Artifact Dataset
Corpus
Canonical ID https://id.searchplex.net/corpus/
A collection of documents, passages, web pages, items, or other retrievable units used for search, retrieval, training, or evaluation.
Also called
- Document Corpus
- Document Collection
Scope
In IR evaluation, the corpus is the searchable collection component of a test collection. The bare term collection is ambiguous: it may mean a corpus or searchable document collection, while test collection usually means the full evaluation package with corpus, topics or queries, and relevance judgments.
Observed terms
- Collection
- Search Collection
Part of
Not equivalent to
- Corpus is not BEIR. A corpus is a collection of retrievable items; BEIR is a benchmark suite aggregating multiple evaluation datasets.
- Corpus is not Qrels. A corpus contains retrievable items; qrels contain relevance judgments for query-item pairs.
- Corpus is not Test Collection. A corpus is the searchable collection; a test collection also includes queries or topics and relevance judgments.
Learn more
ID
- Canonical ID
https://id.searchplex.net/corpus/ - Scheme
Searchplex ID - Machine
JSONJSON-LD