Generating Knowledge Graph Paths from Textual Definitions using Sequence-to-Sequence Models

Proceedings of NAACL-HLT 2019, pages 1968–1976 Minneapolis, Minnesota, June 2 - June 7, 2019. c⃝2019 Association for Computational Linguistics 1968 Generating Knowledge Graph Paths from Textual Definitions using Sequence-to-Sequence Models Victor Prokhorov,1 Mohammad Taher Pilehvar1,2 and Nigel Collier1 1Department of Theoretical and Applied Linguistics, University of Cambridge 2School of Computer Engineering, Iran University of Science and Technology, Tehran, Iran vp361@cam.ac.uk, pilehvar@iust.ac.ir, nhc30@cam.ac.uk Abstract We present a novel method for mapping un- restricted text to knowledge graph entities by framing the task as a sequence-to-sequence problem. Specifically, given the encoded state of an input text, our decoder directly predicts paths in the knowledge graph, starting from the root and ending at the target node fol- lowing hypernym-hyponym relationships. In this way, and in contrast to other text-to-entity mapping systems, our model outputs hierar- chically structured predictions that are fully in- terpretable in the context of the underlying on- tology, in an end-to-end manner. We present a proof-of-concept experiment with encourag- ing results, comparable to those of state-of- the-art systems. 1 Introduction Text-to-entity mapping is the task of associating a text with a concept in a knowledge graph (KG) or an ontology (we use two terms, interchange- ably). Recent works (Kartsaklis et al., 2018; Hill et al., 2015) use neural networks to project a text to a vector space where the entities of a KG are represented as continuous vectors. Despite being successful, these models have two main disadvan- tages. First, they rely on a predefined vector space which is used as a gold standard representation for the entities in a KG. Therefore, the quality of these algorithms depends on how well the vector space is represented. Second, these algorithms are not interpretable; hence, it is impossible to understand why a certain text was linked to a particular entity. To address these issues we propose a novel tech- nique which first represents an ontology concept as a sequence of its ancestors in the ontology (hy- pernyms) and then maps the corresponding textual description to this unique representation. For ex- ample, given the textual description of the concept swift (“small bird that resembles a swallow and is noted for its rapid flight”), we map it to the hier- archical sequence of entities in a lexical ontology: animal → chordate → vertebrate → bird → apod- iform bird. This sequence of nodes constitutes a path.1 Our model is based on a sequence-to-sequence neural network (Sutskever et al., 2014) coupled with an attention mechanism (Bahdanau et al., 2014). Specifically, we use an LSTM (Hochre- iter and Schmidhuber, 1997) encoder to project the textual description into a vector space and an LSTM decoder to predict the sequence of entities that are relevant to this definition. With this frame- work we do not need to rely on the pre-existing vector space of the entities, since the decoder ex- plicitly learns topological dependencies between the entities of the ontology. Furthermore, the pro- posed model is more interpretable for two rea- sons. First, instead of the closest points in …

Content truncated to 500 words