R-Tree Full Text Search Using Token Pair Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Full text search systems using inverted indexes face performance issues when handling multiple search terms, leading to slower query response times due to the need to iterate over large document sets, even if the result set is small.
Innovation Solution
A method and system that generate token pairs and query ranges based on user input, allowing for efficient querying of a multi-dimensional index, such as an R-tree, to improve search query speed by converting data tokens into vectors and using these vectors for window queries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If inverted indexes are used for full text search, then document retrieval is efficient for single term queries, but query performance degrades significantly when handling multiple search terms due to iteration over large document sets
Solution Approach 1:
The patent transforms the traditional one-dimensional inverted index structure into a multi-dimensional index space by creating token pairs from document tokens and mapping them to dimensional coordinates. This dimensional transformation allows the system to query multiple terms simultaneously through geometric operations (window queries) rather than sequential set intersections, resolving the time performance degradation while maintaining search accuracy.
Solution Approach 2:
The patent segments the document collection into multiple dimensional spaces by creating token pairs and organizing them in a multi-dimensional index. Each dimension represents a term combination, allowing the system to divide the search problem into manageable dimensional segments that can be queried efficiently using window operations rather than processing the entire document set.
2Reliability
If separate searches are executed for each search term using inverted indexes, then comprehensive results are obtained, but the number of iterations increases leading to slower query processing
Solution Approach 1:
The patent merges multiple separate term searches into a single unified query operation by representing multiple search terms as a dimensional window. Instead of executing separate inverted index queries for each term and then intersecting results, the system combines all term constraints into one geometric window query that retrieves documents satisfying all conditions simultaneously, thereby maintaining search completeness while dramatically improving processing speed.
Solution Approach 2:
The patent performs preliminary action by pre-computing and storing token pairs in a multi-dimensional index structure before queries are executed. This preprocessing organizes the data into an optimized spatial structure that enables rapid window queries, eliminating the need for runtime iteration over large document sets and significantly boosting query processing productivity.
3Adaptability or versatility
If traditional inverted indexes are used, then the system works well for documents with many words, but performance improves for documents with few words using alternative designs
Solution Approach 1:
The patent creates a universal search system that adapts to documents of any length by using token pairs and multi-dimensional indexing. The system maintains flexibility to handle both short and long documents equally well through its dimensional representation, while achieving improved query response speed for all document types through efficient window queries that avoid unnecessary iterations over large document sets.
Data Source
Figure 1
Figure 2~10
Figure 3
AI summary
A method and system for full text search (FTS) using a navigation system is disclosed. The FTS system generates vectors for documents that are stored in a database. The vectors are formed by converting data token pairs to binary or hexadecimal strings. The FTS system may generate and store vectors in a multi-dimensional index (e.g., an R-Tree)for each document stored in the database. The FTS system may receive two or more search terms inputted by a user and generate one or more vectors based on the inputted search terms. The FTS system may then use the generated vectors to query the multi-dimensional index.