Multi-hop Question Generation with Graph Convolutional Network

Abstract

Multi-hop Question Generation (QG) aims to generate answer-related questions by aggregating and reasoning over multiple scattered evidence from different paragraphs. It is a more challenging yet under-explored task compared to conventional single-hop QG, where the questions are generated from the sentence containing the answer or nearby sentences in the same paragraph without complex reasoning. To address the additional challenges in multi-hop QG, we propose Multi-Hop Encoding Fusion Network for Question Generation (MulQG), which does context encoding in multiple hops with Graph Convolutional Network and encoding fusion via an Encoder Reasoning Gate. To the best of our knowledge, we are the first to tackle the challenge of multi-hop reasoning over paragraphs without any sentence-level information. Empirical results on HotpotQA dataset demonstrate the effectiveness of our method, in comparison with baselines on automatic evaluation metrics. Moreover, from the human evaluation, our proposed model is able to generate fluent questions with high completeness and outperforms the strongest baseline by 20.8% in the multihop evaluation. The code is publicly available at https://github.com/HLTCHKUST/MulQG.

Controllable Open-ended Question Generation with A New Question Type Ontology

“…In particular, question likelihoods and rewards are designed to steer them toward being addressed by the given answers (Zhou et al, 2019a;Zhang and Bansal, 2019). Attempts are also made toward creating complex questions that require multi-hop reasoning over the given text, and graph-based representations have been an enabling tool to facilitate the access to both entities and relations (Pan et al, 2020;Su et al, 2020). While our model also enhances the input with a semantic graph, it boasts a richer representation by including both dependency and semantic relations, with predicted question focuses highlighted via extra node embeddings.…”

“…Significant progress has been made in generating factoid questions (Zhang and Bansal, 2019;Zhou et al, 2019b;Su et al, 2020), yet new challenges need to be addressed for open-ended questions. First, specifying the question type is crucial for constructing meaningful questions (Graesser et al, 1992).…”

A Survey on Multi-hop Question Answering and Generation

“…3.1.3 Question generation: Generating multi-hop QA pairs requires accumulation of information across different contexts which itself is an unsolved problem [54,63,72,79,102,107,119,137,176]. As discussed above, Welbl et al [159] automatically generates questions using existing KBs where WikiData and DrugBank [163] are used as the knowledge bases, and Wikipedia and Medline 7 are used as the document corpus.…”

Read before Generate! Faithful Long Form Question Answering with Machine Reading

“…Grounded generation is the task of leveraging external knowledge sources to enhance the gen-eration. Previous work has either used structured external knowledge source (Liu et al, 2018;Young et al, 2018;Su et al, 2020a) or unstructured data. introduced a document grounded dataset for text conversations, and proposed to extract lexical control phrases to do controllable grounded response generation, while Zhang et al (2021) jointly trained a retriever and generator so that annotated text reference parallel data are not needed.…”