This is an interesting idea - though, I am sure quite difficult to do effectively. I have mentioned this before, but there is a company called Banjo doing something similar: http://ban.jo/ They track public postings (Twitter, Facebook, etc.) that have geographic information and look for anomalies for each geographic location.
The fact that they have paying customers implies that they are able to provide some kind of value from their efforts. Though, since it is a private company, I have no idea what their methods are, and how effective their technology is. It would be interesting to see an approach using HTMs, and in a more public way. All that said, my guess is that what they are providing is a real-time alert system that something anomalous IS happening - as opposed to predicting that something like a terror attack is about to happen - which would be much, much harder. - Jeff On Mon, Nov 16, 2015 at 1:35 PM Chandan Maruthi <[email protected]> wrote: > Pascal and Matt, > There seems to be a huge potential in classifying identified patterns and > being able to trigger actions based on them. > When you see a particular event a few times, it may not be an anomaly any > more, but that does not make it useless. The fact that > this has been seen before and classified as something of interest can be > of significant value. Anomalies lies at the edges > of known and unknown knowledge, they are critical in triggering learning . > but the learnt patterns through classification are > as much or way more important in stable state. > > chandan > > On Mon, Nov 16, 2015 at 1:19 PM, Pascal Weinberger < > [email protected]> wrote: > >> Thanks Matt, >> >> A problem I see now, is that they are under constant attack on Social >> media, so learning any patterns as not anomlous will be a huge problem, and >> possible contextual infromation will need to be taken into consideration... >> >> See for example: >> >> >> http://techcrunch.com/2015/11/16/hacking-collective-anonymous-declares-total-war-on-isis-following-paris-terror-attacks/ >> >> http://www.mirror.co.uk/news/world-news/anonymous-declares-war-islamic-state-6839030 >> and many more... >> >> Still I think it is worth a try, but am not really able to allocate time >> :( >> Is anybody form Cortical listening?? :) (hint) >> >> >> Best, >> >> Pascal >> >> On Mon, Nov 16, 2015 at 9:43 PM, Matthew Taylor <[email protected]> wrote: >> >>> I thought about what happened in Paris all day yesterday. I spent >>> considerable time researching ISIS and I found this: >>> >>> >>> http://www.brookings.edu/~/media/research/files/papers/2015/03/isis-twitter-census-berger-morgan/isis_twitter_census_berger_morgan.pdf >>> >>> It describes it a lot of detail how ISIS uses Twitter, and it also >>> lists a lot of topics and hashtags ISIS promoters have used. >>> >>> Regards, >>> --------- >>> Matt Taylor >>> OS Community Flag-Bearer >>> Numenta >>> >>> >>> On Sat, Nov 14, 2015 at 3:44 AM, miguel angel rodriguez >>> <[email protected]> wrote: >>> > Hi Pascal, >>> > >>> > I really think is a naive idea, but very interesting too. Creating a >>> > WorldWide-SocialInfo processor that can detect these kind of anomalies >>> would >>> > serve a fantastic purpose. But this is a Big project and should be >>> addressed >>> > with the care and workforce appropirate. >>> > I kindo of Isac Asimovs "Multivac" WorldManaging Computer. >>> > >>> > If someone opnes a serious project on the matter I would be in as soon >>> as my >>> > current tasks are finished. >>> > >>> > Best, >>> > >>> > Miguel Angel >>> > >>> > #PrayForParís >>> > >>> > 2015-11-14 11:39 GMT+01:00 Pascal Weinberger < >>> [email protected]>: >>> >> >>> >> Hey NuPIC, Hey Cortical.io >>> >> >>> >> >>> >> After this devastating night with the disturbing and also a bit >>> awakening >>> >> events in Paris, I would like to ask for people to join me on a small >>> and >>> >> probably naive idea: >>> >> >>> >> The cortical.io team has submitted a twitter monitoring service for >>> >> elections, there has been a hack on the hackaton in NY for monitoring >>> news >>> >> and indicating anomalies in them, and now I would like to see if we >>> can find >>> >> indications for the terror-attacks in Paris in Social media. >>> >> >>> >> Twitter, news and "underground" platforms like 4chan are maybe good >>> places >>> >> to start. I was thinking about monitoring the schemantics and numbers >>> of >>> >> posts containing keywords related to terrorism and hatred. >>> >> >>> >> I was hoping for sb from cortical.io and others to help trying to >>> see if >>> >> there are indications in Social media before the events, so anomalies >>> in >>> >> posts and activities which may have been used to warn and/or find >>> >> individuals who are causing these anomalous behavior. >>> >> >>> >> I know this may sound naive (I know it is) but I think it's worth the >>> try! >>> >> Maybe this Leeds us to a warning tool for terrorism, which would be >>> >> incredibly helpful to us. >>> >> >>> >> >>> >> I hope for feedback and ideas here! >>> >> >>> >> >>> >> PS: I hope you had a great rest time yesterday night :) >>> >> >>> >> >>> >> Best, >>> >> >>> >> Pascal >>> >> >>> >> ____________________________ >>> >> >>> >> BE THE CHANGE YOU WANT TO SEE IN THE WORLD ... >>> >> >>> >> >>> >> >>> >> >>> > >>> >>> >> > > > -- > Regards > Chandan Maruthi > >
