In an era when many aspects of artificial intelligence (AI) and machine learning (ML) are increasing in popularity because of the proliferation of technology via a range of representational state transfer (REST) application programming interfaces (APIs) and frameworks--such as those intended for natural language processing (NLP), including text-to-text translation services, conversational user interfaces, and intelligent search engines--smart technology is becoming increasingly accessible for developers and businesses to create their own services.

There is an ever-increasing interest in smarter, more assistive search engines and applications--think of question-answering systems rather than mere keyword-based search. Understanding the intention of a query has always been one of the key intelligent applications necessary to improve the user experience with a computer in matters of getting answers rather than documents from a querying session.

Since query autocompletion was invented, it has been practiced in a plethora of currently available search engines. This paper provides an excellent overview of query autocompletion technology; however, it seeks to improve the performance of query autocompletion technologies by reducing the number of suggestions for completions, some of which are notoriously redundant. Removing the redundant candidates for query completion also increases space for the inclusion of more relevant suggestions; hence, it increases the probability of formulating a more semantically precise query. Another advantage of the proposed technique is that it works as a cold-start approach, that is, when no training data are available. The proposed technique and algorithms have been evaluated on the basis of mean reciprocal ranking (MRR), a well-established evaluation technique for web search.

In that context, the paper is of a rather narrow scope and, therefore, of limited interest when it comes to developing search technologies, engines, and systems. Even though the development of information retrieval systems and search engines has a much broader scope in that there is more than one factor affecting their performance, the paper may only be interesting for researchers in the query autocompletion field. Nonetheless, the paper is very well written and relies on an experimental setup, which can be recommended as guidance for aspiring PhD students, despite questions about whether the results and experiments can be reproduced, an issue recently raised as a concern by the scientific community with regard to many journal publications.