Neural Network Domain-Specific Stopword Removal

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for removing stopwords from unstructured computer text fail to account for domain-specific considerations, leading to inaccurate removal of noise words that degrade the quality of text analysis in downstream applications.

Innovation Solution

A neural network-based system that analyzes unstructured text to identify and remove domain-specific stopwords by generating word vectors in a high-dimensional space, using bootstrap keywords and regular expressions to determine candidate stopwords, and subsequently filtering out noise words.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If predefined lists of generic stopwords are used for removal, then the system can filter common noise words, but it fails to remove domain-specific stopwords that are important to the organization

Engineering Contradiction:

Improvestopword identification accuracyVSAvoiddomain-specific adaptability

Core Design Contradiction:

Measurement precisionVSAdaptability or versatility

Solution Approach 1:

The system changes the parameter of stopword identification from static predefined lists to dynamic neural network-based identification. The neural network learns domain-specific characteristics and adapts the stopword definitions based on the specific domain context, thereby improving both accuracy and domain adaptability simultaneously

Inventive Principle:

Principle #35Parameter changes

Solution Approach 2:

The patent replaces the mechanical approach of manually curated stopword lists with an intelligent neural network system. This substitution enables the system to automatically learn and identify domain-specific stopwords through pattern recognition, eliminating the need for manual domain expertise input while improving identification precision

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

2Quantity of substance

If regular expressions are used to define stopwords, then certain patterns can be captured, but the definitions become cumbersome and the number of possible combinations becomes very high

Engineering Contradiction:

Improvenumber of stopwords identifiedVSAvoidregular expression complexity

Core Design Contradiction:

Quantity of substanceVSDevice complexity

Solution Approach 1:

The patent replaces complex regular expression patterns with a neural network-based identification system. This substitution transforms the complex combinatorial problem of regex pattern matching into a learned pattern recognition task, significantly reducing system complexity while maintaining comprehensive stopword identification

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The neural network system performs self-learning to automatically identify stopword patterns without requiring explicit rule definitions. The system serves itself by learning from training data what constitutes a stopword in the specific domain, eliminating the need for manual regex construction and maintenance

Inventive Principle:

Principle #25Self-service

3Productivity

If statistical quality measures like term frequency are used to identify stopwords, then the system can process large amounts of text, but the techniques are generally inaccurate and fail to capture all critical stopwords

Engineering Contradiction:

Improvetext processing speedVSAvoidstopword identification accuracy

Core Design Contradiction:

ProductivityVSMeasurement precision

Solution Approach 1:

The patent replaces simple statistical measures with a neural network-based identification system. This substitution maintains high processing speed through efficient neural network inference while dramatically improving accuracy by learning complex patterns that statistical measures cannot capture

Inventive Principle:

Principle #28Mechanics substitution (Replace mechanical system)

Solution Approach 2:

The system combines multiple features (word embeddings, contextual information, domain-specific features) into a composite representation that the neural network processes. This composite approach enables simultaneous achievement of high speed and high accuracy by leveraging multiple information sources

Inventive Principle:

Principle #40Composite materials

Data Source

PatentUS10628471B2Domain-specific stopword removal from unstructured computer text using a neural network

Publication Date: 2020.04.21 FMR CORP

AI summary

Methods and apparatuses are described for analyzing unstructured computer text for domain-specific stopword identification and removal. A computer data store stores unstructured text. A server computing device splits the unstructured text into phrases and generates tokens from the phrases. The server computing device generates a set of bootstrap keywords using the tokens. An artificial intelligence neural network executing on the server computing device generates a stopword training model. The server computing device generates a first set of candidate stopwords using the bootstrap keywords and the stopword training model. The server computing device generates regular expressions using the bootstrap keywords, and generates a second set of candidate stopwords using the regular expressions. The server computing device stores the candidate stopwords in the data store, and removes stopwords from the unstructured text using the data store.