Re: [twitter-dev] app to block all users ending with numerals
What if Twitter just suspended anyone who followed more than 1,000 users without ever having tweeted? But then their membership would sink dramatically. How about not allowing following past 100 users without tweeting at least once. What is the point of these accounts anyway, unless they are being built up and then sold? They can't be used for spam, since they don't tweet, and generally don't have URLs in their profiles. On Fri, Mar 25, 2011 at 10:47 AM, Dean Collins d...@cognation.net wrote: Lol, someone want to write me an app that blocks all users where their username ends with two or three numbers. This is getting ridiculous. Seems like something that would be pretty easy to achieve via the API don’t you think? Cheers, Dean -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk -- Adam Green Twitter API Consultant and Trainer http://140dev.com @140dev -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk
Re: [twitter-dev] app to block all users ending with numerals
I know a number of people who use twitter as a read only source of information (for instance they may follow only news outlets and celebrity tweeters) and therefore may have large follow counts with zero tweets. This may not be a use case that you are familiar with, but it is a valid use case. Also, I don't know if you are aware of the current limits on following, etc, which are described here, my apologies if you already are : http://support.twitter.com/forums/10711/entries/15364 As for the OP, well, a) if this is what you (or your users) want, just parse the follow messages looking for numerical postfixes and offer the user the user the option to block them, there is no need for an API call specifically to do this. And b) again, you are missing a use case, there are lots of genuine accounts that have numerics postfixed to them, some people use birth years, and some people - perhaps finding that the screen name they wanted is not available in a naked form - will have chosen [screen name]76 or some similar format, or picked a year with some historical connection with their chosen name. It is not safe to simply assume that ending with numerics is sufficient to indicate that the account is used only in the delivery of spam, be that tweet spam or simply follow spam. While the assumption may hold in a large number of cases - and I am not aware of any empirical data that shows what this number is, though I'd be interested to see one - it will undoubtedly include some false positives. HTH hax0rsteve On 25 Mar 2011, at 15:00, Adam Green wrote: What if Twitter just suspended anyone who followed more than 1,000 users without ever having tweeted? But then their membership would sink dramatically. How about not allowing following past 100 users without tweeting at least once. What is the point of these accounts anyway, unless they are being built up and then sold? They can't be used for spam, since they don't tweet, and generally don't have URLs in their profiles. On Fri, Mar 25, 2011 at 10:47 AM, Dean Collins d...@cognation.net wrote: Lol, someone want to write me an app that blocks all users where their username ends with two or three numbers. This is getting ridiculous. Seems like something that would be pretty easy to achieve via the API don’t you think? Cheers, Dean -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk -- Adam Green Twitter API Consultant and Trainer http://140dev.com @140dev -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk
Re: [twitter-dev] app to block all users ending with numerals
On Fri, 25 Mar 2011 15:17:25 +, hax0rsteve hax0rc...@btinternet.com wrote: I know a number of people who use twitter as a read only source of information (for instance they may follow only news outlets and celebrity tweeters) and therefore may have large follow counts with zero tweets. This may not be a use case that you are familiar with, but it is a valid use case. Also, I don't know if you are aware of the current limits on following, etc, which are described here, my apologies if you already are : http://support.twitter.com/forums/10711/entries/15364 As for the OP, well, a) if this is what you (or your users) want, just parse the follow messages looking for numerical postfixes and offer the user the user the option to block them, there is no need for an API call specifically to do this. And b) again, you are missing a use case, there are lots of genuine accounts that have numerics postfixed to them, some people use birth years, and some people - perhaps finding that the screen name they wanted is not available in a naked form - will have chosen [screen name]76 or some similar format, or picked a year with some historical connection with their chosen name. It is not safe to simply assume that ending with numerics is sufficient to indicate that the account is used only in the delivery of spam, be that tweet spam or simply follow spam. While the assumption may hold in a large number of cases - and I am not aware of any empirical data that shows what this number is, though I'd be interested to see one - it will undoubtedly include some false positives. HTH hax0rsteve On 25 Mar 2011, at 15:00, Adam Green wrote: What if Twitter just suspended anyone who followed more than 1,000 users without ever having tweeted? But then their membership would sink dramatically. How about not allowing following past 100 users without tweeting at least once. What is the point of these accounts anyway, unless they are being built up and then sold? They can't be used for spam, since they don't tweet, and generally don't have URLs in their profiles. On Fri, Mar 25, 2011 at 10:47 AM, Dean Collins d...