Text representation using neural word embeddings has proven efficacy in many NLP applications. Recently, a lot of research interest goes beyond word embeddings by adapting the traditional word embedding models to learn vectors of multiword expressions (concepts/entities). However, current methods are limited to textual knowledge bases only (e.g., Wikipedia). In this paper, we propose a novel approach for learning concept vectors from two large scale knowledge bases (Wikipedia, and Probase). We adapt the skip-gram model to seamlessly learn from the knowledge in Wikipedia text and Probase concept graph. We evaluate our concept embedding models intrinsically on two tasks: 1) analogical reasoning where we achieve a state-of-the-art performance of 91% on semantic analogies, 2) concept categorization where we achieve a state-of-the-art performance on two benchmark datasets achieving categorization accuracy of 100% on one and 98% on the other. Additionally, we present a case study to extrinsically evaluate our model on unsupervised argument type identification for neural semantic parsing. We demonstrate the competitive accuracy of our unsupervised method and its ability to better generalize to out of vocabulary entity mentions compared to the tedious and error prone methods which depend on gazetteers and regular expressions.

本文提出了一种新颖而简单的技术，将来自不同结构的两个大型知识库中的概念知识（Wikipedia和Probase）整合起来，学习概念表示，并在两个任务上评估了概念嵌入模型：类比推理和概念分类，并展示了无监督环境中用于神经语义分析的参数识别的案例研究，相对于gazetteer和正则表达式等繁琐易出错的方法，我们的无监督方法具有更好的泛化能力。

超越词嵌入：从大规模知识库中学习实体和概念表示