As we continue to move away from simply analyzing facts and
figures, and more towards unstructured text analysis, the need for
sentimental analysis is more important than ever. Perfecting this
method will be the next great hurdle of text mining. With the
advent of social media, there is more text input from more people
than ever before, and this large mass of data is just waiting to be
analyzed through text mining methods like sentiment analysis. If
businesses wish to learn as much as they can about potential
customers, a sentiment analysis method can help determine people's
feelings about a particular subject, whether positive, negative, or
neutral. After the removal of stop words and the use of stemming,
the remaining words could be run through a model that's been
previously trained and determinations could be made about most
commonly used words and the opinions attached to those words.
I think for anyone who wants to attempt this kind of text mining
and analysis, the cost is going to be extensive, whether we're talking
about the machinery and storage required to handle that level of
computation, or the power required to run that equipment. The
ability to read the number of posts made in a single day, perform all
sorts of pre-processing (such as fixing typos), and then determine
whether the post has an opinion and if its a useful opinion will
already require sophisticated equipment, and will increase as time
goes on and social media continues to evolve. As with any data
analysis, the challenge comes from figuring out how to appropriately
teach a machine how to read the data and both learn from it and
analyze it. It will be interesting to see if sentiment analysis in it's
current form is still useful in the future, or if some other method has
replaced it.