@cognation.net wrote: Lol, someone want to write me an app that blocks all users where their username ends with two or three numbers. This is getting ridiculous. Seems like something that would be pretty easy to achieve via the API don’t you think? Cheers, Dean -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk -- Adam Green Twitter API Consultant and Trainer http://140dev.com @140dev -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk 1. There are plenty of good spam detection and filtering algorithms. The ones listed here, however, are simple hacks unlikely to work without extensive manual intervention. The same can be said for ManageFlitter, TwitCleaner and similar services. They give you a start, but you still have to wade through hundreds or thousands of positives to weed out the keepers. 2. A first name followed by a few numbers is a common legitimate screen name - just having a name like that isn't necessarily an indication of a spammer. Here's how it works - Bobby asks Kelly if she's on Twitter. Kelly says No and signs up. She starts with the screen name Kelly, finds it's taken, so she adds her age or the year she was born. If that's taken too, she'll maybe get clever and pick something like PiercedChick, or she'll pick a few random numbers and get in as Kelly117. (Now don't go blaming me if you start getting followers with names like PiercedChick117.) 3. The User / Twitter spam reporting process could definitely be improved with a few simple steps. I don't have any data - that would have to come from inside Twitter - but the two most common types of spam I see is spambots riding Trending Topics and spambots replying to keywords. In either case, the actual spam tweets sent are usually easily found via Twitter Search. Given that, what I do when I get a spam tweet is perform the search, then go through the resulting page and manually report a page or so, depending on how much time I'm willing to spend on this. So here's what I'd propose: Twitter sets up an email address or some other mechanism to receive these search patterns. When someone gets spammed, they can send a copy of the tweet to Twitter, in addition to doing a block and report on the spammer. Twitter could then create the search pattern, run the query and suspend
Re: [twitter-dev] app to block all users ending with numerals
On Fri, 25 Mar 2011 11:33:30 -0700, M. Edward (Ed) Borasky zn...@borasky-research.net wrote: [snip] One other note - a tweet that contains multiple Trending Topics is nearly always spam. I haven't gathered any data, mostly because I'm too lazy to write the API call management / rate limit logic to automate this. I'd *almost* be willing to recommend to Twitter that they stop indexing tweets for Search that match more than one Trending Topic, though. ;-) -- http://twitter.com/znmeb http://borasky-research.net A mathematician is a device for turning coffee into theorems. -- Paul Erdős -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk
Re: [twitter-dev] app to block all users ending with numerals
Today someone tweeted a quote on steve jobs to me. I responded to him referencing the same quote. I got two mentions since Steve Jobs was in both my tweets from an id @RT_steve_jobs . I consider this spam. What would the general opinion be. This does not have any numerics :). Regards Umashankar Das On Sat, Mar 26, 2011 at 12:03 AM, M. Edward (Ed) Borasky zn...@borasky-research.net wrote: On Fri, 25 Mar 2011 15:17:25 +, hax0rsteve hax0rc...@btinternet.com wrote: I know a number of people who use twitter as a read only source of information (for instance they may follow only news outlets and celebrity tweeters) and therefore may have large follow counts with zero tweets. This may not be a use case that you are familiar with, but it is a valid use case. Also, I don't know if you are aware of the current limits on following, etc, which are described here, my apologies if you already are : http://support.twitter.com/forums/10711/entries/15364 As for the OP, well, a) if this is what you (or your users) want, just parse the follow messages looking for numerical postfixes and offer the user the user the option to block them, there is no need for an API call specifically to do this. And b) again, you are missing a use case, there are lots of genuine accounts that have numerics postfixed to them, some people use birth years, and some people - perhaps finding that the screen name they wanted is not available in a naked form - will have chosen [screen name]76 or some similar format, or picked a year with some historical connection with their chosen name. It is not safe to simply assume that ending with numerics is sufficient to indicate that the account is used only in the delivery of spam, be that tweet spam or simply follow spam. While the assumption may hold in a large number of cases - and I am not aware of any empirical data that shows what this number is, though I'd be interested to see one - it will undoubtedly include some false positives. HTH hax0rsteve On 25 Mar 2011, at 15:00, Adam Green wrote: What if Twitter just suspended anyone who followed more than 1,000 users without ever having tweeted? But then their membership would sink dramatically. How about not allowing following past 100 users without tweeting at least once. What is the point of these accounts anyway, unless they are being built up and then sold? They can't be used for spam, since they don't tweet, and generally don't have URLs in their profiles. On Fri, Mar 25, 2011 at 10:47 AM, Dean Collins d...