Synonym Extraction Using Noun Type Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for calculating the similarity between compound words using co-occurrence vectors are not accurate enough, leading to difficulties in extracting synonyms from documents.
Innovation Solution
A synonym extraction apparatus that determines the types of nouns in compound words, extracts groups with identical patterns, creates co-occurrence vectors for nouns in the same sentence, and calculates similarity to identify synonyms above a predetermined threshold.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If co-occurrence vectors are used to calculate similarity between compound words, then synonym extraction can be performed, but the calculation accuracy is not high enough
Solution Approach 1:
The patent segments compound words into constituent nouns and analyzes their individual types (Sahen-noun or other noun). By dividing the compound word into parts and examining each noun's characteristics separately, the system can create more nuanced co-occurrence vectors that capture the semantic structure of compound words, thereby improving similarity calculation accuracy and synonym extraction reliability.
2Device complexity
If only co-occurrence vectors of compound words are used for similarity calculation, then the process is simple, but the synonym extraction accuracy is insufficient
Solution Approach 1:
The patent adds a new dimension to the co-occurrence vector by incorporating information about the types of constituent nouns (Sahen-noun or other noun) within compound words. This dimensional enhancement transforms the vector from capturing only external co-occurrence patterns to also encoding internal structural information, thereby improving measurement precision without significantly increasing processing complexity.
Data Source
AI summary
A synonym extraction apparatus determines, for compound words included in the document, that types of nouns constituting the compound word are each a Sahen-noun or another noun type to determine a pattern of a sequence of the types of the nouns constituting the compound word, extracts a group of compound words having an identical pattern of the sequence of the types, extracts compound words having an identical leading or ending word from among them, and creates, for a group of compound words having the identical pattern of the sequence of the nouns and the identical leading or ending word, a co-occurrence vector having, as a component, a noun appearing in the same sentence as the corresponding compound word, and outputs, as synonyms, a group of compound words having a degree of similarity between the co-occurrence vectors of the compound words equal to or greater than a predetermined threshold.