I just read Seth Grimes’ article on Expert Analysis: A Case for Socialization of Data. What a great article and very timely, too! Seth’s comment “people post facts (true or not)… to blogs, Facebook, Twitter, …” provided real food for thought (cheese, if you will) and left me thinking about the need to apply advanced analytics to identify fraudulent social media comments. I think, in time, it will be commonplace! Insurance fraud, tax fraud, identity theft,… (see a huge list of fraud types http://en.wikipedia.org/wiki/Fraud ) are generally considered to be rare events – that is, they d
on’t happen a lot. The analytics around catching fraudulent behavior can be akin to finding a needle in a haystack. But fraudulent actions look different to the actions of the regular Joe in one dimension or another. Depositing amounts of less than $10,000 in an individual bank account over the space of a month would be commonplace. But, if the average banking customer makes 50 such deposits in a month, making >100 deposits would be considered atypical and cause for closer investigation. Analyzing content (via text analytics, text mining, social media analysis, sentiment analysis or other related technologies) provides organizations with the ability to determine typical social media comments vs atypical social media comments. If something stands out as being odd, it probably is. If it is too good to be true, again, it probably is. In the lyrics of the famous kid’s song, The Farmer in the Dell – THE CHEESE(the out of place text comments) STANDS ALONE! The challenge is to use text analytics to identify fact or fiction – a good challenge for the sharp text analytics brains out there!
Photo credit: bbmowery
Tweets that mention Social Media Comments – Fact Or Fiction? Find Out With Text Analytics. | Attensity Blog -- Topsy.com wrote,
[...] This post was mentioned on Twitter by Maria Ogneva and Attensity, Bartholomew Heaven. Bartholomew Heaven said: Social Media Comments – Fact Or Fiction? Find Out With Text …: Social Media Monitoring, Measurement and SocialCR… http://bit.ly/9bXOeI [...]
| Link | July 28th, 2010 at 4:52 pm
Ellie K wrote,
One can’t believe everything one reads/hears from more traditional sources who actually have press credentials. The internet lets EVERYONE jump on in if they wish. And actually, that is why it would be helpful to have some means of gauging credibility / authenticity than what we do now: s slow process of discovery noting inconsistencies, behavioral red flags etc for an on-line community member, “expert” or even colleague, which eventually becomes apparent, but can cause havoc in the interim.
I think this is such a novel concept that it deserves more than a superficial presentation, so that it has a chance to be considered seriously as it deserves.
Re anomalous activity detection for consumer banking transactions you say
First, Federal banking laws require filing paperwork with the IRS when an individual makes three or fewer transactions (withdrawal or deposits) summing to $10,000 or more over 1-month interval. The customer who makes 50 deposits per month: do they total < $10,000? And change in activity level to 100 deposits per month: does total remain < $10,000? If so, that's an average change of $200 per deposit vs $100 per deposit, with lots of assumptions. So this isn't a good example for suggestive behavioral pattern breaks.
There are also some difficult matters pertaining to freedom of speech and privacy, I'd think, regarding the matter of corporations tracking employees activites outside of work
Finally, I do believe that identity theft, tax and insurance theft is actually quite common, else the US Fed Trade Comm wouldn't feel the need to their incredibly easy-to-use and high-traffic identity theft fraud incident reporting system to the public
However, in an appropriate context, what you suggest could be great. And Seth Grimes is such a great subject matter expert!
PS Hope I wasn't too critical. I'm playing a bit of devils advocate, because I'd like to provoke you untill me more about it.
| Link | July 28th, 2010 at 8:20 pm
Manya Mayes wrote,
In response to Ellie K’s comment:
I should have made it clear that the fraud example I used was an example and a rather extreme one at that. In reality fraudsters like to make their activity look as normal as possible. Once fraudulent behavior is identified, the fraudsters change their behavior in an attempt to fly below the radar. Ellie gives specific details of US Federal Banking laws. Undoubtedly these laws would vary across different countries, but the goal remains the same – to identify the characteristics of fraud. Returning to the notion of identifying credible Twitter posts, I’ve heard a lot of people comment that they’d like to identify the useful information on Twitter. I haven’t heard of examples where people are using analytics to identify credibility/authenticity, nor do I think it would be an easy task. And there is always the question of privacy. I do think that analytics could be used to identify areas of expertise, audience reaction to posts, peer comments, crowdsourced peer ratings(much like the ratings attached to the blogs on your blog site – I love the text analytics and the Science Tattoo Emporium posts, BTW) could all potentially be used to identify breadth and depth of knowledge and be factored into a credibility/authenticity rating. Ellie, I’d love to talk to you more about the process you are currently using. Are there any other readers that would be willing to share their experiences?
| Link | August 3rd, 2010 at 2:28 pm
themaria wrote,
.@manyamayes discuss social media comments and text analytics on @attensity blog — http://bit.ly/bmUTit
This comment was originally posted on Twitter
| Link | July 28th, 2010 at 4:11 pm
attensity wrote,
RT @themaria: .@manyamayes discuss social media comments and text analytics on @attensity blog — http://bit.ly/bmUTit
This comment was originally posted on Twitter
| Link | July 28th, 2010 at 4:48 pm
SethGrimes wrote,
RT @themaria: .@manyamayes discuss social media comments and text analytics on @attensity blog — http://bit.ly/bmUTit
This comment was originally posted on Twitter
| Link | July 28th, 2010 at 5:57 pm
ManyaMayes wrote,
New @attensity blog on social media comments – which are true and which are not? http://bit.ly/bmUTit #textmining #analytics #fraud
This comment was originally posted on Twitter
| Link | July 28th, 2010 at 6:48 pm
4westcoast wrote,
RT @attensity Social Media Comments – Fact Or Fiction? Find Out With Text Analytics. http://bit.ly/bmUTit
This comment was originally posted on Twitter
| Link | July 29th, 2010 at 3:49 am
PRwise wrote,
Social media comments: Fact, fiction or fraud? Find out with text analytics. | Attensity Blog — http://ht.ly/2iqX8
This comment was originally posted on Twitter
| Link | July 30th, 2010 at 2:25 pm
newswiseroger wrote,
Social media comments: Fact, fiction or fraud? Find out with text analytics. | Attensity Blog — http://ht.ly/2iqUF
This comment was originally posted on Twitter
| Link | July 30th, 2010 at 2:25 pm
jenniferdelano wrote,
RT @PRwise: Social media comments: Fact, fiction or fraud? Find out with text analytics. | Attensity Blog — http://ht.ly/2iqX8
This comment was originally posted on Twitter
| Link | July 30th, 2010 at 2:31 pm