NLP Command Processing via Charm Feature Extraction

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current natural language processing (NLP) systems face inefficiencies and high resource requirements when processing user-generated speech, particularly in converting speech into machine-executable commands, due to the need for large tagged text corpora to improve accuracy.

Innovation Solution

The method involves a server-based system that uses a training module to generate and analyze features for user-generated commands, employing algorithms like the Charm Bitset algorithm to determine feature patterns without requiring a full text corpus, thereby reducing resource intensity and improving processing efficiency.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If the size of the tagged text corpus is increased to improve accuracy of NLP application, then the accuracy and efficiency of processing user-generated speech is improved, but the resource intensity and time required for processing increases significantly

Engineering Contradiction:
Improveaccuracy of processing user-generated speechVSAvoidresource intensity
Core Design Contradiction:
Measurement precisionVSQuantity of substance

Solution Approach 1:

The patent extracts only the essential feature patterns from a limited tagged text corpus using the Charm algorithm, rather than requiring a complete large-scale corpus. This extraction approach identifies and utilizes only the most relevant feature combinations needed for accurate command detection, significantly reducing the quantity of tagged text data required while maintaining processing accuracy.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The system performs preliminary analysis to identify feature patterns from a small tagged text corpus before actual speech processing. By pre-extracting and storing these feature patterns, the system prepares the essential processing rules in advance, eliminating the need to process large amounts of tagged text during real-time operation and reducing overall resource intensity.

Inventive Principle:
Principle #10Preliminary action

2Measurement precision

If the size of the tagged text corpus is increased to improve accuracy of NLP application, then the accuracy and efficiency of processing user-generated speech is improved, but the time required for processing increases significantly

Engineering Contradiction:
Improveaccuracy of processing user-generated speechVSAvoidtime required for processing
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The Charm algorithm extracts only the essential feature patterns from a limited tagged text corpus, identifying the most relevant feature combinations for command detection. This selective extraction dramatically reduces the time required for processing compared to analyzing complete large-scale corpora, while maintaining the accuracy needed for reliable speech processing.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The system performs feature pattern extraction from the tagged text corpus in advance, before actual speech processing begins. By pre-computing and storing these patterns, the system eliminates the need for time-consuming corpus analysis during real-time operation, significantly reducing processing time while preserving accuracy.

Inventive Principle:
Principle #10Preliminary action

3Reliability

If traditional NLP methods are used to process user-generated speech, then comprehensive analysis is achieved, but the computational demands and resource requirements are excessively high

Engineering Contradiction:
Improvecomprehensive analysis capabilityVSAvoidcomputational demands
Core Design Contradiction:
ReliabilityVSUse of energy by moving object

Solution Approach 1:

The system extracts only the essential feature patterns needed for reliable command detection from the tagged text corpus, rather than processing complete text data. This extraction approach maintains comprehensive analysis capability by identifying the most discriminative feature combinations, while significantly reducing computational demands and energy consumption during speech processing.

Inventive Principle:
Principle #2Taking out (Extraction)

Solution Approach 2:

The system applies partial action by using only the essential feature patterns extracted from the tagged text corpus, rather than processing all available text data. This selective approach provides sufficient analysis capability for reliable command detection while minimizing computational resource requirements and energy consumption.

Inventive Principle:
Principle #16Partial or excessive action

Data Source

PatentEP3186707B1Method of and system for processing a user-generated input command
Publication Date: 2022.11.23 YANDEX EURO
  • EP3186707B1 patent drawingFigure 1
  • EP3186707B1 patent drawingFigure 2
  • EP3186707B1 patent drawingFigure 3A~3B

AI summary

A method of processing a user-generated input command executable at a computing apparatus includes, for each one of the plurality of machine-executable instructions, receiving a plurality of example expressions of the machine-executable instruction, generating a plurality of features for the plurality of example expressions, and analyzing the plurality of features associated with the plurality of example expressions to determine a plurality of feature patterns. The plurality of feature patterns is used to convert the user- generated input command into a machine executable output instruction, the machine executable output instruction being one of the plurality of machine-executable instructions. Systems for processing a user-generated command are also disclosed.