Work | Meltem Odabaş

#BlackLivesMatter Turns 10: Social Media and Online Activism

Sun, 15 Feb 2026 00:00:00 +0000

This report analyzes how the #BlackLivesMatter movement has unfolded online over the past decade and how Americans engage with social and political issues on social media. It combines large-scale computational analysis of more than 44 million public tweets containing the hashtag with a nationally representative survey of U.S. adults to examine discourse patterns, sentiment, topical themes, and public experiences with online activism.

What I did

I contributed to the computational analysis of large-scale social media data, including data labeling, measurement design, and interpretation of patterns in online discourse, and helped translate technical findings into accessible research insights for public audiences.

Adding It Up – 2024: Investing in Sexual and Reproductive Health in Latin America and the Caribbean

Sun, 15 Feb 2026 00:00:00 +0000

The Adding It Up 2024 analysis estimates the need, impact, and cost of fully meeting sexual and reproductive health (SRH) needs in low- and middle-income countries (LMICs). This fact sheet focuses on the estimates of the Latin America and the Caribbean and its subregions. It evaluates services including contraception, maternal and newborn care, abortion services, and treatment of major curable STIs

What I did

I contributed to the data delivery and production of this data essay, including data analysis, visualization feedback, and narrative feedback.

Adding It Up – Country Profile Dashboard

Sun, 15 Feb 2026 00:00:00 +0000

A dashboard presenting sexual and reproductive health estimates for 128 low- and middle-income countries, including indicators on contraceptive use, unintended pregnancy, abortion, and maternal health. The dashboard is designed to make complex research outputs accessible and actionable for policymakers, advocates, and the public.

What I did

I led research translation and data preparation for the dashboard, supporting the design team in transforming large-scale sexual and reproductive health estimates into accessible, policy-relevant outputs.

Computational methods for redacting identifying information in large text data

Sun, 15 Feb 2026 00:00:00 +0000

This blog post explains how researchers used computational methods to redact identifying information from unstructured text data, a set of 1,314 mission statements from U.S. K-12 school districts, before releasing it publicly. Removing identifiers like district names is straightforward in structured datasets, but much harder with free-form text because there are no fixed labels for names or addresses. To tackle this, the researcher combined three different techniques:

Exact name matching against an external list of known district names,
Named Entity Recognition (NER) with pretrained models to detect organization names, and
Regular expressions to spot patterns like capitalized words preceding “school” or “district.”

Each approach had limitations on its own, so they were used together to maximize correctly redacted terms while minimizing false positives.

What I did

I developed and evaluated scalable NLP approaches for detecting and removing personally identifiable information from large text corpora, and narrated the implications of these methods for research transparency and data privacy in a public-facing blog post.

Dataset: U.S. School District Mission Statements

Sun, 15 Feb 2026 00:00:00 +0000

This dataset release provides the underlying text corpus used in Pew Research Center’s analysis of U.S. school district mission statements, enabling researchers to examine how districts articulate goals, values, and priorities in official language. The dataset includes 1,314 mission statements collected from public school district websites across the United States, along with metadata such as district location, political context, and demographic characteristics.

The dataset supports investigation of themes such as diversity, equity, and inclusion and how these references vary across political and geographic contexts, available in this research report: School District Mission Statements and the Politics of DEI in K–12 Education.

The data has been processed to remove personally identifiable information while preserving the integrity of the text for analysis. This process is explained in this blog post: Computational methods for redacting identifying information in large text data.

What I did

I supported preparation and documentation of the dataset for public release, including structuring text data, validating entries, extracting identifiable information and ensuring usability for secondary analysis and reproducibility.

Most Top-Ranked Podcasts Bring On Guests

Sun, 15 Feb 2026 00:00:00 +0000

This report explores how guest appearances function within top-ranked podcasts and how their use varies across genres. It shows that while inviting guests is a common feature of popular shows, the practice is unevenly distributed: some podcasts rely heavily on guests while others rarely include them, and a relatively small group of recurring guests appears across many programs. The analysis highlights how guest participation reflects genre norms, production styles, and network dynamics within the podcast ecosystem.

What I did

I contributed to the computational analysis behind this report by validating and cleaning LLM-assisted extraction data used to identify guests at scale from episode descriptions, and by translating the findings into a public-facing research narrative and visuals.

School District Mission Statements and the Politics of DEI in K–12 Education

Sun, 15 Feb 2026 00:00:00 +0000

This report analyzes how public school districts across the United States describe their educational priorities and values in official mission statements, and what those statements reveal about broader political and social dynamics shaping education. Drawing on a nationally representative collection of 1,314 district mission statements, the study examines which themes districts emphasize—such as student preparation, safety, community involvement, and academic development—and how frequently they reference diversity, equity, and inclusion.

What I did

I led and executed the study’s large-scale text data collection and analysis, including identifying mission statements from district websites, developing coding frameworks, and interpreting linguistic and topical patterns to surface policy-relevant insights for public audiences.