@cognation.net wrote: Lol, someone want to write me an app that blocks all users where their username ends with two or three numbers. This is getting ridiculous. Seems like something that would be pretty easy to achieve via the API don’t you think? Cheers, Dean -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk -- Adam Green Twitter API Consultant and Trainer http://140dev.com @140dev -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk 1. There are plenty of good spam detection and filtering algorithms. The ones listed here, however, are simple hacks unlikely to work without extensive manual intervention. The same can be said for ManageFlitter, TwitCleaner and similar services. They give you a start, but you still have to wade through hundreds or thousands of positives to weed out the keepers. 2. A first name followed by a few numbers is a common legitimate screen name - just having a name like that isn't necessarily an indication of a spammer. Here's how it works - Bobby asks Kelly if she's on Twitter. Kelly says No and signs up. She starts with the screen name Kelly, finds it's taken, so she adds her age or the year she was born. If that's taken too, she'll maybe get clever and pick something like PiercedChick, or she'll pick a few random numbers and get in as Kelly117. (Now don't go blaming me if you start getting followers with names like PiercedChick117.) 3. The User / Twitter spam reporting process could definitely be improved with a few simple steps. I don't have any data - that would have to come from inside Twitter - but the two most common types of spam I see is spambots riding Trending Topics and spambots replying to keywords. In either case, the actual spam tweets sent are usually easily found via Twitter Search. Given that, what I do when I get a spam tweet is perform the search, then go through the
Re: [twitter-dev] app to block all users ending with numerals
On 25 Mar 2011, at 18:49, M. Edward (Ed) Borasky wrote: One other note - a tweet that contains multiple Trending Topics is nearly always spam. I haven't gathered any data, mostly because I'm too lazy to write the API call management / rate limit logic to automate this. I'd *almost* be willing to recommend to Twitter that they stop indexing tweets for Search that match more than one Trending Topic, though. ;-) To pointlessly prolong the discussion - it being Friday :-) ... Of course that suffers from the same problems, my personal feed follows a lot of UK politics, most of this action takes place during the day when much of the UK is busy. It is not all uncommon to see two trending topics or hashtags in these tweets, and this is especially true when something like a televised debate or random bit of civil unrest is taking place. Instructive for anyone who wishes to see this in action will be the large demonstration in central London tomorrow, I would predict that within several hours of the beginning of the march, if not before, you will see many UK tweets containing at least two TTs, if not three. Indeed, it is a common practice amongst such twitterers - particularly the more provocative and/or confrontational - to include as many popular hashtags as they can fit onto the end of their tweet so as to reach the highest number of people who may have added these to their searches, resulting in multiple trends in some otherwise very short tweets. Now it is certainly the case that these tweets amount to a vanishingly small number of the total number of tweets in any given timeframe, and that the number of people who are interested in them is a vanishingly small proportion of even the UK user base, never mind the entire global base, but for those users those tweets are precisely what they are using twitter for. I guess what I'm getting at here is that any automated filtering system ultimately amounts to making value judgements on behalf of your users. That this fails quite often in - for example - corporate email systems gives me no confidence that any similar approach is going to work for twitter, where the diversity of message content, users, and use cases is vastly more pronounced. But I could - of course - be wrong :-) hax0rsteve -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk
Re: [twitter-dev] app to block all users ending with numerals
On Fri, 25 Mar 2011 19:58:58 +, hax0rsteve hax0rc...@btinternet.com wrote: To pointlessly prolong the discussion - it being Friday :-) ... [snip] I guess what I'm getting at here is that any automated filtering system ultimately amounts to making value judgements on behalf of your users. That this fails quite often in - for example - corporate email systems gives me no confidence that any similar approach is going to work for twitter, where the diversity of message content, users, and use cases is vastly more pronounced. But I could - of course - be wrong :-) Well, Twitter is on the one hand a smaller data set than Google, but on the other hand has different usage patterns in the real-time signal-processing sense. So yes, if Google has to mix human judgment and algorithmic judgment to optimize shareholder value, then so does Twitter. I claim, though, that the mere fact that one can buy the number one position in a Twitter Search that otherwise returns total garbage is very much different from buying clicks on Google, where organic search results at least return something that a mix of human and mechanical judgment has determined is relevant to the searchers' intent. Twitter Trending Topics is broken and infested with spam. One shouldn't need Sulia to consume Twitter, and Twitter's own Promoted Trends and Tweets should not have to compete for eyeballs and clicks with spambots. -- http://twitter.com/znmeb http://borasky-research.net A mathematician is a device for turning coffee into theorems. -- Paul Erdős -- Twitter developer documentation and resources: http://dev.twitter.com/doc API updates via Twitter: http://twitter.com/twitterapi Issues/Enhancements Tracker: http://code.google.com/p/twitter-api/issues/list Change your membership to this group: http://groups.google.com/group/twitter-development-talk