Build an index

Search engines don’t read your documents at query time. They prepare them first. Paste some documents (one per line) or pick a preset, then hit Index documents and watch what happens.

Presets

Documents · one per line

A search engine finds the documents most relevant to a query.
An inverted index maps every term to the list of documents that contain it.
Term frequency counts how many times a term appears in a document.
Document frequency counts how many documents contain a term at least once.
Inverse document frequency rewards rare terms and discounts common ones.
TF-IDF multiplies term frequency by inverse document frequency.
BM25 improves on TF-IDF by saturating term frequency and normalizing for document length.
The k1 parameter in BM25 controls how quickly term frequency saturates.
The b parameter in BM25 controls how strongly long documents are penalized.
Stemming reduces words like running, runs and runner to a common root.
Stopwords such as the, and and of are often removed before indexing.
Tokenization splits raw text into terms before anything else happens.
Dense retrieval embeds queries and documents as vectors and compares them by similarity.
Hybrid search combines BM25 scores with vector similarity for better recall.
Precision measures how many returned documents are relevant; recall measures how many relevant documents were returned.

15 / 500 docs1,136 / 250,000 chars

Analyzer