Knowledge-enriched, Type-constrained and Grammar-guided Question Generation over Knowledge Bases

TL;DR

This paper introduces KTG, a novel question generation model over knowledge bases that enhances diversity, fluency, and semantic accuracy by integrating auxiliary KB information, type constraints, and grammar guidance, significantly outperforming existing methods.

Contribution

The paper presents KTG, a knowledge-enriched, type-constrained, and grammar-guided model that effectively addresses low diversity, semantic drift, and fluency issues in KB question generation.

Findings

Outperforms existing methods on benchmark datasets

Improves question diversity and syntactic correctness

Enhances semantic relevance of generated questions

Abstract

Question generation over knowledge bases (KBQG) aims at generating natural-language questions about a subgraph, i.e. a set of (connected) triples. Two main challenges still face the current crop of encoder-decoder-based methods, especially on small subgraphs: (1) low diversity and poor fluency due to the limited information contained in the subgraphs, and (2) semantic drift due to the decoder's oblivion of the semantics of the answer entity. We propose an innovative knowledge-enriched, type-constrained and grammar-guided KBQG model, named KTG, to addresses the above challenges. In our model, the encoder is equipped with auxiliary information from the KB, and the decoder is constrained with word types during QG. Specifically, entity domain and description, as well as relation hierarchy information are considered to construct question contexts, while a conditional copy mechanism is incorporated to modulate question semantics according to current word types. Besides, a novel reward function featuring grammatical similarity is designed to improve both generative richness and syntactic correctness via reinforcement learning. Extensive experiments show that our proposed model outperforms existing methods by a significant margin on two widely-used benchmark datasets SimpleQuestion and PathQuestion.