Most social media analyses such as sentiment analysis
for microblogs are often built as standalone, endpoint to
endpoint applications. This makes the collaboration among distributed
software and data service providers to create composite
social analytic solutions difficult. This paper first proposes a
system of systems service architecture (SoS-SA) design for social
media analytics that support and facilitate efficient collaboration
among distributed service providers. Then we propose a novel
Twitters sentiment analysis service implemented on top of this
design to illustrate its potentials. Current sentiment classification
applications based on supervised learning methods relies too
heavily on the chosen large training datasets; approaches using
automatically generated training datasets also often result in the
huge imbalance between the subjective classes and the objective
classes in the sentiment of tweets, making it difficult to obtain
good recall performance for the subjective ones. To address
this issue, our proposed solution is based on a semi-supervised
learning method for tweet sentiment classification. Experiments
show that the performance of our method is better than those of
the previous work.