Reading Public Discourse Through 5,000 Comments

In college I completed an independent data analysis project around People’s Daily related videos and comments.

The technical workflow itself was familiar: crawling, cleaning, visualization, text analysis, sentiment analysis, and topic clustering. But the project taught me that data becomes useful only when it is held against a clear theoretical question.

I kept bringing the work back to communication theory: agenda setting, framing, and the way emotion and stance shape audience response.

That habit still follows me in RAG, agents, and benchmarks: clarify the problem first, choose the method second, and never confuse a metric with truth.