A deep-learning approach to mining conditions

A condition is a constraint that determines when a consequent holds. Mining them in text is paramount to understand many sentences properly. In the literature, there are a few pattern-based proposals that fall short regarding recall because it is not easy to characterise unusual ways to express cond...

Descripción completa

Detalles Bibliográficos
Autores: Gallego, Fernando O., Corchuelo Gil, Rafael
Tipo de recurso: artículo
Estado:Versión publicada
Fecha de publicación:2020
País:España
Institución:Universidad de Sevilla (US)
Repositorio:idUS. Depósito de Investigación de la Universidad de Sevilla
OAI Identifier:oai:idus.us.es:11441/143218
Acceso en línea:https://hdl.handle.net/11441/143218
https://doi.org/10.1016/j.knosys.2019.105422
Access Level:acceso abierto
Palabra clave:Natural language processing
Text mining
Condition mining
Neural networks
Descripción
Sumario:A condition is a constraint that determines when a consequent holds. Mining them in text is paramount to understand many sentences properly. In the literature, there are a few pattern-based proposals that fall short regarding recall because it is not easy to characterise unusual ways to express conditions with hand-crafted patterns; there is one machine-learning proposal that is bound to the Japanese language, requires specific-purpose dictionaries, taxonomies, and heuristics, works on opinion sentences only, and was evaluated very shallowly. In this article, we present a deep-learning proposal to mine conditions that does not have any of the previous drawbacks; furthermore, we have performed a comprehensive experimental study on a large multi-lingual dataset on many common topics; our conclusion is that our proposals are similar to the state of the art in terms of precision, but improve recall enough to beat them in terms of F1 score.