Bi-directional Inter-dependencies of Subjective Expressions and Targets and their Value for a Joint Model
Roman Klinger and Philipp Cimiano
The 51st Annual Meeting of the Association for Computational Linguistics - Short Papers (ACL Short Papers 2013)
Sofia, Bulgaria, August 4-9, 2013
Opinion mining is often regarded as a classification or segmentation task, involving the prediction of: i) subjective expressions, ii) their target and iii) the polarity. Intu- itively, these three variables are bidirectionally interdependent, but most work has ei- ther attempted to predict them in isolation or proposing pipeline-based approaches that can not model the bidirectional interaction between these variables. Towards better understanding the interaction between these variables, we propose a model in this paper that allows us for studying the interactions between predicting targets and subjectivity terms. In particular, we analyze the relation of target and subjective phrases in both directions and show how this can be modeled when one of them is not known in advance. This allows for analysis of error propagation issues and aims at estimating an upper bound for the im- pact of a joint model in comparison to a pipeline model. We report results on two public datasets (cameras and cars), showing that our model outperforms state-of-the-art models, as well as on a new dataset consisting of Twitter posts.
Conference Manager (V2.61.0 - Rev. 2792M)