Data & Tools

Peer-reviewed corpora on South Korean politics, with the open-source tools built to work with them.

Published as a data paper Lim, T. H. (2025). South Korean Election Campaign Booklet and Party Statements Corpora. Scientific Data, 12, 1030. Publication: https://doi.org/10.1038/s41597-025-05220-4
Data (OSF): https://doi.org/10.17605/OSF.IO/RCT9Y

Data

South Korean Election Campaign Booklet Corpus (2000-2022)

The South Korean Election Campaign Booklet Corpus is a collection of manifesto pamphlets of individual candidates who ran for offices in single-member or multi-member districts during six National Assembly elections, six local elections, and five Presidential elections in South Korea between 2000 and 2022. The dataset has 49,678 observations across 31 variables. It is distributed through krpoltext v0.2.0, which supports managed CSV/Parquet artifacts, metadata and schema helpers, and a refreshed static JSON API.

South Korean Party Statements Corpus (2003-2022)

The South Korean Party Statements Corpus is a collection of official statements from the two major parties in South Korea from 2003 to 2022. The dataset comprises a total of 35,115 entries from the Conservative Party and 48,086 entries from the Progressive Party (83,201 total, 9 variables). It is distributed through krpoltext v0.2.0, which supports managed CSV/Parquet artifacts, metadata and schema helpers, and a refreshed static JSON API.

Tools

krpoltext v0.2.0 — R package

The R package that distributes both corpora. It loads either corpus in a single line, verifies and caches the files with SHA-256, reads CSV or Parquet, exposes metadata and schema helpers, and bridges directly to quanteda. Documentation and a static JSON API are on the package site.

kr-elections-mcp v0.1.0 — MCP server

A Model Context Protocol (MCP) server that lets an AI assistant read the corpora and South Korea's official election records in plain language — search manifesto and statement text, look up candidates and results, and link an official candidate record to their corpus manifesto. It runs locally: the corpus tools work out of the box, while the election-record tools wrap the NEC Open API and need your own NEC service key.

Three ways to reach the data

  1. Download the files. Full CSV and Parquet on the Open Science Framework, with a citable DOI — the shortest path for replication.
  2. Work in R. The krpoltext package loads, filters, and caches either corpus and bridges to quanteda.
  3. Ask an AI agent. The kr-elections-mcp server lets an assistant read the corpora and official election records in plain language.

How the pieces fit

Two corpora source krpoltext R package static JSON API any language kr-elections-mcp AI agents NEC Open API external · gov't records
The same corpora, reachable three ways — files for anyone, R for researchers, and AI agents via the MCP server, which also wraps the government's election records.

How to cite & acknowledge

Data & paper
Cite the data paper and the OSF dataset. The full descriptor lives at Research → two political text data sets.
Software
Acknowledge krpoltext via citation("krpoltext") and its version tag.
MCP server
Cite the data it serves; acknowledge the server by name and version.

Data licensed CC BY-NC-ND 4.0; code licensed MIT. Attribution and the license travel with any copy.

Sibling project: korean-law MCP — a Korean legal-domain server (separate domain, listed for reference).