Convolutional Neural Network Word Extraction via Segmented Context Windows
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for extracting specific words related to event information from text, such as those using convolutional neural networks, face a trade-off where increasing the window size to consider context reduces ambiguity but decreases accuracy in word position detection.
Innovation Solution
An information processing apparatus and method that generates training word pairs by shifting a window of a predetermined length over the text, associating semantic vectors with each word pair, and using a convolutional neural network to calculate probabilities and update the determination model for accurate extraction of specific words.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the window size is increased to consider more context, then the ambiguity of word meaning is reduced, but the accuracy of word position detection decreases
Solution Approach 1:
The patent segments the context window into multiple sub-windows, where each sub-window contains a subset of words from the full context window. This allows the system to process context information in manageable portions while maintaining the ability to detect word positions accurately. The segmentation enables parallel processing of different context regions without diluting the precision of individual position detections.
Solution Approach 2:
The patent introduces a new dimension by creating multiple overlapping windows at different positions around the target word. Instead of using a single large window, the system generates multiple smaller windows that overlap, each centered at different offsets. This dimensional transformation allows the system to capture contextual relationships from multiple angles simultaneously, improving both context consideration and position detection accuracy.
2Quantity of substance
If a single large window is used to extract context, then context information is comprehensive, but the system complexity increases
Solution Approach 1:
The patent divides the large context window into multiple smaller sub-windows, each containing a manageable number of words. This segmentation reduces the complexity of processing individual windows while collectively providing comprehensive context information. Each sub-window can be processed independently using simpler processing units, reducing overall system complexity.
Solution Approach 2:
The patent uses multiple overlapping windows that collectively cover the entire context region, with some words appearing in multiple windows. This partial action approach ensures that sufficient context information is captured without requiring a single enormous window. The overlapping structure provides redundancy that simplifies processing while maintaining comprehensive coverage.
Data Source
AI summary
An information processing apparatus includes a lexical analysis unit that generates a training word string, a pair generation unit that generates a plurality of training word pairs, a matrix generation unit that generates, for each training word pair, a training matrix in which a plurality of words and respective semantic vectors of the words are associated, a classification unit that calculates, for a word of each position of the training word string, a probability of the word corresponding to a specific word, using the training matrices generated by the matrix generation unit and a determination model that uses a convolutional neural network, and an optimization processing unit that updates parameters of the determination model, such that the probability of the word labeled as corresponding to the specific word is high, among the probabilities of the words of the respective positions of the training word string calculated by the classification unit.