Live data from Hacker News

Reproducing Hacker News writing style fingerprinting

antirez.com

71–80 of 168 posts

Re: Reproducing Hacker News writing style fingerprinting

#72
It did find my "alt" (really an old account with a lost password), but the rest of the list – all users with very high match scores (0.8+) – is random.

Taking a look at comments from those users, I think the issue is that the algorithm focuses too much on the topic of discussion rather than style. If you are often in conversations about LLMs or Musk or self driving cars then you will inevitably end up using a lot of similar words as others in the same discussions. There's only so many unique words you can use when talking about a technical topic.

I see in your post that you try to mitigate this by reducing the number of words compared, but I don't think that is enough to do the job.

Re: Reproducing Hacker News writing style fingerprinting

#73
post #67
post #57

Earlier quoted context omitted.

The system uses on purpose those simple words, since they are "tellers" of the style of the user in a context independent way. Burrows papers explain this very well, but in general we want to capture low-level structure, more than topics and exact non obvious words used. I tested the system with 10k words and removing the most common words, and you get totally different results (still useful, but not style matching),…

I think grouping users by interests would be a more interesting application. Most users don't have multiple accounts, but everyone probably shares some interests with other users, whom they might enjoy discovering.

There's already plenty of those running everywhere.

Re: Reproducing Hacker News writing style fingerprinting

#74
post #33

It works for me. The accounts I used long time ago are there in high positions. I guess that my style is very distinctive. But I also have seen some accounts that seem to be from other non-native English speakers. They may even have a Latin language as their native one (I just read some of their comments, and, at minimum, some of them seem to also be from the EU). So, I guess, that it is also grouping people by their…

> Probably, we make the same type of mistakes while using English.

That is most likely the case. Case in point: My native language doesn't have articles, so locally they're a common source of mistakes in English.

Re: Reproducing Hacker News writing style fingerprinting

#75
post #49

Earlier quoted context omitted.

> I prefer "ought to" I too like when others use it, since a very easy and pretty universal retort against "you ought to..." is "No, I don't owe you anything".

Are you saying there's a connection between "ought" and "owe"? All I see is "I don't want to hear any criticism".

Yes, "ought" is the past tense of "owe". At some point, the second alternative spelling "owed" was introduced to better separate the two meanings (literal and figurative), but it's still the same word; a similar thing happened with "flower" and "flour", those used to be interchangeable spellings of the same word but then somebody decided that the two meanings of that word should be separated and given specific spellings.

And the construct "you owe it to to " still exists even today but is not nearly as popular as "you should " precisely because it has to state to whom exactly your owe the duty; with "should" it sounds like an impersonal, quasi-objective statement of fact which suits the manipulative uses much better.

Re: Reproducing Hacker News writing style fingerprinting

#76
post #72

It did find my "alt" (really an old account with a lost password), but the rest of the list – all users with very high match scores (0.8+) – is random. Taking a look at comments from those users, I think the issue is that the algorithm focuses too much on the topic of discussion rather than style. If you are often in conversations about LLMs or Musk or self driving cars then you will inevitably end up using a lot of…

In case you haven't seen it, the author addressed this point of topic vs. style in a comment (albeit in a different context): https://news.ycombinator.com/item?id=43708474.

Re: Reproducing Hacker News writing style fingerprinting

#80

Earlier quoted context omitted.

I think if you rule out insanely common words, it might get scary accurate.

Actually, the way that these things work is usually by focusing exclusively on the usage patterns of very common (top 500) words. You get better results by ignoring content words in favor of the linking words.

Interesting, I think it also doesn't help that outside of a throw away on a blue moon, I don't really use alts...
Post reply on HN