Q: What do you call a language that doesn't change year after year?
A: Dead.
There was a great interview with the folks at the OED about why they added words to the dictionary that someone had just "made up" (like 'selfie'). And their answer was simply that all words are made up at some point, and trying out a new word exposes it to the population and ones that effectively communicate a concept or idea get picked up and distributed. That is how the language evolves to deal with a world and a society that evolves.
On the flip side, many folks are rather annoyed that technologists took previously fine words and repurposed them for their own use and now they speak a language that sounds like English but isn't understandable by non-techies.
I'm guessing the OED folks would have preferred that new words were created to cover that case :-)
> On the flip side, many folks are rather annoyed that technologists took previously fine words and repurposed them for their own use and now they speak a language that sounds like English but isn't understandable by non-techies.
It's worse in eg German, where we take meanings that already have a perfectly fine word, and choose a new English-inspired word for them.
(Or in English as well, where people use `ask' as a noun when `request' is much less gruesome.)
One might also consider that a dictionary should be considered a service, to provide the user a means to identify and understand unfamiliar language. I think, too often, people treat dictionaries as official authority on language. But I'm not aware that there is any such thing as a central language authority for any language other than French. So, for a dictionary to include a word like "selfie" is not necessarily a definitive inclusion of the word in the language (as literally no-one holds that authority), but a statement that new things are going on that readers might, just perhaps, want to know about to stay relevant.
> I think, too often, people treat dictionaries as official authority on language.
They aren't, and a lot of people miss it - they don't know that dictionaries are descriptive, not proscriptive. They do not define words, they only catalogue ones that are used, providing their most popular meanings.
At one time I probably would have tried to argue that "selfie" isn't a "real" word. And yet it's a sequence of letters (or phonemes) that conveys a meaning among a large number of English speakers, so in what sense isn't it a word? Because it wasn't a word last year? Dictionaries are inevitably descriptivist, they just have to decide what their thresholds are, in terms of time and numbers.
"selfie" isn't a made up word; it existed all along, in search of a meaning. It's formed according to rules which give us words like "hankie", "softie", "oldie", "meanie", "baddie".
Incidentally, the pattern generates "fraudie", which is potentially applicable in the given context.
The above sentence looks like English syntax, but doesn't appear in in the Oxford List of Valid English Sentences. Therefore, it isn't English. Clearly, you just made it up.
Please stick to English sentences from the approved list if you want people to understand you.
One of the best things about the culture around English, the thing I absolutely love, is the acceptance for creating ad-hoc - sometimes even throwaway - words. Things like verbing, which are done without thinking by English speakers, are sadly not so welcome in many other languages.
Some useful patterns, but also quite likley noise.
Both Alaska and Delaware are small states, with (relatively) small populations. For classifiers with uneven numbers of members (e.g., states), odds are high that whatever your outlier member is will be a lower-population classifier. It's simply a matter of variance and other elements.
To test for actual significance of those findings, you'd want to look at Monte Carlo simulations through your dataset over time to see if there's a consistent trend for these particular indicators, or if the locus shifts among several other regions.
Other indicators such as multiple accounts and time/day of activity suggest stronger causal relationships.
This post is not useful or meant to be useful, I think. It's just supposed to be entertaining enough to get you to think about Sift. You shouldn't blacklist Delaware addresses or elderly people…
That said, your point about noise here is a good and interesting one
I would also add that even if these cases aren't noise, they're the ones that are easiest to detect due to their "strangeness". The followup question is then whether these cases are representative of all fraud or are just the easy ones to spot. To answer that, you probably need to cross-validate with signals besides just transaction logs, e.g. evidence from a physical raid or insider tips. Otherwise, it seems like you're bound to miss the skillful frauds.
I wonder if the age range identified is because these fraudsters speed through the sign up and select the low value when asked their age. I wonder if they are either 1/1 or 12/31 babies too.
That is definitely a trend. Similarly, Alaska (first in the abbreviated drop down list for states) could be fraudulent due to the same reason. I don't believe that the fraud users are actually from Alaska.
Fraud doesn't just cause merchants lost goods. It makes ordering from stores more difficult when there are extra steps (at least 3-D secure) just for fraud reduction. This both annoys customers and causes sales to be lost because ordering is less convenient.
It causes unnecessary returns when merchants end up sending things to non-existent addresses that someone just invented to see if their stolen card was working or not.
It even makes things like A/B testing less reliable, when your numbers are skewed by fraud. Eliminating an extra step in your order process might look like a great A/B testing win. Unless it was just because that "optimization" just made fraud easier to commit. At the very least it adds noise.
It also makes it more difficult to know where you as a merchant stand financially, as orders can become reversed in the future. Even if you seemingly turned a $1000 profit this month, later you might find out you actually didn't.
It's a sad situation that we have to just try to guess who might be committing fraud or not, sometimes denying service to perfectly legitimate users while still missing many cases of actual fraud.
