Introducing Copilot, a new feature for Scopus AI to handle specific and complex queries

Since the launch of Scopus AI last year, we’ve learned that it delivers great results when exploring a specific research area but can struggle to handle very narrow queries — for example, those asking about a certain gene or rare medical condition. Some topics are so niche there may be only a handful of papers that describe them.

To address this need, we’re pleased to introduce Copilot, the first in a lineup of new features for Scopus AI. Copilot is designed to optimize responses to highly specific or complex queries. It uses both keyword and vector search tools, employing more varied types of search technology to better process specialized queries and provide more specific responses.

Here's how Copilot works:

Copilot looks at the content of your query and decides whether to run a vector search and/or a keyword search.
It then ensures that complex queries are broken down into their component parts and optimized for the search channels it has chosen to use. For example, for a keyword search, it adds Boolean operators.
In doing so, it considers more abstracts than the previous search.

As a result of this process, users can pose more complicated or multifaceted questions, and Copilot will be able to sift through the different aspects to produce the best possible results.

Support of non-English queries

We realize exploring a new topic area can be difficult even if English is your first language. But if English isn’t your native language, delving into new subjects can be even more challenging. To address this issue, Copilot also includes an intelligent language layer, which translates queries from a non-English language into English, which is the language many researchers say they prefer to work in.

Copilot improves your Scopus AI experience in other ways, including:

Automatically correcting spelling mistakes to provide the clearest responses possible
Processing longer text queries, including full abstracts
Expanding the number of search results that Scopus AI provides — up from 10 in the Summary and 20 in the Expanded summary to a new upper limit of 30

In line with Elsevier’s Responsible AI principles, Copilot provides a reflection layer that explains exactly how the tool breaks down and optimizes your query. This is a unique feature among GenAI solutions currently on the market, and another step toward refining this exciting new tool.

The research community has been closely involved in every step of Scopus AI’s development, and the introduction of Copilot is also the result of customer feedback. Testing and offline evaluation shows Copilot greatly improves the quality of the results produced and enables Scopus AI to support a wider range of queries, from highly niche to very broad.

With Copilot, we hope to provide an even more nuanced and specialized view into the evolving world of academic research.