Phrase-Similarity Vector Space for Style Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods fail to effectively transform the style of original works into derivative works without significant alteration of the original content or meaning, and lack efficient data structures for similarity analysis.
Innovation Solution
A method that utilizes a similarity data structure with phrase-similarity vector spaces to replace original phrases in an original work with similar phrases from a corpus, based on predetermined similarity thresholds, to create a derivative work in a specific style, employing a neural network for similarity analysis and style transformation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional text transformation methods are used, then the original content and meaning may be preserved, but the style transformation effectiveness deteriorates
Solution Approach 1:
The patent segments text into phrases rather than processing words individually or treating entire sentences as units. This phrase-level segmentation enables more granular style transformation while preserving original meaning, as phrases capture meaningful linguistic units that balance semantic integrity with stylistic flexibility.
Solution Approach 2:
The patent introduces a similarity data structure as an intermediary between the original text and the transformed output. This data structure stores pre-computed similarity relationships and serves as a mediator to guide phrase substitutions that maintain meaning while achieving style transformation, resolving the contradiction between preservation and transformation effectiveness.
2Manufacturing precision
If comprehensive similarity analysis is performed, then style transformation accuracy is improved, but computational time and resource consumption increase
Solution Approach 1:
The patent performs similarity analysis in advance and stores the results in a similarity data structure before actual style transformation is needed. This preliminary computation of phrase similarities allows the transformation process to simply look up pre-computed values rather than performing complex calculations in real-time, significantly reducing computational time while maintaining accuracy.
Solution Approach 2:
The patent creates a copied representation of phrase similarity relationships in the similarity data structure, which is a simplified version of the full similarity analysis. This copied structure enables fast lookup and comparison operations during style transformation without requiring access to the original, computationally intensive similarity calculation process.
3Measurement precision
If a detailed data structure for similarity analysis is created, then phrase similarity lookup accuracy is improved, but storage requirements and data structure complexity increase
Solution Approach 1:
The patent extracts only the essential similarity information needed for style transformation and stores it in the similarity data structure, rather than maintaining complete and exhaustive similarity relationships. This extraction of necessary data maintains lookup accuracy while reducing data structure complexity and storage requirements by eliminating redundant information.
Data Source
AI summary
Aspects of the invention include receiving an original work written in a first style to be transformed into a derivate work written in a second style. The transforming includes accessing a similarity data structure organized to include a phrase-similarity vector space of phrase vectors representing degrees of similarity between phrases in a corpus of works written in the second style. An original sentence of the original work that includes one or more original phrases is obtained. The original phrases are located in the phrase-similarity vector space. For each of the original phrases, a similar phrase from the phrase-similarity vector space is selected based on the selected similar phrase having a phrase-similarity index corresponding to the original phrase that is greater than a predetermined phrase-similarity index threshold. The derivative work is created by replacing the original phrases in the original work with the selected similar phrases. The derivative work is output.