Distributed Prediction of Relations for Entities: The Easy, The Difficult, and The Impossible

Proceedings of the 6th Joint Conference on Lexical and Computational Semantics (*SEM 2017), pages 104–109, Vancouver, Canada, August 3-4, 2017. c⃝2017 Association for Computational Linguistics Distributed Prediction of Relations for Entities: The Easy, The Difficult, and The Impossible Abhijeet Gupta* and Gemma Boleda† and Sebastian Pad´o* *Stuttgart University, Germany {abhijeet.gupta,pado}@ims.uni-stuttgart.de †Universitat Pompeu Fabra, Barcelona, Spain gemma.boleda@upf.edu Abstract Word embeddings are supposed to provide easy access to semantic relations such as “male of” (man–woman). While this claim has been investigated for concepts, little is known about the distributional behavior of relations of (Named) Entities. We de- scribe two word embedding-based models that predict values for relational attributes of entities, and analyse them. The task is challenging, with major performance dif- ferences between relations. Contrary to many NLP tasks, high difficulty for a re- lation does not result from low frequency, but from (a) one-to-many mappings; and (b) lack of context patterns expressing the relation that are easy to pick up by word embeddings. 1 Introduction A central claim about distributed models of word meaning (e.g., Mikolov et al. (2013)) is that word embedding space provides easy access to semantic relations. E.g., Mikolov et al.’s space was shown to encode the “male-female relation” linearly, as a vector (# » man − # » woman = # » king − # » queen). The accessibility of semantic relations was sub- sequently examined in more detail. Rei and Briscoe (2014) and Melamud et al. (2014) reported success- ful modeling of lexical relations such as hyper- nymy and synonymy. K¨oper et al. (2015) consid- ered a broader range of relationships,with mixed results. Levy and Goldberg (2014b) developed an improved, nonlinear relation extraction method. These studies were conducted primarily on concepts and their semantic relations, like hypernym(politician) = person. Mean- while, entities and the relations they partake in are much less well understood.1 Entities are instances of concepts, i.e., they refer to specific individual ob- jects in the real world, for example, Donald Trump is an instance of the concept politician. Conse- quently, entities are generally associated with a rich set of numeric and relational attributes (for politician instances: size, office, etc.). In contrast to concepts, the values of these attributes tend to be discrete (Herbelot, 2015): while the size of politician is best described by a prob- ability distribution, the size of Donald Trump is 1.88m. Since distributional representations are notoriously bad at handling discrete knowl- edge (Fodor and Lepore, 1999; Smolensky, 1990), this raises the question of how well such models can capture entity-related knowledge. In our previous work (Gupta et al., 2015), we analysed distributional prediction of numeric at- tributes of entities, found a large variance in quality among attributes, and identified factors determin- ing prediction difficulty. A corresponding analysis for relational (categorial) attributes of entities is still missing, even though entities are highly rele- vant for NLP. This is evident from the highly ac- tive area of knowledge base completion (KBC), the task of extending incomplete entity information in knowledge bases such …

Content truncated to 500 words