Methods

My Data

I accessed the Cofacts database of social media posts from the Taiwanese social conglomerate app LINE. The Cofacts database, authored by the Taiwanese think-tank CoLabs, is a repository of approximately 210,000 posts with creation dates ranging from 2017 to 2025. Each entry includes the text body of a given post and several categories of metadata, as well as a crowd-sourced label created by teams at CoLabs regarding each post's relationship to confirmed truth. Labels can take one of four values: RUMOR, NOT_RUMOR, NOT_ARTICLE or OPINIONATED. I accessed the database, publicly available on HuggingFace.com, via Python.

My Methods

To perform my analysis, I measure the incidence of posts containing any of the following six keywords: [Lai, Hou, KMT, president, environment, illegitimate]. This list includes the surnames of the two presidential candidates: Lai Ching-Te, the progressive candidate of the Democratic People's Party (DPP) whose stance on unification remains squarely in the center, and Hou Yu-Ih, the conservative and pro-unification candidate of the Kuomintang (KMT). 

The remaining terms act as controls, with the exception of the term "illegitimate," to which is linked a confirmed disinformation narrative in which Lai Ching-Te is implicated as having an unrecognized, illegitimate child.

So, here's what I've done so far

• Method: heatmap. Created a heatmap of the rates of incidence of the above key terms in all posts across time. I detected and recorded posts that contained one or more instance of any of the key terms. This lacked dimensionality; it did not tell the full story. This led me to:

• Method: time series. Created graphs of the incidence of the terms across time. Used a 45-day sliding average window to create smooth curves telling the story of relative levels of keyword-containing posts across time. These revealed several interesting things elaborated in the following tab. This, in turn, led me to:
  Method: co-incidence across time. For one pair of key terms (Lai, illegitimate), selected for their superlative relevance to a key narrative explored across this research, a graph was made demonstrating a) the incidence of posts containing one or more mention of "Lai" in English or Chinese, b) the incidence of posts containing one or more mention of "illegitimate" in English or Chinese, and c) the co-incidence of posts that were in both of these categories.