organizing a student-led exhibition for the India Pavilion at Expo 2020. I curated a collection of regional Indian folk paintings replicated by my school art class, including my own painted reproduction of “Bani Thani.” I arranged the works by style and narrative theme, writing accompanying labels that explained their symbolism and origin stories. This process led me to question how visual traditions preserve ideals of femininity across generations. Drawing on qualitative visual analysis and postcolonial feminist theory, I compare Indian images like “Bani Thani” with Berger’s critique of Western works such as Ingres’ La Grande Odalisque. Through elements like gaze, gesture, and symbolic ornamentation, I explore how both traditions construct the female subject for cultural consumption. This study reconsiders how we “see” tradition, and how inherited art forms both empower and constrain the modern viewer.
Presenter 4
BOWER, JACOB
To [ř] or not to [ř]: A cross-dialectal sociophonetic study of assibilated rhotics in Latin American Spanish

With the goal of understanding observed cross-dialectal attitude differences, the present study probed linguistic attitudes towards the assibilated rhotic variant ([ř]) in word-final position among speakers from three regions of the Spanish-speaking world: Mexico, the Andes (Peru and Ecuador), and Spain. Utilizing the matched-guise technique, we plan to devise a survey utilizing the SurveyMonkey software, containing stimuli developed spliced audio samples from the PRESEEA Corpus (Moreno-Fernandez 2005) and publicly-available YouTube videos. Participants will be instructed to listen to the stimuli, which will include words with the [ř] variant, words with the traditional [r] variant, and words without rhotics. Upon listening to the stimuli, participants indicate via Likert scales where they believe the speaker lies on various adjective dichotomies (e.g., nice vs. mean, educated vs. uneducated). We expect results will indicate that participants from Mexico rate the speaker using [ř] more positively while participants from the Andean region will rate the speaker using [ř] more negatively, based on prior findings that [ř] is a variant covert prestige in Mexico, while being disfavored in the Andean region. We expect that speakers from Spain, where the [ř] variant is absent, will have more neutral opinions of the variant. Secondarily, a robust acoustic analysis of this variant across dialects will be conducted, investigating potential differences between acoustic properties of [ř] across different dialects and speakers.

BACK