ㅠㅠ Feeling in Numbers

Leiden University · Korean Studies · a PRE-Class text lab

Feeling in Numbers 감정을 세다

50,000 people wrote short reviews, in Korean, of films they loved and films they hated. Now you are going to read all of them at once — by counting.

50,000reviews
½ 긍정 · ½ 부정half loved it, half didn't
0lines of code required
the corpus · 말뭉치

A corpus is a body of text treated as data. Ours: 50,000 short film reviews from Naver, Korea's biggest web portal — typed by ordinary moviegoers (collected 2015), each under 140 characters, slang and typos intact. Half come from 9–10★ ratings (긍정, the loves) and half from 1–4★ (부정, the hates), so for any word you can ask: which side uses it more? Cleaned for classroom use — full details in how this works.

begin experiment 01

Count the words

If you piled up every word in 50,000 reviews, which would rise to the top?

word type

    Click any word to put it under the microscope in experiment 03.

    The tug of war

    Some words pull a review toward love, some toward hate. Which side is each word on?

    ← pulls toward hate 부정 긍정 pulls toward love →

    Probe the corpus

    Pick an expression — or type your own — and see how reviewers actually use it.

    The corpus is waiting. ㅋㅋ is a good place to start.

    The face-off 대결

    Two expressions enter; 50,000 reviewers judge. Click two contenders from the roster below, bet on which one shows up in happier reviews — then fight.

    pick from the roster ↓
    …or type any Korean

    the roster — click a contender to put it in the ring

    Each fighter's meter shows how often it appears in loving (긍정) versus hating (부정) reviews. Two ways of laughing — ㅋㅋ and ㅎㅎ — make a good first match. Place your bet, then fight!

    방법 How this works

    The data

    You explored a curated subset of the Naver Sentiment Movie Corpus — 200,000 real movie reviews collected from Naver, Korea's largest web portal, and released into the public domain. We removed duplicates and filtered strong language, then kept a balanced 50,000: half from ratings of 9–10 (긍정), half from 1–4 (부정). About 400 frequently shown example reviews carry hand-made English translations, emoticons intact.

    The method

    This way of reading is called text as data. Words were split with a Korean morphological analyzer (Korean glues grammar onto words, so 영화가, 영화를 and 영화는 must all count as 영화). Then everything is just careful counting — the same first step researchers use on parliamentary records, newspapers, and historical archives.

    What counting misses

    Sarcasm. Context. The difference between crying from grief and crying from joy. Numbers found the pattern in ㅠㅠ; explaining it took a human who understands how Koreans talk about films. That partnership — computation plus interpretation — is what Korean Studies looks like now.

    Want more?

    This lab is part of the PRE-Class Asian Studies at Leiden University, where language, culture, and computational tools meet. Curious how deep this goes? Join the PRE-Class — or peek at the source of this page.