Neural Network Model for Simultaneous Sentence and Word Intent Recognition

Solution Overview

Problem

Existing intent recognition systems in human-machine dialogue interactions can only recognize either sentence-level or word-level intents, failing to recognize both simultaneously, which limits their effectiveness.

Innovation Solution

A neural network model is constructed with a feature extraction layer and a recognition layer to output intent results and scores for both sentence-level and word-level intents, enabling simultaneous recognition of both through training with annotated data and word segmentation results.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Adaptability or versatility

If only sentence-level intent recognition is implemented, then the system structure is simple, but word-level intent recognition capability is lost

Engineering Contradiction:

Improveintent recognition capabilityVSAvoidsystem structure

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent merges sentence-level intent recognition and word-level intent recognition into a single neural network model. The model integrates multiple recognition layers that simultaneously process both sentence-level and word-level intents, allowing the system to recognize both types of intents without requiring separate systems. This resolves the contradiction by combining multiple functions into one unified structure.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The neural network model is designed with multi-functionality to perform both sentence-level intent recognition and word-level intent recognition. The model includes multiple recognition layers that can process different levels of intent simultaneously, making the system universal in its capability to handle various intent recognition tasks without requiring separate specialized systems.

Inventive Principle:

Principle #6Universality (Multi-functionality)

2Adaptability or versatility

If only word-level intent recognition is implemented, then the system structure is simple, but sentence-level intent recognition capability is lost

Engineering Contradiction:

Improveintent recognition capabilityVSAvoidsystem structure

Core Design Contradiction:

Adaptability or versatilityVSDevice complexity

Solution Approach 1:

The patent merges sentence-level intent recognition and word-level intent recognition into a single neural network model. The model integrates multiple recognition layers that simultaneously process both sentence-level and word-level intents, allowing the system to recognize both types of intents without requiring separate systems. This resolves the contradiction by combining multiple functions into one unified structure.

Inventive Principle:

Principle #5Merging (Combining)

Solution Approach 2:

The neural network model is designed with multi-functionality to perform both sentence-level intent recognition and word-level intent recognition. The model includes multiple recognition layers that can process different levels of intent simultaneously, making the system universal in its capability to handle various intent recognition tasks without requiring separate specialized systems.

Inventive Principle:

Principle #6Universality (Multi-functionality)

3Measurement precision

If simultaneous sentence-level and word-level intent recognition is implemented, then intent recognition accuracy is improved, but model complexity increases

Engineering Contradiction:

Improveintent recognition accuracyVSAvoidmodel structure

Core Design Contradiction:

Measurement precisionVSDevice complexity

Solution Approach 1:

The patent segments the intent recognition task into multiple independent recognition layers within the neural network. Each layer is responsible for recognizing specific levels of intent (sentence-level or word-level), allowing the model to handle complex recognition tasks through divided functional components. This segmentation enables simultaneous multi-level recognition while maintaining manageable model complexity through modular architecture.

Inventive Principle:

Principle #1Segmentation

Solution Approach 2:

The patent introduces an additional dimension to the intent recognition model by incorporating multiple recognition layers that operate simultaneously at different levels (sentence-level and word-level). This dimensional expansion allows the model to process multiple types of intent information concurrently, improving overall recognition accuracy without requiring a complete redesign of the fundamental model structure.

Inventive Principle:

Principle #17Another dimension (Dimensionality change)

Data Source

PatentUS20230004798A1Intent recognition model training and intent recognition method and apparatus

Publication Date: 2023.01.05 BEIJING BAIDU NETCOM SCI & TECH CO LTD

AI summary

The present disclosure provides intent recognition model training and intent recognition methods and apparatuses, and relates to the field of artificial intelligence technologies. The intent recognition model training method includes: acquiring training data including a plurality of training texts and first annotation intents of the plurality of training texts; constructing a neural network model including a feature extraction layer and a first recognition layer; and training the neural network model according to word segmentation results of the plurality of training texts and the first annotation intents of the plurality of training texts to obtain an intent recognition model. The method for intent recognition includes: acquiring a to-be-recognized text; and inputting word segmentation results of the to-be-recognized text to an intent recognition model, and obtaining a first intent result and a second intent result of the to-be-recognized text according to an output result of the intent recognition model.