When it's all piling up: investigating error propagation in an NLP pipeline

We present an analysis of a high-level semantic task, the construction of cross-document event timelines from SemEval 2015 Task 4: TimeLine, to trace down errors to the components of our pipeline system. Event timeline extraction requires many different Natural Language Processing tasks among which entity and event detection, coreference resolution and semantic-role-labeling are pivotal. These tasks yet depend on other low-level analysis. This paper shows where errors come from and whether they are propagated through the different layers. We also show that performance of each of the subtasks is still insufficient for the complex task considered. Finally, we observe that there is not enough semantics and inferencing within the standard NLP techniques to perform well.

• Type: Journal
• Year: 2015
• Where: NLDB
• Authors: Tommaso Caselli, Piek Vossen, Marieke van Erp, Antske Fokkens, Filip Ilievski, Rubén Izquierdo, Minh Le, Roser Morante, Marten Postma
• Keywords: Natural Language Processing, NLDB 2015