Journalism Portfolio

One of my unique strengths is organizing and analyzing unstructured or semi-structured qualitative data. In an academic research project, rather than a conventional literature review, I built a topic taxonomy and a six-point sentiment scale to examine how sampling choices (especially snowball sampling) can influence tone in academic studies.

Sentiment analysis involves at least some subjective judgment, so I prioritized transparency and replicability. I designed a structured rubric that paired guiding questions with point values so it’s possible to audit edge cases or extend the dataset without rebuilding the coding framework.

Artificial intelligence (AI) was used selectively to support early-stage content review, data cleaning and preliminary categorization. However, at the time, the model consistently underperformed at identifying complex sentiment. It’s possible the problem is that academic papers use neutral scholarly language even while implicitly endorsing or critiquing multi-level marketing (MLM) structures, so final sentiment decisions were made through manual review guided by the rubric.

One revision I would make in a future iteration is to merge the topic category “exploitation” with “regulation.” When sentiment was analyzed by category, it’s hard to define a situation where “exploitation” could skew toward positive sentiment, which limited its usefulness as a standalone category. In practice, exploitation themes frequently overlapped with regulatory discussions, and integrating exploitation into “regulation.”