How BibGenie Understands Your Library

Why this page exists

When you search in BibGenie, the goal is not only to find papers that contain the exact same words you typed. In many cases, you want BibGenie to find papers that are about the same idea, even if they use different terminology.

That is where embeddings and indexing come in.

What is an embedding?

An embedding is a way to turn a piece of text into a numeric representation that captures its meaning.

You do not need to think about the math behind it. The practical idea is simple:

• Similar text tends to produce similar embeddings
• Different topics tend to produce more distant embeddings
• This lets BibGenie compare meaning, not only exact wording

For example, a paper about drug discovery with knowledge graphs may still be found when you search for graph methods for biomedical research, even if the wording is not identical.

How BibGenie uses embeddings

BibGenie currently builds semantic search from the title and abstract of your regular Zotero items.

The process is:

1. BibGenie reads the title and abstract of a paper.
2. It converts that text into an embedding.
3. It stores the result in a local search index for your Zotero library.
4. When you ask a topic-style question, BibGenie converts your query into an embedding too.
5. It compares your query with the indexed papers and returns the closest matches.

This is why semantic search can surface papers that are relevant in meaning, not only in wording.

What is indexing?

Indexing is the preparation step that makes semantic search fast and practical.

Instead of recomputing everything every time you search, BibGenie first prepares a local index of embeddings for your library. Once that index exists, BibGenie can search across your papers much more efficiently.

In short:

• Embedding helps BibGenie represent meaning
• Indexing helps BibGenie search that meaning quickly

What does this help with?

Semantic indexing is especially useful when you want to:

• Find papers related to a research topic
• Discover work that uses different wording for a similar idea
• Explore a method, theory, or problem area
• Search a large Zotero library more naturally

Keyword search is still useful when you know the exact author, title, term, or phrase. Semantic search is more helpful when your question is conceptual.

What content is included?

BibGenie currently uses:

• Paper titles
• Paper abstracts

BibGenie does not currently build this semantic index from:

• PDF full text
• Notes
• Annotations
• Attachments

This helps keep indexing focused, predictable, and fast.

Do I need to rebuild the index often?

Usually, no.

Most of the time, BibGenie handles indexing updates automatically in the background as your library changes. If you add papers, modify records, or remove items, BibGenie will usually take care of the necessary updates for you.

You may want to rebuild the index in a few situations:

• The first time you set up semantic indexing
• After switching to a different embedding model
• After a large import, if results seem incomplete or outdated
• If BibGenie tells you the current index needs attention
• If semantic results look clearly wrong and you want a clean rebuild

What do the buttons mean on the Indexing page?

Build / Rebuild

This creates the index for the first time, or rebuilds it from scratch when needed.

Use it when:

• You are setting up indexing for the first time
• You changed embedding models
• You want to refresh the entire index

Retry Failed Items

This only retries papers that failed during previous indexing attempts.

Use it when:

• The failed item count is greater than zero
• You want a lighter recovery step before trying a full rebuild

Common questions