FBK_NK: A WordNet-Based System for Multi-Way Classification of Semantic Relations

Proceedings of the 5th International Workshop on Semantic Evaluation, ACL 2010, pages 202–205, Uppsala, Sweden, 15-16 July 2010. c⃝2010 Association for Computational Linguistics FBK NK: a WordNet-based System for Multi-Way Classification of Semantic Relations Matteo Negri and Milen Kouylekov FBK-Irst Trento, Italy {negri,kouylekov}@fbk.eu Abstract We describe a WordNet-based system for the extraction of semantic relations be- tween pairs of nominals appearing in English texts. The system adopts a lightweight approach, based on training a Bayesian Network classifier using large sets of binary features. Our features con- sider: i) the context surrounding the an- notated nominals, and ii) different types of knowledge extracted from WordNet, in- cluding direct and explicit relations be- tween the annotated nominals, and more general and implicit evidence (e.g. seman- tic boundary collocations). The system achieved a Macro-averaged F1 of 68.02% on the “Multi-Way Classification of Se- mantic Relations Between Pairs of Nom- inals” task (Task #8) at SemEval-2010. 1 Introduction The “Multi-Way Classification of Semantic Re- lations Between Pairs of Nominals” task at SemEval-2010 (Hendrickx et al., 2010) consists in: i) selecting from an inventory of nine possi- ble relations the one that most likely holds be- tween two annotated nominals appearing in the in- put sentence, and ii) specifying the order of the nominals as the arguments of the relation. In con- trast with the semantic relations classification task (Task #4) at SemEval-2007 (Girju et al., 2007), which treated each semantic relation separately as a single two-class (positive vs. negative) classifi- cation task, this year’s edition of the challenge pre- sented participating systems with a more difficult and realistic multi-way setup, where the relation Other can also be assigned if none of the nine re- lations is suitable for a given sentence. Examples of the possible markable relations are reported in Table 11. The objective of our experiments with the pro- posed task is to develop a Relation Extraction sys- tem based on shallow linguistic processing, taking the most from available thesauri and ontologies. As a first step in this direction, our submitted runs have been obtained by processing the input sen- tences only to lemmatize their terms, and by using WordNet as the sole source of knowledge. Similar to other approaches (Moldovan and Badulescu, 2009; Beamer et al., 2009), our sys- tem makes use of semantic boundaries extracted from the WordNet IS-A backbone. Such bound- aries (i.e. divisions in the WordNet hierarchy that best generalize over the training examples) are used to define pairs of high-level synsets with high correlation with specific relations. For in- stance, <microorganism#1, happening#1> and <writing#1, consequence#1> are extracted from the training data as valid high-level collocations respectively for the relations Cause-Effect and Message-Topic. Besides exploiting the Word- Net IS-A hierarchy, the system also uses the holo-/meronymy relations, and information de- rived from the WordNet glosses to capture specific relations such as Member-Collection and Product- Producer. In addition, the context surrounding the annotated nominals is represented as a bag-of- words/synonyms to enhance the relation extraction process. Several experiments have been …

Content truncated to 500 words