The task of fine-grained opinion analysis deals with the
identification of opinions in text at the sentence-level. A goal of
such an analysis may for instance be to identify the actual opinion
expressed, the opinion holder and the opinion target (who or what the
opinion is about) for a given sentence.
For instance, for the sentence "Users hate the new Iphone", "Users" would be the opinion holders, "hate" would be the opinion expressed, and "the new Iphone" would be the target of the opinion. Since opinions can be expressed linguistically in a multitude of different
ways, this is considered a challenging task and there are a range of
different methods that have been proposed for solving various aspects
of this task. Several of these rely on syntactic analysis in some form
or other for identifying the various components of a sentence. Even
though syntactic parsing has been shown to be useful for
fine-grained opinion analysis, relatively little is known regarding
how the choice of specific syntactic formalism and/or parser influences performance.
This project has the aim of investigating the use of syntactic parsing
for the task of fine-grained, sentence-level opinion analysis and in
particular examining the effect of different syntactic formalisms.
Prerequisites for this project are some programming skills and an
interest in syntactic parsing. Details and further specification of
the project can be discussed with Lilja Øvrelid or Stephan Oepen.