Learning Adaptable Patterns for Passage Reranking

摘要

This paper proposes passage reranking models that (ⅰ) do not require manual feature engineering and (ⅱ) greatly preserve accuracy, when changing application domain. Their main characteristic is the use of relational semantic structures representing questions and their answer passages. The relations are established using information from automatic classifiers, i.e., question category (QC) and focus classifiers (FC) and Named Entity Recognizers (NER). This way (ⅰ) effective structural relational patterns can be automatically learned with kernel machines; and (ⅱ) structures are more invariant w.r.t. different domains, thus fostering adaptability.