A little off-topic, but today there was a pretty entertaining post on Gawker about ISIS follower accounts who were caught talking about mundane stuff and having run-of-the-mill "Twitter" drama:
What was funny was not just the purported content of the tweets -- now apparently removed by Twitter -- but how these guys were identified:
> ...Abu Yusuf Al-Jabarti is an avid tweeter (his handle, @AlJabarti42, indicates he’s been banned 41 times) and supporter of the Islamic State. Most of his tweets are like this, just trying to expand his brand like everyone else...
I'm not saying it's easy to write a general algorithm that follows a rule of "If an account gets banned an another account with the same name but a Levenshtein distance of 1 sprouts up from the same IP block and its first tweets contain similar content to the deleted account, then ban that new account, too"...at least, it wouldn't be easier than removing these accounts ad-hoc (i.e. after Gawker discovers them)...but some problematic users don't even make themselves hard to find and yet the prospective computational solution isn't necessarily practical to implement or particularly worth anyone's time (at the moment...).
Systems like Sift are interesting. Machine learning can pick up a lot of trends over time that humans can't. The problem is that they often fail to react quickly to new trends, many of which hit hard quickly. I've talked to a lot of these vendors and most of them are complimenting machine learning systems with traditional rule-based systems and human review because ML by itself is too slow in adapting to attackers.
29 comments
[ 6.0 ms ] story [ 83.6 ms ] threadA: Dead.
There was a great interview with the folks at the OED about why they added words to the dictionary that someone had just "made up" (like 'selfie'). And their answer was simply that all words are made up at some point, and trying out a new word exposes it to the population and ones that effectively communicate a concept or idea get picked up and distributed. That is how the language evolves to deal with a world and a society that evolves.
On the flip side, many folks are rather annoyed that technologists took previously fine words and repurposed them for their own use and now they speak a language that sounds like English but isn't understandable by non-techies.
I'm guessing the OED folks would have preferred that new words were created to cover that case :-)
It's worse in eg German, where we take meanings that already have a perfectly fine word, and choose a new English-inspired word for them.
(Or in English as well, where people use `ask' as a noun when `request' is much less gruesome.)
I do like `prepone', and `whelmed' though.
"Sorry, I didn't get the TPS report done. I was circumwhelmed by Hacker News and thedailywtf.com"
They aren't, and a lot of people miss it - they don't know that dictionaries are descriptive, not proscriptive. They do not define words, they only catalogue ones that are used, providing their most popular meanings.
Incidentally, the pattern generates "fraudie", which is potentially applicable in the given context.
The above sentence looks like English syntax, but doesn't appear in in the Oxford List of Valid English Sentences. Therefore, it isn't English. Clearly, you just made it up.
Please stick to English sentences from the approved list if you want people to understand you.
Both Alaska and Delaware are small states, with (relatively) small populations. For classifiers with uneven numbers of members (e.g., states), odds are high that whatever your outlier member is will be a lower-population classifier. It's simply a matter of variance and other elements.
To test for actual significance of those findings, you'd want to look at Monte Carlo simulations through your dataset over time to see if there's a consistent trend for these particular indicators, or if the locus shifts among several other regions.
Other indicators such as multiple accounts and time/day of activity suggest stronger causal relationships.
That said, your point about noise here is a good and interesting one
So that explains the holds I sometimes get on orders, where I am like why hold up a $20 order.
Regarding Deleware, I bet there are drop mailers there, surprised there isn't a database of those.
It causes unnecessary returns when merchants end up sending things to non-existent addresses that someone just invented to see if their stolen card was working or not.
It even makes things like A/B testing less reliable, when your numbers are skewed by fraud. Eliminating an extra step in your order process might look like a great A/B testing win. Unless it was just because that "optimization" just made fraud easier to commit. At the very least it adds noise.
It also makes it more difficult to know where you as a merchant stand financially, as orders can become reversed in the future. Even if you seemingly turned a $1000 profit this month, later you might find out you actually didn't.
It's a sad situation that we have to just try to guess who might be committing fraud or not, sometimes denying service to perfectly legitimate users while still missing many cases of actual fraud.
http://gawker.com/even-isis-guys-have-twitter-drama-17405414...
What was funny was not just the purported content of the tweets -- now apparently removed by Twitter -- but how these guys were identified:
> ...Abu Yusuf Al-Jabarti is an avid tweeter (his handle, @AlJabarti42, indicates he’s been banned 41 times) and supporter of the Islamic State. Most of his tweets are like this, just trying to expand his brand like everyone else...
I'm not saying it's easy to write a general algorithm that follows a rule of "If an account gets banned an another account with the same name but a Levenshtein distance of 1 sprouts up from the same IP block and its first tweets contain similar content to the deleted account, then ban that new account, too"...at least, it wouldn't be easier than removing these accounts ad-hoc (i.e. after Gawker discovers them)...but some problematic users don't even make themselves hard to find and yet the prospective computational solution isn't necessarily practical to implement or particularly worth anyone's time (at the moment...).
what does that even mean? it's always 3 am somewhere in the world.