Semantic Text Matching for Search Result Relevance

The patent segments the search ranking process into multiple independent scoring components: term match score, concept match score, and semantic match score. Each component evaluates different aspects of relevance independently, allowing the system to achieve high precision through aggregation of multiple simple scores rather than one complex scoring mechanism.

The patent introduces concept tags as an intermediary layer between query terms and document content. Instead of directly matching words to concepts, the system uses concept tags as mediators that bridge the gap between literal word matching and semantic understanding, enabling relevance improvement without direct complex analysis of the entire document.

The patent performs preliminary action by pre-computing and storing concept tags for documents during indexing. The concept tags are extracted and stored in advance, so during search, the system only needs to match pre-computed tags rather than performing complex semantic analysis in real-time, significantly reducing search processing time while maintaining accuracy.

The patent creates a simplified copy of the document's semantic structure through concept tags. Instead of working with the full complex document content during search, the system uses these compressed conceptual representations (copies) that capture the essential meaning without the redundant details, enabling fast matching while preserving accuracy.