Search Apparatus Using Frequency-Based Document Segmentation

The patent divides the large document set into multiple regions based on document frequencies. The word document matrix is segmented into multiple sub-matrices, each representing a specific frequency range. This allows the system to process and store data in smaller, manageable chunks rather than requiring all data to be loaded into memory simultaneously, thus reducing memory capacity requirements while maintaining search efficiency.

The patent pre-calculates and stores frequency information in an abstract matrix before the actual search operation. By performing frequency calculations in advance and organizing data by frequency ranges, the system prepares the data structure optimally for quick access during search operations, eliminating the need to load and process entire document sets during searching.

The patent segments the document set into frequency-based regions and processes only the relevant regions for each search query. By dividing the large document set into smaller frequency ranges and using region upper limit calculations, the system can determine which regions need to be examined, thereby reducing processing time while maintaining accurate frequency measurements for the queried keywords.

The patent implements a region upper limit calculation that determines the minimum necessary processing scope. Instead of processing the entire document set, the system calculates and processes only up to the upper limit of relevant frequency regions, achieving sufficient accuracy for search purposes without the excessive time cost of complete processing.