Baseline System

To lower the obstacles for participation, we provide a baseline system that may or may not be used as a starting point for the participating system.

An executable and all source can be found here.


System

The baseline system contains two classifiers. A SVM is used for relevance classification, sentiment analysis and aspect categorization. For opinion target identification, it uses a CRF classifier.

The SVM uses the following features:

  • term frequency
  • German sentiment lexica

The CRF classifier uses these features:

  • the token (without standardization/lemmatization/lowercasing)
  • the POS tag

Both features are unigram features on the current token, no preceding/following tokens are taken into account.