Semantic Weight Adjustment for Text Representation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The Bag-of-Clusters (BOC) representation in text classification and clustering results in coarse-grained word clusters, leading to loss of information, as it fails to differentiate between texts that are semantically similar but not explicitly sharing the same words.
Innovation Solution
An information processing system that adjusts word weights in a weight vector based on semantic similarity, using a distance storing mechanism to increase weights of related words that do not explicitly occur in the text, thereby enhancing the semantic representation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of information
If Bag-of-Clusters (BOC) representation is used to group semantically similar words, then word similarity is captured, but information is lost due to coarse-grained clustering
Solution Approach 1:
The patent applies local quality by adjusting word weights individually based on their semantic distance to other words in the text, rather than applying uniform cluster-level weights. This allows each word to be represented with precision tailored to its specific semantic relationships, resolving the contradiction between capturing similarity and preserving information.
Solution Approach 2:
The patent segments the coarse cluster representations into individual word contributions, where each word's weight is adjusted separately based on its distance to other words. This segmentation preserves fine-grained information while still benefiting from semantic similarity, overcoming the information loss problem of BOC.
2Reliability
If word weights are adjusted based on semantic similarity, then texts with synonymous words are detected as similar, but computational complexity increases
Solution Approach 1:
The patent implements partial action by adjusting weights only for words that have semantic relationships with other words in the text, rather than computing all possible word relationships. This selective approach maintains high reliability in similarity detection while reducing computational complexity compared to exhaustive semantic analysis.
3Loss of information
If semantic similarity is used to adjust word weights, then zero entries are reduced in weight vector, but calculation time increases
Solution Approach 1:
The patent applies preliminary action by pre-calculating or efficiently computing semantic distances between words before weight adjustment. This allows the system to reduce zero entries in the weight vector while minimizing calculation time, as the semantic relationship computations are performed in advance or optimized during the weighting process.
Data Source
AI summary
An information processing system to increase weights of words that are related to a text, but that do not explicitly occur in the text, in a weight vector representing the text, is provided. An adjusting system (100) includes a distance storing unit (110) and an adjusting unit (120). The distance storing unit (110) stores distances between any two terms of a plurality of terms. The distance between two terms becomes smaller as the two terms are semantically more similar. The adjusting unit (120) adjusts a weight of each term of the plurality of terms in a weight vector including weights of the plurality of terms and representing a text, on the basis of a distance between each term and other term in the weight vector and a weight of the other term.