This website requires JavaScript.
DOI: 10.1561/1900000078

Natural Language Interfaces to Data

Abdul QuamarVasilis EfthymiouChuan LeiFatma \"Ozcan
Dec 2022
Recent advances in NLU and NLP have resulted in renewed interest in naturallanguage interfaces to data, which provide an easy mechanism for non-technicalusers to access and query the data. While early systems evolved from keywordsearch and focused on simple factual queries, the complexity of both the inputsentences as well as the generated SQL queries has evolved over time. Morerecently, there has also been a lot of focus on using conversational interfacesfor data analytics, empowering a line of non-technical users with quickinsights into the data. There are three main challenges in natural languagequerying (NLQ): (1) identifying the entities involved in the user utterance,(2) connecting the different entities in a meaningful way over the underlyingdata source to interpret user intents, and (3) generating a structured query inthe form of SQL or SPARQL. There are two main approaches for interpreting a user's NLQ. Rule-basedsystems make use of semantic indices, ontologies, and KGs to identify theentities in the query, understand the intended relationships between thoseentities, and utilize grammars to generate the target queries. With theadvances in deep learning (DL)-based language models, there have been manytext-to-SQL approaches that try to interpret the query holistically using DLmodels. Hybrid approaches that utilize both rule-based techniques as well as DLmodels are also emerging by combining the strengths of both approaches.Conversational interfaces are the next natural step to one-shot NLQ byexploiting query context between multiple turns of conversation fordisambiguation. In this article, we review the background technologies that areused in natural language interfaces, and survey the different approaches toNLQ. We also describe conversational interfaces for data analytics and discussseveral benchmarks used for NLQ research and evaluation.