Change in indexing technology

Why do you change the indexing?

We need our indexing technology to be more flexible and modular for future innovation (new content types, new information extracted from existing content).

What changes exactly?

We are improving the technology we use for the automatic indexing step of processing Embase’s records. For this, we are switching to a different technology: SciBite with its named entity recognition (NER) and extraction engine TERMite.

This technological improvement is the first step to providing our customers with improved indexing quality. This enhancement has been released on November 20th.

What stays as it used to?

• No change in format.
• Indexing Embase involves a two-step process for most articles (preconditions are explained in our indexing guideline): the first step is automatic indexing on the title and abstract of the Embase article. Then, after verification the preconditions, the next step is an in-depth manual indexing based on the content in the full text, usually available a few weeks later. This will continue to be done as it always has been, and as documented in this FAQ.
• This new technology will deliver a similar indexing quality (recall and precision) compared to our previous process which also leverages Emtree terms to automatically index content, making the Embase content available as soon as possible to customers. You will not need to make any changes to your workflow. In a comparable way to any time our indexing is updated (e.g., during the 3x/year Emtree releases), we recommend reviewing saved search/email alert queries and update them if needed.

How does this look like concretely?

Here are two examples where you can see that there can be slight differences with the new technology indexing more terms, different terms, or less terms.

What will happen in the future?

The next step will be for the initial indexing (on titles and abstracts) to be improved and complemented with medical devices, tradenames, and manufacturers from the full text article. This will enable our customers to retrieve the most relevant records sooner to improve precision and recall at an earlier stage while manual indexing is still performed in a second step. This is an improvement that our medical devices customers have been requesting. On a long term, the new technology will also allow the processing of new content types (new sources or new extracted information), which will provide further advancements in the upcoming years.