Enterprise Document Search Indexing via Contextual Passage Analysis

The patent segments the search process into multiple stages: initial broad search using simple text indexing, followed by targeted refinement using variation indexing only on promising results. This segmentation allows the system to use lightweight indexing methods initially, then apply more computationally intensive methods only when necessary, resolving the contradiction between search capability and computing power requirements.

The patent applies partial variation indexing rather than complete indexing of all documents. By using text indexing for initial retrieval and only applying variation indexing to a subset of relevant documents, the system achieves adequate search capability without the full computational burden of indexing every document with all its variations.

The patent performs preliminary text indexing and basic term matching before applying more complex co-occurrence analysis. This preliminary action filters out clearly irrelevant documents early, so that co-occurrence indexing is only applied to a reduced set of candidate documents, thereby improving search relevance without the full computational cost of analyzing all documents.

The patent applies different indexing strategies to different parts of the search process: simple text indexing for initial retrieval, and more complex variation indexing and co-occurrence analysis only for refining results. This local quality approach ensures high search relevance for final results while minimizing overall computational resource consumption.