Semi-automated Methods for the Annotation and Design of a Semantic Network Designed for Sentiment Analysis of Social Web Content

We have designed a new semantic network for the English language annotated with two types of domain knowledge: 1) general domain knowledge, represented as a sentiment polarity attached to relationships between words and 2) contextual knowledge, such as domain-specific proper names. In the context of social networks, such as Twitter, the evolution and maintenance of the network are particularly critical tasks for analysts not to incur in a quick obsolescence. We show how a set of semi-automated methodologies applied to our new semantic network show promising results on preliminary tests run using Twitter data.