Context-Enriched Intent Extraction from Service Usage Language
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing intent extraction methods using natural language-based input information fail to consider background information, leading to inaccurate extraction due to errors in keyword identification and lack of consideration for user-specific conditions and network resources.
Innovation Solution
A data processing apparatus that supplements natural language input with user-specific service information and network resource knowledge, using models like BERT and K-BERT to accurately extract intents by considering context and resource status.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If intent is extracted using only user's natural language-based input information, then the extraction process is simple, but the accuracy of intent extraction deteriorates due to ignoring background information
Solution Approach 1:
The system performs preliminary actions by collecting and storing background information (user profiles, service usage history, network status) before the intent extraction process. This pre-prepared context is then integrated with the user's input to improve extraction accuracy without significantly increasing the complexity of the extraction process itself.
Solution Approach 2:
The system introduces an intermediary component (information supplement unit) that bridges the gap between the simple extraction process and the need for accurate intent recognition. This intermediary supplements the user's input with relevant background information, enabling accurate extraction without making the core extraction process overly complex.
2Ease of manufacture
If mapping is set after keyword extraction from input information, then the process follows a clear sequence, but the intent is not correctly extracted when there is an error in keyword extraction or mapping rule is not set
Solution Approach 1:
The system implements feedback mechanisms where the information supplement unit provides contextual information back to the extraction process. This feedback loop allows the system to correct potential errors in keyword extraction by comparing against background information and usage patterns, thereby improving reliability while maintaining the clear sequential structure of the process.
Solution Approach 2:
The system prepares cushioning measures in advance by storing comprehensive background information and multiple mapping rules. When keyword extraction encounters errors or ambiguities, these pre-prepared resources serve as a safety net, allowing the system to recover and correctly extract intent despite errors in the initial keyword extraction phase.
3Productivity
If usage conditions and resource status are not considered in intent extraction, then the extraction process is faster, but it is not possible to extract an optimal intent according to the current state
Solution Approach 1:
The system applies local quality by selectively incorporating background information and resource status only when relevant to the specific intent extraction task. Rather than processing all possible data uniformly, the information supplement unit identifies and integrates only the necessary contextual elements, maintaining extraction speed while improving adaptability to the current state.
Solution Approach 2:
The system employs partial action by supplementing the user's input with a selective subset of background information rather than processing all available data. This partial supplementation provides just enough contextual information to improve adaptability to the current state without significantly slowing down the extraction process.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
A data processing apparatus according to one embodiment has: an input unit that receives an input of information in natural language related to a status when a user who uses a service uses the service; an information supplement unit that supplements the information with information specific to the user about usage of the service; and an extraction unit that extracts a desire related to provision of the service made by the user from the information supplemented by the information supplement unit.