Filtering and Fusion of Question-Answering Streams by Robust Textual Inference Ingo Glöckner FernUniversität in Hagen , 58084 Hagen , Germany iglockner

The paper presents an answer validation system based on robust textual inference. The system is evaluated on the AVE06 test set for German. Ablation experiments are used to determine the contribution of functional components and knowledge sources to the achieved validation quality. We demonstrate the utility of the approach for filtering answers and for the fusion of several answer sources into a single most trusted answer. Topics: Answer Validation, Robust Textual Inference, Information Fusion/Integration, Evaluation.