Dialogue Generation Model Fusion for Accuracy and Stability
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing dialogue understanding technologies face challenges in accurately understanding user intentions due to the need for setting thresholds to select between small sample learning models and deep learning models, leading to reduced accuracy with increasing sample sizes and instability in model performance.
Innovation Solution
A method that combines small sample learning models, deep learning models, and fusion models to generate dialogue information. This involves obtaining problem information, generating first and second features from these models, combining them to form a feature sequence, and inputting this sequence to a fusion model to produce relevant dialogue information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a threshold is set to select between small sample learning models and deep learning models, then model selection becomes manageable, but dialogue understanding accuracy decreases with increasing sample sizes and model performance becomes unstable
Solution Approach 1:
The patent merges small sample learning models and deep learning models into a unified dialogue understanding system. The fusion model integrates features from both model types, allowing the system to leverage the advantages of both approaches simultaneously rather than requiring threshold-based selection between them.
Solution Approach 2:
The fusion model serves multiple functions by processing features from both small sample learning and deep learning models. This multi-functional approach allows the system to handle varying sample sizes effectively without requiring separate model selection logic or threshold settings.
2Reliability
If small sample learning models are used, then performance is maintained with few samples, but accuracy deteriorates as sample sizes increase
Solution Approach 1:
The fusion model combines the output features from small sample learning models and deep learning models, allowing the system to maintain reliability with few samples while improving accuracy with increasing samples through the complementary information provided by both model types.
Solution Approach 2:
The system dynamically adjusts the contribution of different model features through the fusion model based on the available sample size, allowing optimal performance across varying data conditions without requiring manual threshold settings.
3Measurement precision
If deep learning models are used, then accuracy improves with more samples, but performance becomes unstable with few samples
Solution Approach 1:
The fusion model integrates features from deep learning models with those from small sample learning models, providing stability with few samples through the latter while capturing the accuracy improvements from the former when more samples are available.
Solution Approach 2:
The fusion model acts as an intermediary that mediates between the deep learning model features and small sample learning model features, balancing their contributions to achieve stable and accurate performance across different sample sizes.
Data Source
AI summary
The present application discloses a method and an apparatus for generating a dialogue, and a storage medium. The method includes: obtaining problem information; inputting the problem information to a small sample learning model to generate a first feature; inputting the problem information to a deep learning (DL) model to generate a second feature; combining the first feature and the second feature to generate a feature sequence; and inputting the feature sequence to a fusion model to generate dialogue information corresponding to the problem information.