Polyseme Word Clustering via Iterative Label Assignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for word distributed learning fail to accurately cluster words with multiple meanings, leading to insufficient clustering accuracy due to the lack of consideration for the identified meanings.
Innovation Solution
A learning apparatus and method that performs clustering of documents for each word, assigns different labels to clusters, and re-clusters words with assigned labels to improve accuracy, specifically addressing the meanings of words like 'notebook' and 'table' by generating and updating context vectors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional clustering techniques are used for polysemes, then the processing is simple, but the clustering accuracy is insufficient because different meanings are not distinguished
Solution Approach 1:
The patent segments polysemes into multiple meaning-specific clusters by performing iterative clustering. In each iteration, words are clustered based on their context vectors, and when a polyseme is identified as having multiple meanings (through label assignment), it is segmented into separate meaning-specific clusters. This segmentation continues across multiple iterations until convergence, thereby improving clustering accuracy by distinguishing different meanings while managing complexity through systematic iteration.
2Measurement precision
If multiple iterations of clustering and re-clustering are performed, then the clustering accuracy improves, but the computation time increases
Solution Approach 1:
The patent performs preliminary actions by pre-processing the corpus to extract context vectors for each word before the iterative clustering begins. This preliminary preparation of context vectors and co-occurrence statistics allows the iterative clustering process to focus only on grouping words based on pre-computed semantic representations, reducing the computational burden during iterations while maintaining improved clustering accuracy through multiple passes.
3Measurement precision
If context vectors are generated and updated for each word, then the distributed learning accuracy enhances, but the memory requirements increase
Solution Approach 1:
The patent merges the context vector representation with the clustering structure by updating context vectors in-place during the iterative clustering process. Instead of maintaining separate large-scale storage for context vectors and cluster assignments, the system combines these representations, allowing context vectors to be updated and reused across iterations without duplicating storage. This merging approach enables improved distributed learning accuracy while controlling memory requirements through efficient data structure integration.
Data Source
AI summary
A learning apparatus includes a memory and a processor configured to acquire a plurality of documents, perform clustering of the plurality of documents for each of a plurality of words included in the plurality of document, when a plurality of clusters are generated for a first word among the plurality of words by the clustering, perform assignment of different labels corresponding to the plurality of clusters to the first word included in the plurality of documents, and perform re-clustering of the plurality of documents including the first word with the assigned different labels, for other words among the plurality of words.