Database Reference
In-Depth Information
to discover and build your models and to gain business insight. Then you have
the option of continuing to use at-rest analytics to harvest call interactions at
much lower latency, or to build up these models and then promote them
back to the frontier of the business, using Streams to examine and analyze
calls as quickly as they can be converted, so that insights can be acquired in
near-real time. This transforms the business from forecasting (we think cus-
tomers will leave if…) to nowcasting (we know this customer is about to leave
because…). The results of Streams analytics flow back into BigInsights, creat-
ing a closed-loop feedback mechanism, because BigInsights can then iterate
over the results to improve the models.
Social Media Techniques
Make the World Your Oyster
In the previous section we talked about social media analytics drawing a lot
of chatter and fatigue—it's potentially too hyped up. We thought we'd expand
your scope of where this fits in; indeed, if you have the technology to draw
structure and understanding from unstructured text, you can imagine an
endless number of usage patterns.
The same text analytics technology used for social media can be applied to
do a number of different things. For example, one client wanted to investi-
gate the dissemination of piracy and copyright infringement of their intel-
lectual property (it was video-based). It's a very interesting area because in
social media there are often discussions about types of piracy, and this client
was able to build patterns that identify these kinds of conversations that take
place in various micro-blog sites around the world. This ultimately led them
to web sites that contain illegally posted copyrighted materials that belonged
to our client. How did they do it? They developed dictionaries and patterns
that identified the names of sports teams involved in the “stolen” materials,
“pull” language that surrounds enticing links for free content, URL resolu-
tion from shortened URLS (tinyURLs) to actual sites, and so on. So while
they weren't analyzing who said what, they did create structure around a
problem domain from unstructured text-based data. Same technology, same
toolset, same approach, just a different use case.
 
Search WWH ::




Custom Search