Posting List Storage via Semantic Role Hierarchy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Inverted indices used in search engines can grow large and suffer from poor performance due to arbitrary layouts of posting lists on storage media, especially when performing intersection operations for related terms in a strict dominance relation.
Innovation Solution
A hierarchy of semantic roles is defined using a role tree, where posting lists for related semantic roles are stored contiguously on a physical storage medium, allowing for efficient retrieval as a single contiguous block through pre-order depth-first traversal and additional metadata for location and size.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If posting lists are stored in an arbitrary layout on physical storage medium, then storage flexibility is maintained, but retrieval performance deteriorates when performing intersection operations on related terms
Solution Approach 1:
The patent segments the inverted index into hierarchical groups based on semantic role relationships. Posting lists are organized into parent-child groups where related terms (e.g., synonyms, hypernym-hyponym pairs) are grouped together. This segmentation allows the system to retrieve only relevant subsets of posting lists for intersection operations, improving retrieval speed while maintaining manageable storage complexity through structured organization.
Solution Approach 2:
The patent introduces a hierarchical dimension to the storage layout by organizing posting lists according to semantic role relationships (parent-child hierarchies). This hierarchical organization adds a structural dimension to the storage layout, enabling efficient navigation and retrieval of related posting lists through the hierarchy, thereby improving retrieval performance without creating arbitrary complexity.
2Measurement precision
If separate posting lists are created for each term-role pair in semantic indexing, then search accuracy is improved, but index size increases significantly
Solution Approach 1:
The patent merges posting lists of terms that share the same parent role in the semantic hierarchy. By combining posting lists of child roles under a common parent, the system maintains the semantic differentiation needed for accurate search while reducing redundancy in the index structure. This merging approach preserves search accuracy for role-specific queries while decreasing overall index size through consolidation of related posting lists.
Solution Approach 2:
The patent creates a multi-functional index structure where each posting list entry can serve multiple purposes depending on the query type. The hierarchical organization allows the same posting list to be used for both specific role-based searches and broader semantic searches, increasing the utility of each stored entry and reducing the total quantity of data needed to maintain search accuracy.
3Loss of time
If posting lists are loaded individually from storage medium, then memory usage is optimized, but retrieval time increases for related semantic roles
Solution Approach 1:
The patent implements preliminary organization of posting lists into hierarchical groups during the indexing phase. By pre-grouping posting lists according to semantic role relationships and storing them in an organized hierarchical structure, the system enables faster retrieval of related posting lists without requiring individual loading. This preliminary organization reduces retrieval time for related semantic roles while maintaining efficient memory usage through selective loading of hierarchical groups.
Data Source
AI summary
A role tree having nodes corresponding to semantic roles in a hierarchy is defined. A posting list is generated for each association of a term and a semantic role in the hierarchy. The posting lists are stored contiguously on a physical storage medium such that a subtree of the hierarchy of semantic roles can be loaded from the storage medium as a single contiguous block. The posting lists for a subtree of the hierarchy are retrieved by obtaining data identifying the beginning location on the physical storage medium of the posting lists for the term at the top of a desired subtree of the hierarchy and data identifying the length of the posting lists of the desired subtree of the hierarchy. A single contiguous block that includes the posting lists for the desired subtree of the hierarchy is then retrieved from the beginning location through the specified length.