A systematic review of intelligent assistants

An intelligent assistant (IA) is a computer system endowed with artificial intelligence and/or machine learning techniques capable of intelligently assisting people. IAs have gained in popularity over recent years due to their usefulness, significant commercial developments, and a myriad of scientific and technological advances resulting from research efforts by the computer science community. In particular, these efforts have led to an increasingly extensive and complex state of the art of IAs, making evident the need to carry out a review in order to identify and catalogue the advances in the construction of IAs as well as to detect potential areas of further research. This paper presents a systematic review aiming to (i) describe, classify, and organize recent advances in IAs; (ii) characterize IAs' objectives, application domains, and workings; (iii) analyze how IAs have been evaluated; and (iv) identify what artificial intelligence and machine learning techniques are used to enable the intelligence of IAs. As a result of this systematic review, it is also proposed a taxonomy of IAs and a set of potential future research directions to further improve IAs. A set of research questions was formulated to guide this systematic review and address the proposed objectives. Well-known scientific databases were searched for articles on IAs published from January 2015 to June 2021 using 172 search strings. A total of 22,554 articles were retrieved and after applying inclusion, exclusion, and quality criteria, an overall number of 99 articles were selected, which are the basis for this systematic review. (c) 2021 Elsevier B.V. All rights reserved.

Automated summary

An intelligent assistant (IA) is a computer system equipped with artificial intelligence and machine learning techniques to assist people intelligently. The systematic review in this paper aims to describe, classify, and organize recent advances in IAs, characterize their objectives, application domains, and workings, analyze evaluation methods, and identify AI and ML techniques used. The review also proposes a taxonomy of IAs and future research directions to enhance IA.