AI-based Virtual Assistants
New AI-Based Virtual Assistants in Web of Science and Scopus Databases
With the emergence of Artificial Intelligence, some might say that the only skill we need is to know how to write a prompt—a notion we’ve seen before in reference to Google’s search engine. This, of course, is not entirely accurate. However, alongside the limitations of artificial intelligence, there is also immense potential.
Major players in the academic world, such as longstanding databases, advanced search systems, and more, are now cautiously adopting this new technology. Their aim is to leverage its advantages without compromising the academic value of the content: the quality and reliability of results, and of course, fair use—core principles of academia as we know it.
The first to implement an AI-based “virtual assistant” are the WOS and Scopus databases.
The virtual assistants were specifically designed for these databases and are based on GPT-4-o technology. What sets them apart from generic artificial intelligence is the scope (and type) of information they handle and their ability to cite sources.
These assistants use a Retrieval-Augmented Generation (RAG) language model, a technology that defines and limits the information retrievable through the chat and can refer to references, meaning it cites the sources of the content. This is in contrast to general-purpose AI, which operates on an unlimited amount of material and cannot specify the origin of the content.
As a result, these virtual assistants perform searches exclusively within specific sources—the database/catalog content—rather than drawing from an infinite pool of online content. The model allows users to write searches in completely free-form language, in English or Hebrew, and provides references to actual articles with specific citations. The prompts we define will yield clear results and concrete answers based on the abstracts of the articles.
Each Virtual Assistant Is Slightly Different, with Unique Professional Adaptations
For example, in the summaries generated by Scopus’ virtual assistant, some of the articles will have many citations, while others will only have few or even none. This is done intentionally, in order to reduce the researchers’ tendency to repeatedly cite from the same studies.
In a designated category called “Foundation Articles,” the system cross-references information from the bibliographic list of all the articles included in the summary and highlights the common source shared by all these articles—identifying the foundational studies.
Link to the webinar recording on Scopus AI
Link to the webinar recording on WoS Research Assistant
We would love to hear from you
The Technion libraries are here to assist you with research, teaching and studying.
We are happy to help with any question or request. Please send us your comments and suggestions for the newsletter in the form.
Any feedback is welcomed and will help us create more relevant and up-to-date content for you.
2025 \ Issue #11 \ Writing, editing and graphic design: Reference & Instruction team (Lior Porat, Ella Bogillo, Karen Dabran & Ortal Renford)