Predictive Text Compression Using NLG Continuations and Dictionary Fallback
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text compression techniques struggle to efficiently reduce the amount of computer data required to represent text strings, particularly due to the redundancy in text characters, strings, and sentences.
Innovation Solution
The method involves recognizing a prefix string preceding a target string in a text and using a natural language generation (NLG) model to predict continuations. If a matching prediction is found, the next characters are compressed as an NLG-type representation; otherwise, a compression dictionary is used to identify and compress the next characters as a dictionary-type representation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional compression techniques are used, then text can be compressed, but compression efficiency is insufficient due to text redundancy
Solution Approach 1:
The patent replaces traditional mechanical compression algorithms with a neural network-based predictive system. The neural network learns patterns in text data and predicts future characters, enabling more efficient compression by exploiting linguistic redundancy that traditional algorithms cannot capture. This substitution of mechanical processing with intelligent prediction directly addresses the insufficient compression efficiency problem.
Solution Approach 2:
The patent dynamically adjusts compression parameters based on the learned text patterns. The neural network modifies prediction probabilities and compression ratios according to the specific characteristics of the input text, allowing adaptive optimization of compression efficiency. This parameter adjustment enables the system to achieve better compression ratios while maintaining data integrity.
2Quantity of substance
If compression is applied to reduce data storage, then storage requirements decrease, but the complexity of compression and decompression processes increases
Solution Approach 1:
The patent performs preliminary training of the neural network model offline before actual compression operations. During this preliminary phase, the system learns text patterns and stores them in the network weights. This pre-learning eliminates the need for complex real-time pattern analysis during compression, reducing operational complexity while maintaining high compression efficiency. The decompression process becomes simpler as it only requires running the pre-trained model to generate predictions.
3Productivity
If more data is retained to improve compression accuracy, then compression efficiency improves, but network bandwidth usage increases
Solution Approach 1:
The patent extracts only the essential predictive information from the training data and stores it compactly in the neural network weights. During compression, the system generates predictions using this extracted knowledge rather than transmitting or processing the entire training dataset. This extraction of essential patterns enables high compression efficiency while minimizing network bandwidth usage, as only the compressed output and minimal model parameters need to be transmitted.
Data Source
AI summary
A method for text compression comprises recognizing a prefix string of one or more text characters preceding a target string of a plurality of text characters to be compressed. The prefix string is provided to a natural language generation (NLG) model configured to output one or more predicted continuations each having an associated rank. If the one or more predicted continuations include a matching predicted continuation relative to the next one or more text characters of the target string, the next one or more text characters are compressed as an NLG-type compressed representation. If no predicted continuations match the next one or more text characters of the target string, a longest matching entry in a compression dictionary is identified. The next one or more text characters of the target string are compressed as a dictionary-type compressed representation that includes the dictionary index value of the longest matching entry.