Automatic Extraction of Opinion Propositions and their Holders

Exploring Attitude and Affect in Text: Theories and Applications

Volume

Papers from the 2004 AAAI Spring Symposium

Contents

We identify a new task in the ongoing analysis of opinions: finding propositional opinions, sentential complements which for many verbs contain the actual opinion, rather than full opinion sentences. We propose an extension of semantic parsing techniques, coupled with additional lexical and syntactic features, that can produce labels for propositional opinions as opposed to other syntactic constituents. We describe the annotation of a small corpus of 5,139 sentences with propositional opinion information, and use this corpus to evaluate our methods. We also present results that indicate that the proposed methods can be extended to the related task of identifying opinion holders and associating them with propositional opinions.