Better Call the Plumber: Orchestrating Dynamic Information Extraction Pipelines

In the last decade, a large number of Knowledge Graph (KG) information\nextraction approaches were proposed. Albeit effective, these efforts are\ndisjoint, and their collective strengths and weaknesses in effective KG\ninformation extraction (IE) have not been studied in the literature. We propose\nPlumber, the first framework that brings together the research community's\ndisjoint IE efforts. The Plumber architecture comprises 33 reusable components\nfor various KG information extraction subtasks, such as coreference resolution,\nentity linking, and relation extraction. Using these components,Plumber\ndynamically generates suitable information extraction pipelines and offers\noverall 264 distinct pipelines.We study the optimization problem of choosing\nsuitable pipelines based on input sentences. To do so, we train a\ntransformer-based classification model that extracts contextual embeddings from\nthe input and finds an appropriate pipeline. We study the efficacy of Plumber\nfor extracting the KG triples using standard datasets over two KGs: DBpedia,\nand Open Research Knowledge Graph (ORKG). Our results demonstrate the\neffectiveness of Plumber in dynamically generating KG information extraction\npipelines,outperforming all baselines agnostics of the underlying KG.\nFurthermore,we provide an analysis of collective failure cases, study the\nsimilarities and synergies among integrated components, and discuss their\nlimitations.\n