Magnitude-Aware Skip CNN for Phrase Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current machine learning models, particularly convolutional neural networks (CNNs), face challenges in accurately interpreting the semantic meaning of natural human language due to their inability to effectively handle skipped ngrams, which can alter the meaning of phrases and lead to incorrect predictions.
Innovation Solution
The implementation of a magnitude-aware skip CNN that determines the semantic meaning of skipped ngrams using a skip word setting and a skip word scalar engine, allowing the model to incorporate the semantic impact of skipped words into its predictions, thereby improving the accuracy of phrase recognition and interpretation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If a filter with skip word setting is applied to ngrams during execution, then the model can handle variations in natural language expressions, but the semantic meaning of skipped ngrams is lost leading to inaccurate predictions
Solution Approach 1:
The patent extracts the skipped ngrams from the filtering process and processes them separately through a skip word scalar engine. This allows the main filter to handle language variations while the extracted skip words are independently evaluated for their semantic contribution and used to adjust the final prediction score.
Solution Approach 2:
The skip word scalar engine acts as an intermediary between the filtering process and the final prediction. It takes the skipped ngrams, calculates their semantic scalar values, and mediates their impact on the final score by combining them with the intermediate score from the filter match.
2Measurement precision
If all ngrams are processed to determine semantic meaning, then prediction accuracy improves, but computational complexity and processing time increase
Solution Approach 1:
The patent segments the ngram processing into two distinct paths: ngrams that match the filter are processed through the standard matching process, while skipped ngrams are processed separately through the skip word scalar engine. This segmentation allows efficient handling of different ngram types with appropriate processing depth.
Solution Approach 2:
The patent applies different processing quality to different ngrams based on their role. Matched ngrams receive standard processing, while skipped ngrams receive focused semantic analysis only when they significantly impact the prediction. This local quality approach optimizes computational resources by not uniformly processing all ngrams at maximum detail.
3Productivity
If skip words are completely ignored to simplify processing, then computational efficiency improves, but semantic accuracy deteriorates
Solution Approach 1:
The skip word scalar engine provides self-service by automatically identifying skipped ngrams that have significant semantic impact and processing them accordingly. It uses learned scalar values to determine when skip words should influence the prediction, eliminating the need for manual configuration or complete processing of all skip words.
Data Source
AI summary
A method including inputting, into a phrase recognition model comprising a neural network, a vector comprising a plurality of ngrams of text. The method also includes applying, using the phrase recognition model, a filter to the plurality of ngrams during execution. The filter has a skip word setting of at least one. The method also includes determining, based on the skip word setting, at least one ngram in the vector to be skipped to form at least one skip word. The method also includes outputting an intermediate score for a set of ngrams that match the filter. The method also includes calculating a scalar number representing a semantic meaning of the at least one skip word. The method also includes generating based on the scalar number and the intermediate score, a final score for the set of ngrams. A computer action is performed using the final score.