Weighted Character N-gram Vector Generation for Text Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating word vectors do not adequately consider the importance of character strings, leading to inadequate representation and analysis of natural language texts.
Innovation Solution
A word coding device and method that uses a pre-learned neural network to convert words into composite vectors by calculating weights for character N-grams, combining these vectors to generate a weighted composite vector representing the word, which is then used for accurate text analysis and language model learning.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If character N-gram vectors are combined with equal weights, then the word vector can be generated simply, but the importance of different character strings is not considered leading to inadequate representation
Solution Approach 1:
The patent changes the parameter of vector combination by introducing weight coefficients that vary for different character N-grams. Instead of uniform combination, each character N-gram vector is multiplied by a learned weight parameter that reflects its importance, thereby improving representation accuracy while maintaining a relatively simple additive combination mechanism.
Solution Approach 2:
The patent applies local quality by assigning different weights to different character N-grams within the same word. Each character N-gram can have its own weight parameter, allowing the system to emphasize important character sequences (like meaningful morphemes) while de-emphasizing less important ones, thereby creating a more nuanced and accurate word vector representation.
2Measurement precision
If traditional word embedding methods are used, then the processing is simple and fast, but the analysis accuracy of natural language texts is insufficient
Solution Approach 1:
The patent segments words into character N-grams before vectorization. Instead of treating each word as a single unit, it breaks down words into smaller character sequences (N-grams), vectors each segment separately, and then combines them with weights. This segmentation allows the model to capture sub-word patterns and improve text analysis accuracy, especially for out-of-vocabulary words.
Solution Approach 2:
The patent creates composite word vectors by combining multiple character N-gram vectors with different weights. This composite approach integrates information from different character-level representations to form a more comprehensive and accurate word embedding, improving text analysis performance while maintaining computational efficiency.
3Manufacturing precision
If character N-grams are decomposed into individual characters, then the vector representation becomes simpler, but the importance of character sequences is lost
Solution Approach 1:
The patent introduces dynamic weight parameters for different character N-grams that are learned during training. These weights allow the system to adaptively determine the importance of each character sequence based on the data, making the representation flexible and accurate without requiring manual intervention or complex rule-based systems.
Data Source
AI summary
To generate a vector representing a word in consideration of the importance of the character string of the word. A character N-gram vector acquisition unit acquires a vector representing each of character N-grams included in a word. A weight calculation unit 40 calculates a weight for a vector representing each of the character N-grams on the basis of the vector representing each of the character N-grams included in the word, and a character N-gram vector combining unit 42 generates a composite vector obtained by combining the vector representing each of the character N-grams on the basis of the vector representing each of the character N-grams included in the word and the weight for the vector representing each of the character N-grams.