Virtual Assistant Intent Recognition via Relational Language Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current intelligent virtual assistants (IVAs) face challenges in accurately interpreting user intentions due to unnecessary background information in user inputs, leading to confusion and degradation of user experience, as they often fail to assess the relevance of input portions effectively.
Innovation Solution
The proposed solution involves a technique where IVAs parse user queries to identify concepts and context, mapping these to intents, and then provide responses based on the context, allowing for more relevant and human-like interactions by focusing on the essential information from the user's input.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If IVAs process all user input including background information, then more context is available for understanding, but processing time increases and accuracy decreases due to unnecessary information
Solution Approach 1:
The patent segments user input into different components (relational language vs. task-oriented language) and processes them separately. This allows the system to efficiently identify and extract only the relevant portions for intent recognition while still considering contextual information, thereby improving accuracy without proportionally increasing processing time.
Solution Approach 2:
The patent extracts and identifies relational language strategies (such as self-exposure and justification) from user input separately from the core task-oriented intent. By taking out these relational components, the system can focus processing resources on the essential information needed for accurate intent recognition, reducing unnecessary processing of background information.
2Reliability
If IVAs interpret all user input including relational strategies, then user trust and engagement improve, but intent recognition becomes confused and less accurate
Solution Approach 1:
The patent divides user input into relational language segments (which build trust) and task-oriented segments (which convey intent). By segmenting these functions, the system can maintain user trust through acknowledging relational strategies while separately and accurately processing the core intent without confusion.
Solution Approach 2:
The patent introduces an intermediary layer that identifies and separates relational language strategies from task-oriented intent. This intermediary processing layer allows the system to handle both trust-building relational content and accurate intent recognition simultaneously, preventing the confusion that would occur if all input were processed uniformly.
3Adaptability or versatility
If IVAs use human-like relational strategies in conversation, then user engagement increases, but system complexity increases
Solution Approach 1:
The patent segments the conversation processing into distinct functional areas: relational language detection and task-oriented intent recognition. This segmentation allows the system to implement human-like relational strategies without requiring complete redesign of the entire system, thereby increasing conversation naturalness while managing complexity through modular processing.
Data Source
AI summary
Virtual assistants intelligently emulate a representative of a service provider by providing variable responses to user queries received via the virtual assistants. These variable responses may take the context of a user's query into account both when identifying an intent of a user's query and when identifying an appropriate response to the user's query.