Multiagent Approach to Coreference Resolution Based on the Multifactor Similarity in Ontology Population
- Authors: Garanina N.O.1, Sidorova E.A.1, Seryi A.S.1
-
Affiliations:
- Ershov Institute of Information Systems, Siberian Branch
- Issue: Vol 44, No 1 (2018)
- Pages: 23-34
- Section: Article
- URL: https://journals.rcsi.science/0361-7688/article/view/176570
- DOI: https://doi.org/10.1134/S0361768818010036
- ID: 176570
Cite item
Abstract
The multiagent approach to coreference resolution in the process of extracting information from texts in natural languages for ontology population is described. Special class agents corresponding to ontology classes are defined. They analyze the available information in the corresponding ontology instances. The results of this analysis are used for values to instance attribute, for detecting duplicates and equivalents of instances, for fixing coreferential relations, and for determining the weights of information connections used to resolve ambiguities. Coreferences are resolved taking into account a multifactor similarity measure between extracted objects that combines the semantic, context, positional, and grammar similarity measures. The class agents work within the multiagent approach to text analysis aimed at ontology population.
About the authors
N. O. Garanina
Ershov Institute of Information Systems, Siberian Branch
Author for correspondence.
Email: garanina@iis.nsk.su
Russian Federation, Novosibirsk, 630090
E. A. Sidorova
Ershov Institute of Information Systems, Siberian Branch
Email: garanina@iis.nsk.su
Russian Federation, Novosibirsk, 630090
A. S. Seryi
Ershov Institute of Information Systems, Siberian Branch
Email: garanina@iis.nsk.su
Russian Federation, Novosibirsk, 630090