Neural Network Vector Normalization for Noncanonical Text
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing technologies face challenges in normalizing electronic communications that contain noncanonical terms, such as misspelled words, shorthand, and informal language, due to the variety of causes and types of noncanonical communications, which complicates textual analysis and normalization.
Innovation Solution
A system using a neural network, specifically a bidirectional gated recurrent neural network (BGRNN), preprocesses noncanonical communications by generating vectors and determining normalized versions based on probabilities generated by the neural network, deciding whether to output the normalized or original terms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional text processing methods are used to handle noncanonical communications, then the processing is simple and fast, but the accuracy of normalization is poor due to the variety of noncanonical forms
Solution Approach 1:
The patent introduces character n-gram features and contextual embeddings as intermediary representations between the raw noncanonical text and the normalization output. These intermediaries capture the variety of noncanonical forms (misspellings, slang, abbreviations) in a structured way, enabling accurate normalization without requiring complex rule-based systems for each variant.
Solution Approach 2:
The system transforms the input text into different parameter representations including character n-gram sequences, contextual embeddings, and probability distributions over possible canonical forms. By changing the parameter space from raw text to these transformed representations, the system can accurately handle the variety of noncanonical forms while maintaining a relatively simple processing architecture.
2Measurement precision
If a complex neural network model is used to improve normalization accuracy, then the normalization precision improves, but the processing time increases
Solution Approach 1:
The patent segments the text processing into character-level n-gram features rather than requiring full-word or sentence-level analysis. This segmentation allows the neural network to process smaller, more manageable units that can be quickly transformed and compared, reducing overall processing time while maintaining accuracy in handling various noncanonical forms.
Solution Approach 2:
The system uses character n-grams of limited size (e.g., 2-grams, 3-grams) rather than analyzing entire words or contexts. This partial action approach captures sufficient information for accurate normalization of common noncanonical forms while avoiding the computational burden of processing complete contextual information, thus reducing processing time.
3Measurement precision
If character n-gram features are extracted to capture various noncanonical forms, then the normalization accuracy improves, but the data processing complexity increases
Solution Approach 1:
The character n-gram feature extraction serves multiple functions simultaneously: it captures misspellings, slang, abbreviations, and other noncanonical forms within a single unified framework. This universal approach handles diverse noncanonical variations without requiring separate processing logic for each type, thereby improving accuracy while limiting the increase in processing complexity.
Data Source
AI summary
Electronic communications can be normalized using a neural network. For example, a noncanonical communication that includes multiple terms can be received. The noncanonical communication can be preprocessed by (I) generating a vector including multiple characters from a term of the multiple terms; and (II) repeating a substring of the term in the vector such that a last character of the substring is positioned in a last position in the vector. The vector can be transmitted to a neural network configured to receive the vector and generate multiple probabilities based on the vector. A normalized version of the noncanonical communication can be determined using one or more of the multiple probabilities generated by the neural network. Whether the normalized version of the noncanonical communication should be outputted can also be determined using at least one of the multiple probabilities generated by the neural network.