Xart: Discovery of correlated arguments of n-ary relations in text

Soumia Lilia Berrahou (1, 2) , Patrice Buche (2, 3) , Juliette Dibie-Barthelemy (4) , Mathieu Roche (5, 6)

1 LIRMM - Laboratoire d'Informatique de Robotique et de Microélectronique de Montpellier
2 UMR IATE - Ingénierie des Agro-polymères et Technologies Émergentes
3 GRAPHIK - Graphs for Inferences on Knowledge
4 MIA-Paris - Mathématiques et Informatique Appliquées
5 ADVANSE - ADVanced Analytics for data SciencE
6 UMR TETIS - Territoires, Environnement, Télédétection et Information Spatiale

Here we present the Xart system based on a three-step hybrid method using data mining approaches and syntactic analysis to automatically discover and extract relevant data modeled as n-ary relations in plain text. A n-ary relation links a studied object with its features considered as several arguments. We addressed the challenge of designing a novel method to handle the identification and extraction of het- erogeneous arguments such as symbolic arguments, quantitative arguments composed of numbers and various measurement units. We thus developed the Xart system, which relies on a domain ontology for discovering patterns, in plain text, to identify arguments involved in n-ary relations. The discovered pat- terns take advantage of different ontological levels that facilitate identification of all arguments and pool them in the sought n-ary relation.

Keywords: Ontology; Linguistic pattern; Information extraction IE; Data mining DM; Information extraction; N-ary relation; Data mining; Sequential pattern; Quantitative data

Domains: Computer Science [cs]

Soumia Lilia Berrahou, Patrice Buche, Juliette Dibie-Barthelemy, Mathieu Roche. Xart: Discovery of correlated arguments of n-ary relations in text. Expert Systems with Applications, 2017, 73, pp.115-124. ⟨10.1016/j.eswa.2016.12.028⟩. ⟨hal-01508801⟩