Semantic Search Using 3-Tuple Keys and Context Information
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing similar document search technologies using semantic structures face inefficiencies due to high noise levels and increased calculation complexity when using 2-tuples as search keys, leading to missed desired documents and excessive noise inclusion, while 3-tuples offer a higher narrowing effect but require significant database matching and compatibility issues.
Innovation Solution
A search apparatus and method utilizing 3-tuples as search keys, generating context information based on appearance frequencies, and associating semantic structures with context information to improve document relevance and reduce noise, employing a correspondence relationship between semantic structures and context information to enhance search precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If 2-tuples are used as search keys in similar document search, then the search system can operate with simpler database matching, but the noise level increases and desired documents may be missed
Solution Approach 1:
The patent segments the search key from 2-tuples to 3-tuples by adding an additional semantic code dimension. This segmentation increases the specificity of search keys, allowing the system to distinguish between different semantic structures more effectively and reduce noise in search results.
Solution Approach 2:
The patent transitions from 2-dimensional search keys (2-tuples) to 3-dimensional search keys (3-tuples) by incorporating a third semantic code. This dimensional change enhances the narrowing effect and improves search precision without proportionally increasing database matching complexity.
2Measurement precision
If 3-tuples are used as search keys, then the narrowing effect increases ten times compared to 2-tuples, but the database matching and compatibility issues increase significantly
Solution Approach 1:
The patent performs preliminary action by pre-processing text data into semantic codes and organizing them into semantic structures before search operations. This preparation includes creating an inverted index that maps semantic codes to documents, which simplifies the actual search process even when using complex 3-tuple keys.
Solution Approach 2:
The patent introduces semantic codes as intermediaries between natural language text and the search system. These codes act as a bridge that transforms complex semantic relationships into a structured format that can be efficiently processed, reducing the complexity of database matching while maintaining high narrowing effects.
3Adaptability or versatility
If semantic analysis is performed on text data to enable similar document search, then the ability to find semantically similar documents improves, but the calculation complexity and processing time increase
Solution Approach 1:
The patent performs semantic analysis as a preliminary action during the text processing stage, converting text data into semantic codes and constructing semantic structures before the actual search operation. This pre-processing enables efficient search execution without requiring complex calculations during the search phase itself.
Solution Approach 2:
The patent creates a copied representation of the text data in the form of semantic codes and inverted indexes. This copied structure preserves the semantic information needed for search while being much more efficient to process than the original text, reducing calculation complexity during search operations.
Data Source
AI summary
A computer generates searching context information relating to appearance frequencies of words or semantic codes in a searching query sentence from the searching query sentence. Semantic analysis is performed on sentences included in documents so as to generate semantic structures, and each of the semantic structures includes three or more semantic codes and a relationship type among the semantic codes. A plurality of context information is generated and each context information relates to an appearance frequency of a word or a semantic code in a group of documents that match each of the semantic structures. The computer specifies one semantic structure of the semantic structures in accordance with the searching context information and the plurality of context information, and searches for a similar document by using the specified semantic structure.