Framework for Weakly Supervised Causal Knowledge Extraction from Text

Anonymous

Abstract: In this paper, we address the problem of extracting causal knowledge from text documents in a weakly supervised manner. We target use cases in decision support and risk management, where causes and effects are general phrases without any constraints. We present a unified framework that supports three classes of tasks with varying degrees of available information. We provide approaches for each of the tasks using pre-trained, Natural Language Inference (NLI) and Question Answering (QA) models. We present a novel evaluation scheme and use existing and new benchmark data sets to measure the relative performance of each of the approaches.

Paper Type: long