TY - JOUR AU - AB - nd In this paper we describe our 2 place FEVER shared-task system that achieved a FEVER score of 62.52% on the provisional " " $ %$&’ ($ test set (without additional human evaluation), " " )&$*) " # " $ %$&’ ($ and 65.41% on the development set. Our sys- " " )&$*) tem is a four stage model consisting of docu- ment retrieval, sentence retrieval, natural lan- " " " " guage inference and aggregation. Retrieval " # " is performed leveraging task-specific features, " " and then a natural language inference model takes each of the retrieved sentences paired with the claimed fact. The resulting predic- tions are aggregated across retrieved sentences with a Multi-Layer Perceptron, and re-ranked corresponding to the final prediction. 1 Introduction We often hear the word “Fake News” these days. Figure 1: Illustration of the model pipeline for a claim. Recently, Russian meddling, for example, has been blamed for the prevalence of inaccurate news a complete set of relevant evidence sentences has stories on social media, but even TI - UCL Machine Reading Group: Four Factor Framework For Fact Finding (HexaF) JF - Proceedings of the First Workshop on Fact Extraction and VERification (FEVER) DO - 10.18653/v1/w18-5515 DA - 2018-01-01 UR - https://www.deepdyve.com/lp/unpaywall/ucl-machine-reading-group-four-factor-framework-for-fact-finding-hexaf-yur4haGzb5 DP - DeepDyve ER -