Hacker News Reader

Comment by ebg13 | Hacker News Reader

It took me a minute to realize that "classified" in the headline means labeled, not secret.

Sorry, I wanted to change it after I push the submit button... it was too late.

Maybe @dang can help, here? :)

I disagree that there is something that needs to be changed here. "Dataset of classified X" is commonly used to refer to this exact thing. (i.e. dataset of data points with human-made labels)

But the word has multiple meanings and both interpretations are equally plausible. Personally I also thought the title was about secrets. Changing from classified to labeled will remove ambiguity completely.

It's the ordering here:

"... classified screenshots ..." would be usually be understood in en-gb to be "screenshots given a security rating requiring restricted access".

"... screenshots classified ..." would be "screenshots given a classification of some sort".

I disagree - the former would be used in both cases in my opinion. I think it just depends which usage you use more.

It is used in both situations when context makes it clear, there's no [or very little, at least] context in a title. Ordering disambiguates the intended meaning when context is insufficient.

Disagree; "classified" is normally used before the noun, not after, when referring to government classification. E.g. it's "classified material" not "material that's classified".

Erm, that's what I did. You said disagree and then repeated my first point.

I think something like "categorization" would be less ambiguous though

But "classification" is the terminology. "Categorization" would be ok if we go back X many years in CS/Stats literature and rename it as "categorization". I think "categorized" would be more confusing here. In this case, I would compromise to something like "Dataset of labeled screenshots from [...]".

HN is not an audience of CS and Stats academics, or those familiar with the literature in those fields.

You don't need to be an academic to be familiar with a discipline's terminology. If you keep reading articles like OP, at some point you'll learn that by convention people use the word "classification" for things like this.

But in HN, the more common usage is government classification, so headlines should be based on that precedent.

It's ambiguous, that's the problem. My first reading was it was a dataset of government information that had leaked on Tor (despite having a computer science background - that's not necessarily the context on HN). Even if it's accurate, changing it to something that is less ambiguous is useful. "Labelled dataset of screenshots" would remove the ambiguity, and it uses fairly standard terminology.

Reply on news.ycombinator.com