TALN-UPF: Taxonomy Learning Exploiting CRF-Based Hypernym Extraction on Encyclopedic Definitions

Proceedings of the 9th International Workshop on Semantic Evaluation (SemEval 2015), pages 949–954, Denver, Colorado, June 4-5, 2015. c⃝2015 Association for Computational Linguistics TALN-UPF: Taxonomy Learning Exploiting CRF-Based Hypernym Extraction on Encyclopedic Definitions Luis Espinosa-Anke and Horacio Saggion and Francesco Ronzano Tractament Autom`atic del Llenguatge Natural (TALN) Department of Information and Communication Technologies Universitat Pompeu Fabra Carrer T`anger, 122-140 08018 Barcelona, Spain {luis.espinosa, horacio.saggion, francesco.ronzano}@upf.edu Abstract This paper describes the system submitted by the TALN-UPF team to SEMEVAL Task 17 (Taxonomy Extraction Evaluation). We present a method for automatically learning a taxonomy from a flat terminology, which benefits from a definition corpus obtained by querying the BabelNet semantic network. Then, we combine a machine-learning al- gorithm for term-hypernym extraction with linguistically-motivated heuristics for hyper- nym decomposition. Our approach performs well in terms of vertex coverage and newly added vertices, while it shows room for im- provement in terms of graph topology, edge coverage and precision of novel edges. 1 Introduction Learning semantic relations out of flat terminologies is an appealing task due to its potential application in tasks like Question Answering (Cui et al., 2005; Boella et al., 2014), automatic glossary construction (Muresan and Klavans, 2002), Ontology Learning (Navigli et al., 2011) or Textual Entailment (Roller et al., 2014). Today, in the context of massive web- enabled data, hypernym (is-a) relations are the focus of much research, as they constitute the backbone of ontologies (Navigli et al., 2011). However, one chal- lenge remains open in the automatic construction of knowledge bases that exploit this type of relation. It is unfeasible to have up-to-date semantic resources for each domain, as they are limited in scope and domain, and their manual construction is knowledge intensive and time consuming (Fu et al., 2014). Given this rationale, Task 17 (Bordea et al., 2015) in the SEMEVAL 2015 set of shared tasks focuses on Taxonomy Extraction Evaluation, i.e. the con- struction of a taxonomy out of a flat set of terms be- longing to one of the four domains of choice (food, chemical, equipment and science). These terms have to be hierarchically organized, and new terms are al- lowed to be included in the taxonomy. As for eval- uation, for each domain, two taxonomies were used as gold standard: One created by domain experts; and one derived from the WordNet taxonomy rooted at the domain node, e.g. food1. Finally, evaluation is carried out from two standpoints: (1) The taxon- omy topology and the rate of replicated nodes and edges are taken into account when compared to a gold standard taxonomy; and (2) Human experts val- idated as correct or incorrect a subset of the newly added edges. In this paper we describe our contribution to this shared task. Our approach relies on a set of definitional sentences for each term, from which term→hypernym relations are extracted using a machine-learning classifier. In a second step, linguistically-motivated rules are applied in order to (1) extract a hypernym candidate when the confi- dence of the classifier was below …

Content truncated to 500 words