Context-Aware Data Correction Using Multi-Dimensional Word Vectors
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data correction techniques for the Know Your Customer (KYC) process in finance institutions are not scalable, fail to consider contextual semantic correctness, and struggle with correcting numerical data and noise characters effectively.
Innovation Solution
A method and device that encode each word in a document into a multi-dimension word vector, generate context word vectors based on preceding and succeeding words, and decode these vectors to produce corrected context word vectors for accurate data correction and validation using sequence-to-sequence machine learning algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If taxonomy based or dictionary based approaches are used to correct erroneous data, then spelling mistakes and punctuation errors can be corrected, but the method is not scalable and cannot correct numerical data or identify noise characters without context
Solution Approach 1:
The patent replaces the mechanical taxonomy-based correction system with a neural network-based semantic analysis system. The neural network automatically learns contextual relationships and semantic meanings from training data, eliminating the need for manual dictionary updates and enabling scalable correction of various error types including numerical data and noise characters.
Solution Approach 2:
The patent transforms the correction approach from fixed categorical matching (taxonomy) to continuous semantic space representation. By encoding words into vector representations and measuring semantic similarity through distance metrics in this continuous space, the system achieves both high precision and scalability simultaneously.
2Extent of automation
If conventional techniques generate text from noise using RNN and validate using generative adversarial network, then text generation is achieved, but the method does not correct data based on semantic context
Solution Approach 1:
The patent introduces context words as intermediaries between the erroneous word and its correction. The neural network identifies context words surrounding the error, encodes their semantic meanings, and uses these contextual representations to guide the correction process, thereby preserving semantic context throughout automatic text generation.
Solution Approach 2:
The patent implements feedback mechanisms where the generated text is continuously validated against semantic context. The neural network uses contextual information to evaluate and refine corrections iteratively, ensuring that generated text maintains semantic coherence and correctness.
3Ease of manufacture
If teacher forcing algorithm is used to train recurrent networks with observed sequence values, then training is achieved, but semantic context is not considered and errors are not corrected
Solution Approach 1:
The patent performs preliminary encoding of context words into semantic vectors before the correction process. This pre-computation of contextual representations enables the model to consider semantic context during training and inference, improving reliability while maintaining training efficiency through the preliminary preparation of contextual information.
Data Source
AI summary
A method and device for automatic data correction using context and semantic aware learning techniques is disclosed. The method includes extracting data within a document as machine readable text in a predefined format. The method further includes encoding each word of each line in the machine readable text to a multi-dimension word vector. The method includes generating a context word vector for each word in each line based on multi-dimension vectors associated with words succeeding and preceding the word in a line comprising the word. The method further includes decoding the context word vector associated with each word in each line to generate a corrected context word vector for each word. The method includes validating the corrected context word vector associated with each word in each line.