First Steps in Recognizing Relational Entailment – Experimental Corpus and Baselines

Warning

This publication doesn't include Faculty of Economics and Administration. It includes Faculty of Informatics. Official publication website can be found on muni.cz.
Authors

VÍTA Martin KLÍMEK Jakub

Year of publication 2019
Type Article in Proceedings
Conference Human Language Technologies as a Challenge for Computer Science and Linguistics – 2019
MU Faculty or unit

Faculty of Informatics

Citation
Keywords NLI; relational entailment; textual entailment; annotated corpus
Description Recently, a task of recognizing relational entailment (RRE) was introduced as a task to decide whether the meaning of a given textual $n$-tuple $t$, i. e., the semantic relationship expressed by $t$, can be inferred from a given text $T$. Since then-tuples are obtained from theopen information extraction process, this task naturally connects two NLP fields: natural language inference (NLI) and open information extraction (open IE). The task has a “practical” counterpart: checking/proving facts stored in open knowledge bases. However, no corresponding annotated corpus has been available yet as well as baselines for this task. In this paper, we present a corpus derived fromthe well known SNLI corpus and provide baselines based on LSTM architectures and state also a baseline using “hypothesis-only”-like approach.
Related projects:

You are running an old browser version. We recommend updating your browser to its latest version.