We participated in Aspect Based Sentiment Analysis task at SemeEval 2015, obtaining best results in the Opinion Target Extraction (OTE) subtask.
We address the OTE task as a sequence labeling problem. We use the ixa-pipe-ote system off-the-shelf to train our OTE models; ixa-pipe-ote uses the Apache OpenNLP project implementation of the Perceptron algorithm customized with its own local and word representation features.
If you use the system, please the following paper: