CONTEXT Index

The CONTEXT index is Oracle Text’s primary index type, designed for full-text search on columns containing unstructured text: documents, articles, legal opinions, technical notes. Unlike a standard B-tree index, it does not index discrete values but linguistic tokens, building an inverted structure that maps each word to the documents where it appears.

How it works #

At creation time, Oracle Text tokenizes the column content (typically a CLOB), applies linguistic filters (stemming, stopwords, optional thesaurus) and populates a set of internal tables — $I, $K, $R, $N — that form the inverted index. Queries use the CONTAINS operator instead of LIKE:

SELECT doc_id, SCORE(1) AS relevance
FROM documents
WHERE CONTAINS(body, 'contract AND (lease OR rental)', 1) > 0
ORDER BY relevance DESC;

The index avoids a full sequential scan of the CLOB column on every query, dramatically reducing response times on large datasets.

When to use it #

The CONTEXT index is the right choice when:

• the column holds long, variable text (documents, converted PDFs, XML);
• queries require boolean operators, proximity search, or concept-based retrieval;
• data volume makes any LIKE '%...%' approach impractical.

One significant operational constraint: the CONTEXT index does not update in real time. Rows inserted or modified after index creation become searchable only after an explicit synchronization (CTX_DDL.SYNC_INDEX) or via a scheduled job. For write-heavy workloads, a sync strategy aligned with the acceptable search latency must be planned upfront.