Abstract
Knowledge graphs such as DBpedia, Freebase or Wikidata always contain a taxonomic backbone that allows the arrangement and structuring of various concepts in accordance with hypo-hypernym (“class-subclass”) relationship. With the rapid growth of lexical resources for specific domains, the problem of automatic extension of the existing knowledge bases with new words is becoming more and more widespread. In this paper, we address the problem of taxonomy enrichment which aims at adding new words to the existing taxonomy. We present a new method which allows achieving high results on this task with little effort. It uses the resources which exist for the majority of languages, making the method universal. We extend our method by incorporating deep representations of graph structures like node2vec, Poincaré embeddings, GCN etc. that have recently demonstrated promising results on various NLP tasks. Furthermore, combining these representations with word embeddings allows us to beat the state of the art. We conduct a comprehensive study of the existing approaches to taxonomy enrichment based on word and graph vector representations and their fusion approaches. We also explore the ways of using deep learning architectures to extend taxonomic backbones of knowledge graphs. We create a number of datasets for taxonomy extension for English and Russian. We achieve state-of-the-art results across different datasets and provide an in-depth error analysis of mistakes.
Highlights
The central idea of Semantic Web is to make the content of the Internet pages machine-interpretable
This work is an extended version of the work described in [57,58,77]. The novelty of this particular article as compared to the previous publications is as follows: 1. We present a new taxonomy enrichment method Distributional Wiktionary-based synset Ranking (DWRank) which combines distributional information and the information extracted from Wiktionary
As for the taxonomy enrichment task, we are only aware of a recent approach T TaxoExpan [71] which applies position-enhanced graph neural networks (GCN [35] and GAT [82]) that we evaluate on our datasets
Summary
The central idea of Semantic Web is to make the content of the Internet pages machine-interpretable. In order to speed up and simplify this task, it becomes more and more important to develop systems that could automatically enrich the existing knowledge bases with new words or at least facilitate the manual extension process. T The state-of-the-art taxonomy enrichment methods have two main drawbacks First of all, they often use unrealistic formulations of the task. Taxonomy is a special case of a knowledge graph G It is a tree-based structure where nodes from the set E are connected with the hypernymy relation r. Taxonomy enrichment can be considered as a special case of the knowledge base completion task. This task aimed at associating each new word q ∈ Q, which is not yet included in the taxonomy T , with the appropriate hypernyms from it
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
