A Survey on Complex Knowledge Base Question Answering: Methods, Challenges and Solutions

Sep 19, 2026

Knowledge base question answering (KBQA) targets natural-language questions grounded in a knowledge base (KB). Research increasingly tackles semantically or syntactically complex questions that involve multi-hop reasoning, constrained relations, and numerical operations. This survey summarizes typical challenges and corresponding solutions for complex KBQA. It reviews two dominant method families—semantic parsing-based and information retrieval-based—then discusses advanced approaches through these lenses and concludes with promising future research directions.

LanguageEnglish

CategoryPresentations

FormatPDF

Pages9

Reading Time3 min

AccessFree to download

Introduction

Background: KB and KBQA

Complex KBQA and its characteristics

Main method categories: SP-based vs IR-based

Key challenges in complex KBQA

Proceedings of the Thirtieth International Joint Conference on Arti􀀂cial Intelligence (IJCAI-21) Survey Track A Survey on Complex Knowledge Base Question Answering: Methods, Challenges and Solutions Yunshi Lan 1 􀀃 , Gaole He2 ;3 􀀃 , Jinhao Jiang4 , Jing Jiang 1 , Wayne Xin Zhao3 ;4y and Ji-Rong Wen2 ;3 ;4 1 School of Computing and Information Systems, Singapore Management University 2 School of Information, Renmin University of China 3Beijing Key Laboratory of Big Data Management and Analysis Methods 4 Gaoling School of Arti􀀂cial Intelligence, Renmin University of China fyslan, [jingjiang](jingjiangg@smu.edu.sg)[g](jingjiangg@smu.edu.sg)[@smu.edu.sg](jingjiangg@smu.edu.sg), fhegaole, [jrwen](jrweng@ruc.edu.cn)[g](jrweng@ruc.edu.cn)[@ruc.edu.cn](jrweng@ruc.edu.cn), fbatman􀀃y, jiangjinhaonlpg@gmail.com Abstract Knowledge base question answering (KBQA) aims to answer a question over a knowledge base (KB) . Recently, a large number of studies focus on semantically or syntactically complicated questions. In this paper, we elaborately summarize the typical challenges and solutions for complex KBQA. We begin with introducing the background about the KBQA task. Next, we present the two mainstream categories of methods for complex KBQA, namely semantic parsing-based (SP-based) methods and information retrieval-based (IR-based) methods. We then review the advanced methods comprehensively from the perspective of the two categories. Speci􀀂cally, we explicate their solutions to the typical challenges. Finally, we conclude and discuss some promising directions for future research. 1 Introduction A knowledge base (KB) is a structured database that contains a collection of facts in the form (subject, relation, object) . Large-scale KBs, such as Freebase [Bollacker et al., 2008], DBPedia [Lehmann et al., 2015] and Wikidata [Tanonet al., 2016], have been constructed to serve many downstream tasks. Based on available KBs, knowledge base question answering (KBQA) is a task that aims to answer natural language questions with KBs as its knowledge source. Early work on KBQA [Bordes et al., 2015; Dong et al., 2015; Hu et al., 2018a; Lan et al., 2019a] focuses on answering a simple question, where only a single fact is involved. For example, 􀂓Where was JK Rowling born?􀂔 is a simple question which can be answered using just the fact 􀂓(J.K. Rowling, birthplace, United Kingdom) 􀂔 . Recently, researchers start paying more attention to answering complex questions over KBs, i.e., the complex KBQA task [Hu et al., 2018b; Luo et al., 2018] . Complex questions usually contain multiple subjects, express compound relations and include numerical operations. Take the 􀀃 Equal contribution. y Corresponding author. Figure 1: An example of complex KBQA for the question 􀂓Whois the 􀀂rst wife of TV producer that was nominated for The Jeff Probst Show?􀂔 . We present the related KB subgraph for this question. The ground truth path to answer this question is annotated with colored borders. The topic entity and the answer entity are shown in the bold font and shaded box respectively. 􀂓multi-hop􀂔 reasoning,􀂓constrained􀂔 relations and 􀂓numerical􀂔 operation are highlighted in black dotted box. We use different colors to indicate different reasoning hops to reach each entity from the topic entity. question in Figure 1 as an example. This example question starts with the subject 􀂓The Jeff Probst Show􀂔 . Instead of querying a single fact, the question requires the composition of two relations, namely, 􀂓nominee􀂔 and 􀂓spouse􀂔 . This query is also associated with an entity type constraint 􀂓 (Jeff Probst, is a, TV producer) 􀂔 . The 􀀂nal answer should be further aggregated by selecting the possible candidates with the earliest marriage date. Generally, complex questions are questions involving multi-hop reasoning, constrained relations, numerical operations, or some combination of the above. Tracing back to the solutions for simple KBQA, a numb

FAQs

What is complex knowledge base question answering (KBQA)?

Complex KBQA asks natural-language questions over a KB that require composing multiple relations and may include multi-hop reasoning, constrained relations, and numerical operations.

What are the two mainstream categories of methods for complex KBQA?

The survey organizes methods into semantic parsing-based (SP-based) and information retrieval-based (IR-based) approaches, each with different mechanisms for deriving answers.

What challenges do existing SP-based and IR-based methods face on complex questions?

SP parsers may not cover diverse complex query structures, while IR methods can fail because ranking may rely on limited-scope entities without traceable reasoning.

Sep 19, 2026

Uploaded by Felix Montgomery