What does this have that other slang lists do not?

Three columns. register says who you can say a term to, which is what actually gets learners in trouble. channel separates the terms people type constantly from the ones people say out loud. status with checked tells you whether a term is still current and when a person last looked, so you can judge how much to trust it.

Can I use this dataset commercially?

Yes. The license is CC BY 4.0, so you can use it in a paid app, a course, a model, or any other product, as long as you credit HearMandarin with a link.

Why are there no example sentences or audio?

The example sentences are teaching content written for the site rather than collected data, and audio does not belong in a spreadsheet row. Every row carries a url column pointing at its entry page, where both live.

How reliable are the currency judgements?

They are judgements by a native speaker in Shanghai, checked against dated sources, not corpus frequency counts. Slang varies by age, region and platform. Where a year of origin could not be established from evidence, since says unknown rather than a guess.

How often is this dataset updated?

Whenever terms are added or a currency judgement changes on the site. The version date above is the date of the last such change, not a fixed release schedule.

Open data

Chinese slang, labelled by who you can say it to

Name: HearMandarin Chinese Slang Dataset
Creator: HearMandarin
License: https://creativecommons.org/licenses/by/4.0/

185 terms as CSV and JSON, CC BY 4.0. Each one carries its register, channel and status.

What this is

Slang lists are easy to find and hard to trust. They tell you what a term means and leave out the two things that decide whether you can use it: who you can say it to, and whether anyone still says it. This dataset carries both as columns, on every one of the 185 terms.

The three columns this exists for

register: who you can say it to
Meaning is the easy part. A term marked mates carries a swear or an insult and belongs only among close friends; one marked anyone is fine in a meeting. Here: 42 anyone, 131 casual, 12 mates.
channel: spoken or typed only
23 terms are marked typed rather than spoken: pinyin initialisms such as yyds and xswl, plus written reactions people post but would not say out loud.
status and checked: still current, and when we last looked
Slang lists go stale silently. Every term carries a hot, steady or fading judgement and the month it was reviewed, so you can judge the age of the claim yourself. Here: 24 hot, 156 steady, 5 fading.

Download

Version: 2026-09-09, the date a term or a judgement last changed.

chinese-slang.csvComma-separated, UTF-8, one header row.
chinese-slang.jsonOne object with metadata, field descriptions and a terms array.

Fields12 columns

Field	What it holds
`slug`	Stable identifier, also the last part of the entry page address.
`term`	The term in simplified Chinese. A slash means two related forms taught together.
`pinyin`	Pinyin with tone marks. For letter abbreviations, the phrase the letters stand for.
`category`	Topic: work, internet, feelings, relationships, money, food, or study.
`register`	Who you can say it to. anyone, casual, or mates.
`channel`	spoken, typed, or both.
`since`	Year it came into wide use, pre-internet for older spoken slang, or unknown where no year could be established from evidence.
`status`	hot, steady, or fading.
`checked`	The month (YYYY-MM) that status judgement was last reviewed.
`literal_en`	What the characters literally say, which is often absurd and is half the point.
`real_en`	What it actually means in use.
`url`	The entry page, where the examples and audio live.

License and attribution

CC BY 4.0. Copy, adapt, and use it for any purpose, including commercially, as long as you give credit.

Suggested attribution: HearMandarin (https://hearmandarin.com)

Read the full CC BY 4.0 license

Example sentences and audio are not part of the dataset. They stay on the entry pages, one of which every row links to.

Source and limits

Written in-house by a native speaker in Shanghai, with each currency judgement checked against dated sources in the month the checked column gives. These are judgements, not corpus frequency counts, and slang varies by age, region and platform.

Where a year of origin could not be established from evidence, since says unknown rather than a guess: 109 of 185 rows say that.

Browse the same terms on the site, with audio.

Questions

What does this have that other slang lists do not?: Three columns. register says who you can say a term to, which is what actually gets learners in trouble. channel separates the terms people type constantly from the ones people say out loud. status with checked tells you whether a term is still current and when a person last looked, so you can judge how much to trust it.
Can I use this dataset commercially?: Yes. The license is CC BY 4.0, so you can use it in a paid app, a course, a model, or any other product, as long as you credit HearMandarin with a link.
Why are there no example sentences or audio?: The example sentences are teaching content written for the site rather than collected data, and audio does not belong in a spreadsheet row. Every row carries a url column pointing at its entry page, where both live.
How reliable are the currency judgements?: They are judgements by a native speaker in Shanghai, checked against dated sources, not corpus frequency counts. Slang varies by age, region and platform. Where a year of origin could not be established from evidence, since says unknown rather than a guess.
How often is this dataset updated?: Whenever terms are added or a currency judgement changes on the site. The version date above is the date of the last such change, not a fixed release schedule.

HearMandarin Today Learn Practice Review Me Pro