Abstract
Lexicons are a crucial aspect of any method that aims analyze textual sentiment. In this paper we aim to provide an algorithm in which we can obtain meaningful polarity values for words using a graph constructed from their synonyms. We show that this polarity graph can provide polarity values within a lexicon where needed (i.e. where another method such as a classifier left gaps). We then assess the value of this algorithm against other recent methods of sentiment lexicon expansion. We conclude that this method supplies significant, yet highly nuanced utility in terms of supplying necessary semantic orientation for words without.