No abstract

In this paper, we conduct an empirical investigation of neural query graph ranking approaches for the task of complex question answering over knowledge graphs. We experiment with six different ranking models and propose a novel self-attention based slot matching model which exploits the inherent structure of query graphs, our logical form of choice. Our proposed model generally outperforms the other models on two QA datasets over the DBpedia knowledge graph, evaluated in different settings. In addition, we show that transfer learning from the larger of those QA datasets to the smaller dataset yields substantial improvements, effectively offsetting the general lack of training data.

Question answering has emerged as an intuitive way of querying structured data sources and has attracted significant advancements over the years. A large body of recent work on question answering over knowledge graphs (KGQA) employs neural network‐based systems. In this article, we provide an overview of these neural network‐based methods for KGQA. We introduce readers to the formalism and the challenges of the task, different paradigms and approaches, discuss notable advancements, and outline the emerging trends in the field. Through this article, we aim to provide newcomers to the field with a suitable entry point to semantic parsing for KGQA, and ease their process of making informed decisions while creating their own QA systems. This article is categorized under: Technologies > Machine Learning Technologies > Prediction Technologies > Artificial Intelligence

Answering simple questions over knowledge graphs is a well-studied problem in question answering. Previous approaches for this task built on recurrent and convolutional neural network based architectures that use pretrained word embeddings. It was recently shown that finetuning pretrained transformer networks (e.g. BERT) can outperform previous approaches on various natural language processing tasks. In this work, we investigate how well BERT performs on SIMPLEQUESTIONS and provide an evaluation of both BERT and BiLSTMbased models in limited-data scenarios.

Knowledge graphs are composed of different elements: entity nodes, relation edges, and literal nodes. Each literal node contains an entity's attribute value (e.g. the height of an entity of type person) and thereby encodes information which in general cannot be represented by relations between entities alone. However, most of the existing embedding-or latent-feature-based methods for knowledge graph analysis only consider entity nodes and relation edges, and thus do not take the information provided by literals into account. In this paper, we extend existing latent feature methods for link prediction by a simple portable module for incorporating literals, which we name Lit-eralE. Unlike in concurrent methods where literals are incorporated by adding a literal-dependent term to the output of the scoring function and thus only indirectly affect the entity embeddings, LiteralE directly enriches these embeddings with information from literals via a learnable parametrized function. This function can be easily integrated into the scoring function of existing methods and learned along with the entity embeddings in an end-to-end manner. In an extensive empirical study over three datasets, we evaluate LiteralE-extended versions of various state-of-the-art latent feature methods for link prediction and demonstrate that LiteralE presents an effective way to improve their performance. For these experiments, we augmented standard datasets with their literals, which we publicly provide as testbeds for further research. Moreover, we show that LiteralE leads to an qualitative improvement of the embeddings and that it can be easily extended to handle literals from different modalities.