Neural Network Text Analysis Multi-Class Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing text analysis methods for predicting the association of compounds with a specific domain rely on binary decisions, which are not sufficient for nuanced classification, and lack the ability to distinguish between varying degrees of association.
Innovation Solution
A computer-implemented method using an artificial neural network is trained with features and identifiers to classify term candidates from a domain-specific corpus into multiple classes, allowing for a finer distinction of compound associations by considering the components and their productivity and frequency within the domain.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If binary decision methods are used for compound classification, then the method complexity is reduced, but the classification precision and ability to distinguish degrees of association deteriorates
Solution Approach 1:
The patent segments the classification task into multiple discrete classes (NONTERM, SIMTERM, TERM, SPECTERM) rather than using a single binary decision. This segmentation allows the system to distinguish between different degrees of association while maintaining manageable complexity through structured classification categories.
Solution Approach 2:
The patent transitions from binary (2 classes) to multi-class (4 classes) classification, adding a dimension of granularity to the classification system. This dimensional change enables the system to capture nuanced differences in compound-domain association that binary methods cannot detect.
2Measurement precision
If multi-class classification is implemented, then the classification precision and ability to distinguish degrees of association is improved, but the device complexity and training requirements increase
Solution Approach 1:
The neural network is designed with segmented processing stages: input layer for feature extraction, hidden layers for pattern learning, and output layer for multi-class prediction. This segmentation of the computational task enables multi-class classification while managing complexity through modular architecture.
Solution Approach 2:
The system uses parameter learning through training to adapt the neural network weights and biases, enabling it to automatically distinguish between different association degrees. The training process adjusts parameters based on labeled data to optimize classification performance across multiple categories.
3Measurement precision
If features like word vectors and productivity/frequency are added, then the prediction accuracy is improved, but the data processing complexity and computational requirements increase
Solution Approach 1:
The patent merges multiple feature types (word vectors, productivity, frequency) into a unified input representation for the neural network. By combining these features, the system achieves higher prediction accuracy while managing processing complexity through integrated feature representation rather than separate analysis streams.
Solution Approach 2:
The system performs preliminary feature extraction and preprocessing (computing word vectors, calculating productivity and frequency) before feeding data to the neural network. This preliminary action organizes and prepares the data in advance, reducing the computational burden during the main classification process and improving overall efficiency.
Data Source
AI summary
A computer-implemented method for training an artificial neural network using training data, which include features and identifiers, the features characterizing term candidates from a corpus, the corpus including a text from a domain, the identifier characterizing a degree of association to at least three classes for term candidates that differ from one another, different classes indicating different degrees of association of the term candidates to the domain, the training data including an assignment of features to identifiers. An artificial neural network, method for classifying term candidates, and computer-implemented method for generating training data, are also described.